diff --git a/.dockerignore b/.dockerignore new file mode 100644 index 0000000..d15bb58 --- /dev/null +++ b/.dockerignore @@ -0,0 +1,33 @@ +# Keep the build context small and the image reproducible. Anything the build +# needs but that is gitignored (web/dist) is produced inside the image instead. + +.git +.gitignore +.env + +# Built by stage 1 — never copy a stale local build into the image. +web/dist +web/node_modules + +# Local runtime state: the live database, images and TTS cache must never end +# up baked into an image layer. +data/ +*.db +*.db-shm +*.db-wal +backups/ + +# Local build outputs +/petal +*.test +*.out +*.log + +# Docs and tooling that don't affect the binary +*.md +!web/**/*.md +deploy/ +scripts/ +.vscode/ +.idea/ +.DS_Store diff --git a/.env.example b/.env.example index 75a22d5..836d41f 100644 --- a/.env.example +++ b/.env.example @@ -10,6 +10,12 @@ DATABASE_PATH=./data/petal.db # On-disk store for images pasted/dropped/inserted in the editor IMAGE_DIR=./data/images +# DreamDict's built dictionary (French, European Portuguese, Spanish, Mandarin), +# opened read-only beside petal.db. Optional: with no file here, word lookups use +# the embedded English/Chinese datasets, which is how a laptop checkout runs. +# Build one with `go run ./cmd/dictimport` in the dreamdict repo. +DICT_PATH=./data/dict.db + # LLM LLM_BACKEND=vllm # vllm | ollama LLM_ENDPOINT=http://localhost:8000 # vLLM :8000, Ollama :11434 @@ -27,17 +33,29 @@ TTS_ENDPOINT= # e.g. http://127.0.0.1:5005 — empty disable TTS_ENDPOINT_ZH= # e.g. http://127.0.0.1:5006 — Chinese Piper instance TTS_VOICE_EN=en_US-amy-medium # Piper voice id for English TTS_VOICE_ZH=zh_CN-huayan-medium # Piper voice id for Chinese +TTS_PATH=/ # path Piper serves synthesis on: "/" up to piper-tts 1.5, "/synthesize" from 1.6.0 TTS_CACHE_DIR=./data/tts # on-disk store for synthesized clips (content-addressed) TTS_TIMEOUT=15s TTS_AUDIO_FORMAT=mp3 # mp3 | opus | wav — mp3/opus transcode Piper's WAV via ffmpeg -# --- Deferred (not wired in the local-dev build) --- - -# Auth (Authentik OIDC) — deferred; single hardcoded local user for now -# SESSION_SECRET=change-me-to-random-64-char-string -# AUTHENTIK_URL=https://auth.parodia.dev +# --- Auth (Authentik OIDC) --- +# +# Login turns on only when the issuer, client id and secret are all set. Leave +# them commented out for local development and Petal runs as the single +# hardcoded `local` user, exactly as it did before auth landed. +# +# AUTHENTIK_URL is the issuer of the Petal provider in Authentik (the value of +# its "OpenID Configuration Issuer" field). The redirect URI to register there +# is BASE_URL + /auth/callback. +# AUTHENTIK_URL=https://auth.parodia.dev/application/o/petal/ # AUTHENTIK_CLIENT_ID=petal # AUTHENTIK_CLIENT_SECRET= +# +# Who may sign in: comma-separated OIDC subject ids and/or email addresses. +# Empty = anyone Authentik authenticates. +# PETAL_ALLOWED_SUBS=her@example.com,me@example.com + +# --- Deferred (not wired in the local-dev build) --- # Copyleaks (Tier-2 plagiarism) — deferred; needs a public webhook # COPYLEAKS_ENABLED=false diff --git a/.gitignore b/.gitignore index f4a6dc9..bcb879a 100644 --- a/.gitignore +++ b/.gitignore @@ -10,6 +10,10 @@ web/dist/* !web/dist/.gitkeep *.log +# Python +__pycache__/ +*.pyc + # Local env & data .env *.db diff --git a/BUILD_PLAN.md b/BUILD_PLAN.md index 3c27394..99d0ab1 100644 --- a/BUILD_PLAN.md +++ b/BUILD_PLAN.md @@ -135,9 +135,209 @@ Multi-session build. **Source of truth for what's done and what's next.** Update - [x] Verified: tsc clean, vite build OK, companion vitest 45/45. **Real-browser screenshots** (local Playwright + Chromium, clock mocked to 23:30): day = warm cream + pink sakura petals; night = dark plum-indigo + twinkling stars + glowing sleepy kitten. Both pretty (acceptance criterion). ### Deferred (post-v1-local) -- [ ] Authentik OIDC auth + session middleware ← **on hold: user doing foundational work first** -- [ ] Copyleaks Tier-2 + webhook HMAC -- [ ] Dockerfile, docker-compose, Traefik, deploy to write.parodia.dev +- [x] **Multi-user groundwork** (2026-07-26) — request-scoped identity. New `internal/auth`: `Middleware(Resolver)` resolves the caller once per API request and stores the id in the context; handlers read it via `auth.UserID(r.Context())` instead of naming `db.LocalUserID`. `StaticResolver(db.LocalUserID)` keeps Petal single-user today. **Auth itself is still deferred** — but every query is now scoped to whoever the resolver says is calling, so landing Authentik is a one-line change in `main.go` plus a `Resolver` implementation. +- [ ] Copyleaks Tier-2 + webhook HMAC (still parked — needs a public webhook; revisit after Phase 15) +- Authentik OIDC + deploy: **no longer deferred — expanded into Phases 15–17 below** (decisions ratified 2026-07-26; see `MULTIUSER_PLAN.md`). Deploy landed 2026-07-26 (Phase 15); Authentik itself already runs on the same VPS, so Phase 16 has its IdP waiting. + +## Execution phases 15–22 (added 2026-07-26) + +Decisions behind these are ratified in `MULTIUSER_PLAN.md` (all OPENs settled) and `SUGGESTIONS.md` (the *why*; Q1–Q3 settled). **Standing rules for every phase below:** +- **Isolation tests in the same commit** as any new user-scoped endpoint (the `docs/isolation_test.go` suites are the template — this discipline caught a real unscoped-query bug once already). +- **LLM-minimalism** (SUGGESTIONS §6): the LLM never gates essential functionality; new essential features are code+data first. +- **Aesthetic + bilingual-in-the-pair copy remain acceptance criteria** on every user-visible change. +- Verify per project convention: go build/vet/test, tsc, vite build, vitest, live smoke on a throwaway DB/port. + +### Phase 15 — Deploy plumbing (parodia.dev + headscale) ✅ (2026-07-26/27) +Petal hosted on the parodia.dev VPS; vLLM stays on millenia over headscale. Auth (Phase 16) needs the stable `BASE_URL`/redirect URI this phase creates. **Hostname: `petal.parodia.dev`** (DNS already pointed at the VPS). Runbook: `deploy/README.md`. +- [x] Dockerfile (multi-stage: `npm run build` → `go build` → alpine runtime) + docker-compose. CGO stays off (modernc SQLite is pure Go), so the runtime layer exists only for **ffmpeg** (read-aloud transcode) and **tzdata** (the bedtime nag + night mode read the local clock). Non-root; `/data` is the single writable mount. **`.dockerignore`** keeps the live DB and a stale local `web/dist` out of the image. +- [x] Traefik route + HTTPS on `petal.parodia.dev` — labels follow the host's existing convention (external `traefik` network, `web-secure` entrypoint, `default` cert resolver, `compression@file`) plus Petal's own header middleware. No host port is published; Traefik is the only way in. `BASE_URL=https://petal.parodia.dev` set for Phase 16's redirect URI. +- [x] `LLM_ENDPOINT` → millenia's headscale address (`100.64.0.2:8000`); `LLM_TIMEOUT` **30s → 90s** for the WAN+VPN round trip (the voice/collocation passes send a whole document and the timeout is a hard deadline on `Complete`). **Exposed with a forwarder, not a rebind** (`deploy/vllm-headscale-proxy.service`, socat): `vllm-chat.service` is shared — Petal, **Gogobee** and Open WebUI all point at `127.0.0.1:8000`, and Open WebUI keeps its endpoint in its own database rather than in env, so rebinding meant editing three consumers and reloading a 35B AWQ model. The forwarder adds a second listener on `100.64.0.2` only (never `0.0.0.0` — the far end is a public host), zero downtime, zero consumer changes. Model is `qwen3.6-35b`. **Verified end to end: a grammar checkpoint from `petal.parodia.dev` returns real suggestions in ~3s over the VPN.** +- [x] TTS — **deviation from the plan, deliberate**: Piper was *not* actually installed on parodia, and the `reala` account has no lingering session to keep user systemd units alive. Runs as **two sibling containers** (`piper-en`, `piper-zh`) off one image, models cached in a shared volume, on an internal network with no published ports. pt-PT in Phase 21 is a fourth service, not a new image. **Found + fixed while wiring**: piper-tts 1.6.0 moved synthesis from `POST /` to `POST /synthesize` (identical body); rather than pin both deployments to one release, the path is now config (`TTS_PATH`, default `/` so millenia is untouched). +- [x] Backups — `db.Backup` uses **`VACUUM INTO`**, not a file copy: in WAL mode the newest committed pages may live in `petal.db-wal`, and copying the three files separately can capture a torn mid-checkpoint state. `VACUUM INTO` reads one coherent snapshot including the WAL, takes no write lock (safe against the live app), and emits a single file with no companions; it refuses an existing destination so a failed run can't destroy the last good backup. Driven by a `-backup` flag on the binary. + - **On the VPS: folded into the host's existing `parodia-backup`** (age-encrypted, offsite to S3, 14-day retention, dead-man snitch) rather than a parallel cron — the user pointed out that layer already existed. **Found a real bug while doing it:** that script's `sqlite_dump` helper uses Python `iterdump`, which **does not reproduce an FTS5 virtual table** — it emits `documents_fts` as a raw `sqlite_master` row plus shadow tables, and replaying the result dies with `no such table`. Cross-document search would have been silently missing after any restore. Added a `sqlite_file_dump` helper using `VACUUM INTO` instead; round-trip verified (counts + a live FTS `MATCH`). + - **On millenia: `petal-backup.timer`** — the canonical instance had **no scheduled backup at all** (newest snapshot a month old), which mattered far more than the staging one. Nightly 03:20, `Persistent=true` (the box isn't on 24/7), snapshot → gzip → **age-encrypt with the parodia public recipient** → push to the VPS over headscale with a size check → prune both ends. Verified the pushed archive is real age ciphertext and that neither box can decrypt it. +- [x] Migration decision: **millenia stays canonical** (user's call). The VPS runs an empty staging DB so she moves accounts exactly once, when Phase 16/17 land. +- [x] **Encryption at rest** (not in the original plan — the user raised it mid-session, correctly). VPS data dir is now a **LUKS2 volume** (`deploy/setup-encrypted-data.sh`), covering `petal.db`, `images/` **and the TTS cache** (synthesized audio of her sentences). LUKS-on-a-file rather than gocryptfs because SQLite in WAL mode needs a shared-memory index mapped consistently across processes and FUSE has a long history of mmap/locking differences. Key on the same box — a deliberate availability tradeoff, documented honestly: it stops a decommissioned disk or a raw block-device read, **not** anyone holding the whole VM image. Canary-verified: a marker written through the app is absent from the raw image and present through the mount. **Two bugs caught by rehearsing a reboot rather than trusting the clean run** — (1) mounting over a directory *hides* its contents rather than removing them, so the first pass left the original plaintext `petal.db` and WAL on the unencrypted root filesystem, invisible under the mount (now shredded pre-mount, with a refusal if the mountpoint won't come up empty); (2) **`systemd-cryptsetup` wasn't installed**, so `/etc/crypttab` was ignored entirely and the volume would never have unlocked at boot. Added a **mount-liveness guard** (`.volume-ok` bind-mounted with `create_host_path: false`) so an unmounted volume is a loud container start failure instead of Petal quietly serving a blank database. ⚠️ A true reboot is untested — the VPS also runs matrix/lemmy/akkoma/gitea/authentik, so that's the user's call. millenia remains unencrypted at rest (LVM, no LUKS). +- [x] **Supervision** (also not in the original plan). millenia's Petal had been running as a bare `./petal` with **PPID 1** — no unit, no screen session — so a crash or reboot left it silently down; now `petal.service`, verified by `kill -9`. **Piper's silent-failure mode fixed**: with `RestartSec=3` against systemd's default 10s window the burst limit was never reached, so a dead service looped **26,800+ times over a day without entering `failed`**; both units now set `StartLimitIntervalSec=300`/`StartLimitBurst=5`. Still missing: an external probe (uptime-kuma monitors on `/api/health` and `/api/tts` — needs the UI, written up in `deploy/README.md` §7). +- [x] **Interim edge gate** (not in the original plan; added once the instance was live). Petal authenticates nobody yet — `StaticResolver` hands every request the same `local` user — so on a public host the whole API was open to read/write and image upload. Traefik basic auth holds the door until Phase 16, with `/api/health` exempt on its own higher-priority router. Deleted when OIDC lands. +- [x] Acceptance — verified over public HTTPS **with the LLM link down** (it genuinely is): Hunspell dictionaries 200, gloss + word lookup (incl. phonetic) 200, doc create/save, FTS search on 春天, md + docx export, vocab capture/list, read-aloud EN + zh (real mp3 via ffmpeg, cache hit on repeat, 404 for an unconfigured language so the client falls back). `POST /check` → the warm 502 that renders as 小助手在休息. `/api/health` public; HTTP 301 → HTTPS with a valid cert. + +### Phase 16 — Auth (in-app OIDC) + image-store ownership ✅ (2026-07-27) — live on petal.parodia.dev +Option B ratified. `go-oidc` + `x/oauth2`; config fields already existed. The `Resolver` seam from Phase 0 was the only integration point — no handler or query moved. +- [x] OIDC login flow (`internal/auth/oidc.go`): `/auth/login` → Authentik → `/auth/callback` → provision → session. **state** (cookie vs param, constant-time) + **nonce** (ID-token claim vs cookie, so a token minted for another attempt is refused) + **PKCE S256**. Discovery is **lazy and retried**: an Authentik outage blocks new logins but leaves every existing session working, since those need only Petal's own DB — the app must not fail to boot because the IdP is briefly down. +- [x] `sessions` table (migration `0010`); opaque token in `petal_session` (`HttpOnly`, `SameSite=Lax`, `Secure` only when `BASE_URL` is https — flagging it on a plain-http dev server makes the browser silently drop it). **The table stores only the token's SHA-256**, so a DB copy yields no usable session. **30-day sliding expiry**, all time math in SQLite `datetime()` (canonical UTC), the extension throttled to one write per hour per session. `/auth/logout` deletes the row, not just the cookie; `RevokeAll` signs one writer out everywhere; expired rows pruned at startup. +- [x] Allowlist: `PETAL_ALLOWED_SUBS`, comma-separated. **Matches a subject id *or* an email**, case-insensitively — a deliberate widening of the plan: a subject is an opaque uuid that doesn't exist until first login, so a subject-only list means letting someone in, reading a log line, and editing config. Empty = anyone Authentik authenticates. A rejected valid login gets the warm bilingual "这个 Petal 不是给你写的 · This Petal isn't yours to write in" page and **no provisioned account**. +- [x] `main.go` picks the resolver from config: `SessionStore` (which is itself the `Resolver`) when `AUTHENTIK_URL`/id/secret are all set, `StaticResolver(local)` otherwise — so local dev and every pre-auth deployment behave exactly as before. `/auth/*` mounts on the root router, outside `/api`. +- [x] Frontend: one 401 interceptor in `api/client.ts` (`UnauthorizedError` + an `onUnauthorized` hook covering `req`, the image upload and the SSE chat stream) → `useSession` → `SignInOverlay` (warm bilingual, editor still visible behind it — nothing has been taken away). **Draft rescue** (`lib/drafts.ts`): a save that 401s stashes its body in `localStorage` keyed by doc id *before* anything else, auto-save then stops (further attempts would only 401 and re-stash), and opening that doc after re-login merges it back and schedules a save. StatusBar says 已保存在本机 · Kept on this device — where the writing is, not what failed. Sidebar footer gains the account + 退出 · Sign out (hidden when the id is still `local`). +- [x] **Image store ownership** (OPEN #5): `images` table (`PRIMARY KEY (name, user_id)`) — **one row per owner, not one owner per file**, so the same picture uploaded by two people is still stored once and dedup survives; the file is deleted only with its last row. Fetch joins on the caller and answers **404, not 403** (whether a hash exists is itself information). `Cache-Control` went `public` → `private` — a shared cache must never hand one writer's image to another. Files already on disk are claimed for the local user at startup (idempotent), because a row is now what makes an image fetchable and every picture already pasted into a document would otherwise 404. +- [x] `users.pair_lang` (default `'zh'`) added in the same migration; login refreshes email/display name but never touches it — it's Petal's setting, not the IdP's. +- [x] Tests: `session_test.go` (lifecycle, expiry + prune, sliding renewal, hash-not-token storage, cross-user non-interchangeability, `RevokeAll`, FK cascade, middleware wiring, user upsert, allowlist matrix) and `oidc_test.go` — **the whole round trip against a stub IdP** (RSA-signed ID tokens, real discovery + JWKS): PKCE challenge present, verifier reaches the token endpoint, state mismatch → 400, **replayed nonce from another attempt → 400**, allowlist refusal → the bilingual 403 with no account created, provider error → 403, already-signed-in login short-circuits home. `images/handler_test.go` gained two-user isolation, cross-user dedup + last-owner file deletion, and backfill idempotency. `web/src/lib/drafts.test.ts` covers the rescue (round-trip, per-doc, take-consumes, expiry, corrupted entry, storage that throws). +- **A real bug the round-trip test caught:** the one-shot state/nonce/PKCE cookies were cleared with `defer o.clearTemp(w)` — which runs *after* the redirect has written the response header, so the `Set-Cookie` was silently dropped and they lingered in the browser for their full 10 minutes. Now cleared up front. +- Verified: go build/vet/test, tsc, vite build, vitest 76/76 all clean. Migration `0010` applied to **a copy of the live millenia DB** (`VACUUM INTO` snapshot): 10 migrations apply, documents/vocab/versions counts unchanged, FTS search still returns hits, the one existing image claimed, `pair_lang` defaulted. Live smoke against the binary on a throwaway DB: auth-off → `/api/me` is `local` and everything 200s; auth-on → `/api/docs` and `/api/me` 401, `/api/health` still public, `/auth/login` with an unreachable IdP renders the warm 503 page, `/auth/logout` redirects home; a hand-inserted session row → 200 with the cookie, 401 without it, with a bad one, and once expired. +- [x] **Deployed** (user: "do it! register it!"). Provider + application registered in Authentik (slug `petal`, confidential, strict redirect `https://petal.parodia.dev/auth/callback`, openid/profile/email, implicit-consent authorization flow), created through `ak shell` since the available API token is a limited invite-minter bot. `.env` filled in on the VPS, image rebuilt, container recreated. **The Traefik basic-auth gate is gone** along with the separate unauthenticated `/api/health` router that existed only to escape it — Petal 401s every `/api` route without a session, so an anonymous visitor gets the app shell and a redirect, and a second password in front of a real login is one more thing to lose. Verified over public HTTPS: `/api/health` 200, `/api/docs` **401 with no basic-auth challenge**, `/auth/login` → Authentik with state+nonce+PKCE in the URL, following it lands on the real sign-in page, `/petal.svg` served as the favicon. The final step — typing her password — is hers. +- **Two bugs deploying caught that the whole test suite could not**, both fatal before the login page ever renders: (1) the code trimmed the issuer's **trailing slash**, and Authentik's issuer has one — OIDC requires a byte-for-byte match, so discovery failed every time while the stub IdP (which advertised a slashless issuer) kept passing. Fixed, and the stub's issuer is now a knob with a regression test that ends in a slash. (2) A provider created through `ak shell` rather than the admin UI comes up with **`grant_types = []`**, which authentik reads as "no grant type is permitted here" and answers with `invalid_request` / *The request is otherwise malformed*. Both are written up in `deploy/README.md` §4. +- **Allowlist is currently `prosolis@proton.me` only.** That Authentik instance fronts ~40 accounts across several applications, so an empty list was not an option, and guessing which account is hers would either lock her out or let a stranger in. Adding her is one line in `.env` plus a restart. Note that an Authentik account with **no email set** (e.g. `akadmin`) can't match an email-based entry — use its subject id. + +### Phase 17 — Migrate the `local` user ✅ (2026-07-27) — her writing now lives on her account +Script, app stopped, backup first (OPEN #4). **The "she logs in once first" dependency turned out not to exist**: authentik's default `hashed_user_id` sub mode makes the subject `User.uid`, which is derived from her user id and the instance secret — stable, and readable before she has ever signed in (`ak shell -c "…User.objects.get(username='claire').uid"`). So the data can move *first*, and she signs in to find her writing already there rather than to an empty Petal that fills in later. +- [x] `scripts/migrate_local_user.*`: single transaction, `PRAGMA foreign_keys=OFF`, re-point `documents`/`tags`/`vocab_words` **and `images`** (versions/suggestions follow parents; `images` is new in Phase 16 and carries `user_id` directly — miss it and every pasted picture 404s), delete the empty provisioned row, verify row counts before commit; refuses to run if the app is up or the target has data +- [x] `scripts/migrate_local_user.py` — dry-run by default, `VACUUM INTO` backup before touching anything, one transaction with `PRAGMA foreign_keys=OFF`, re-points `documents`/`tags`/`vocab_words`/`images`, deletes the old user row, and **verifies every expected row actually moved (and that the source is left owning nothing) before it commits**, rolling back otherwise. Refuses to merge into an account that already owns writing. Runbook in the script header. +- **The "is the app stopped?" guard needed a second attempt.** `BEGIN EXCLUSIVE` — the obvious check — passes straight through against a *running but idle* Petal, because in WAL mode it only conflicts with another writer. That is exactly the case the guard exists to catch, and it would have failed silently. `PRAGMA locking_mode = EXCLUSIVE` conflicts with any connection at all, since it locks the shared-memory index every WAL reader maps; verified against a live server. +- [x] **Run for real.** Her 8 documents, 33 snapshots, 103 suggestions, 3 vocabulary words and 1 image moved from `local` onto `5f47d955…` (Claire, `clairew8@pm.me`). Sequence: `-backup` snapshot of millenia's live DB (taken while it kept running — `VACUUM INTO` needs no write lock), shipped to the VPS, installed over the throwaway staging database (kept as `petal.db.staging-*`), **started once so migration `0010` applied**, stopped, migrated, started. Verified over public HTTPS with a short-lived probe session, then removed: `/api/me` is her, 8 documents listed, her image 200s, vocabulary garden and version history intact, search returns hits — and 401 without the cookie. +- **millenia is a frozen fallback, not a mirror** (user's call: "both, VPS first"). It was left running and completely untouched, still serving the same writing under the pre-auth `local` user. The two diverge the moment anything is written on either, so it wants retiring rather than syncing. +- **A second bug in the guard, found by running it against production rather than a test file.** `PRAGMA locking_mode = EXCLUSIVE` keeps holding the lock after being set back to `NORMAL` — SQLite only releases it on that connection's next database access — so on a **WAL** database the script locked itself out of its own `VACUUM INTO` backup. It passed locally because the test database had come out of `VACUUM INTO` and so wasn't in WAL mode at all — the same shape of miss as the trailing-slash issuer: the fixture didn't look like production. The probe now runs on its own connection and closes it, and the fix was re-verified against a database that had genuinely been served in WAL mode. +- **Startup crash averted while sequencing this**: the image backfill claims unowned files for `local`, which stops existing after the migration — a foreign-key error inside `images.New`, which `main.go` treats as fatal. Petal would have entered a crash loop the first time it started on a migrated database. The backfill now skips a missing owner (there is nothing to claim in that case anyway; the migration moves the image rows itself). + +### Phase 18 — Per-user, per-language client state ✅ (2026-07-27) +- [x] **Preferences namespaced by account** (`web/src/lib/prefs.ts`) — `petal.sound`, `petal.petals` and `petal.companion` now read/write `.u.`. The wrinkle is timing: `sounds.ts` and `petals.ts` read their value at *import* time, long before `/api/me` answers, so rather than block startup on the network for a mute flag, a read before the answer sees the **legacy un-namespaced key** (on a single-writer browser, exactly the right value) and `setPrefsScope` — called from `useSession` the moment `/api/me` resolves — adopts it and fires `onPrefsScopeChange` so each module re-reads. `PetalCompanion` re-reads too, unless she's already swapped mascots in the meantime. +- [x] **Legacy adoption is a move, not a copy** (user's call): the first account to sign in on a browser inherits whatever was set back when Petal had no accounts, and the key is then deleted so the *second* account starts from Petal's defaults rather than from a stranger's choices. An existing scoped value is never overwritten by the legacy one. +- [x] **Personal spell dictionary promoted to a server table** (the plan's nice-to-have; user chose it) — new `internal/spell` package + migration `0011_personal_dictionary`: `personal_words (user_id, lang, word, created_at)`, `PRIMARY KEY (user_id, lang, word)`, cascading with the account. `GET/POST/DELETE /api/spell/words`; adds are idempotent, a delete of a word that was never there is a success (the caller's intent already holds), and **every response carries the full resulting list** so the client never has to merge two views of one set. Namespacing it in `localStorage` instead would have *fragmented* the list she already has across her laptop and tablet — strictly worse than before; a table means it follows her. +- [x] `lang` is the **dictionary's** language, not the writer's — an en-US personal word must not silence a pt-PT flag once the second pair ships. Normalised (`trim`/lowercase, default `en`) so `EN`/`en`/absent can't split one list into three. +- [x] `useSpellChecker` is server-backed: nspell loads, then the word list arrives from her account and is replayed in. A browser still holding the Phase-7 `petal.spell.personal` key hands it over on first load — but **only lets go of it once the server has accepted it**, so a failed request costs nothing. `addWord` takes effect in the editor immediately and persists in the background: the underline goes away the instant she asks, whatever the network is doing. A word list that fails to load costs correct words being flagged, never writing. +- [x] Tests: `internal/spell/handlers_test.go` — lifecycle (idempotent add, bulk add, repeat delete, `[]` not `null`), languages-don't-merge incl. the case-normalisation case, junk rejection, and the standing-rule **two-user isolation** (mount twice behind two resolvers over one DB: Bob sees none of Alice's words, his identical word is his own row, his delete doesn't reach hers, deleting the account takes the dictionary with it). `web/src/lib/prefs.test.ts` — legacy adoption, move-not-copy, two accounts on one browser, never-overwrite, listener fires once per real change, storage-throws safety. +- Verified: go build/vet/test, tsc, vite build, vitest 82/82 all clean; live smoke on a throwaway DB (:8073) — add/bulk-add/list/delete, pt-PT list independent of en, CJK word accepted, 400 on empty. +- [x] **Rehearsed, then deployed** (2026-07-27). The rehearsal the previous session couldn't run: `VACUUM INTO` snapshot of the live VPS database, pulled down, migrated by the Phase-18 binary — 11 migrations apply, every count unchanged (2 users, 8 documents, 33 versions, 103 suggestions, 3 vocab words, 1 image), FTS still matching, `integrity_check` and `foreign_key_check` both clean, `personal_words` present and empty. Then the deploy: off-box encrypted backup first, `git pull` + `docker compose up -d --build`, all three containers healthy, `0011` applied to the live DB with her writing untouched. Verified over public HTTPS: `/api/health` 200, `/api/docs` and the new `/api/spell/words` **401 without a session**. +- ⚠️ The authenticated live probe of `/api/spell/words` (hand-inserted session row, as Phases 16/17 used) was **blocked by this session's permission classifier** — minting a session token reads as credential fabrication. Not worked around. The endpoint's full lifecycle is covered by `internal/spell/handlers_test.go` and was smoke-tested end to end on a throwaway DB when it was built; what remains unproven in production is only that it answers 200 for a real cookie, which the shared middleware already governs for every other route. + +### Phase 19 — Langpack extraction (the copy chore) ✅ (2026-07-27) +Pure refactor, zero visible change; prerequisite for every new pair (SUGGESTIONS §2, Q2 settled). +- [x] Every `中文 · English` string from the ~29 frontend files (plus `tips.ts`, `prose.ts`, `companions.ts`, `stats.ts`) now lives in `web/src/i18n`: `types.ts` (the `Pack` shape), `packs/zh.ts` (today's copy, **verbatim** — sentinel assertions in `i18n.test.ts` guard against a quiet rewording), `index.ts` (the accessor). +- [x] Two access paths, matching where copy is built: `usePack()` for components (a `useSyncExternalStore` subscription, so a pack arriving after first paint re-renders), and `pack()` for the modules that compose a line when something *happens* rather than when something renders — the companion and the prose checker read it at call time, never at import time. +- [x] **Anything with a value in it is a function on the pack**, not a template assembled at the call site (`reviewDue(n)`, `daysAgo(n)`, `duplicateTitle(title)`, every prose rule). Word order isn't universal; a pack author must be able to move the number. English pluralisation moved into the pack with it. +- [x] `Line` renamed its Mandarin half `zh` → `native` throughout (companion bubbles, tone/style pills, history badges, stat rows). `gradeBand` now returns a band *name* rather than a label, and the roster constants (`TONES`, `REWRITE_STYLES`, export formats, companions) keep only value + emoji — the label is a pack lookup keyed by the same value, with a test asserting no roster entry is unlabelled. +- [x] `internal/llm/lang.go`: the three prompts that *name* the writer's language — the collocation gloss, Ask Petal's "answer in her language", the explanation translator — take a `Lang` instead of saying "Simplified Chinese" outright. pt-PT is spelled **"European Portuguese (pt-PT, never Brazilian Portuguese)"** in the prompt itself, since a model that has read far more pt-BR needs telling. `Why` carries her word for "why" (为什么 / porquê / …) so the tutor prompt still recognises the question. An unknown code falls back rather than erroring — a prompt is the wrong place to discover a config problem. +- [x] Wired to `users.pair_lang` on both sides: `useSession` calls `setPackLang` the moment `/api/me` answers, and each LLM handler reads the column **in the row-scoped query it already ran** (the one that proves she owns the document) rather than in a second lookup that could disagree with it. +- Tests: `internal/llm/lang_test.go` (fallback matrix; each prompt names the writer's language and *not* Chinese; the zh pair reads exactly as before), `internal/suggestions/pairlang_test.go` (the column reaches the model for collocation + translate, zh unchanged — **verified to fail when the join is removed**), `web/src/i18n/i18n.test.ts` (default before `/api/me`, fallback for an unshipped pair, no spurious notifications, verbatim sentinels, interpolation incl. plurals, no empty string anywhere in the pack, every companion/tone/style labelled). +- Verified: go build/vet/test, tsc, vite build, vitest 90/90 clean; live smoke on a throwaway DB (:8074) — doc create/save, search, md export, spell add, warm 502 from the collocation pass with the LLM down; the built bundle still carries the zh strings. +- [x] **Deployed 2026-07-27**, together with Phase 20. Pure refactor with no migration, so it was a rebuild; the langpack is live and reads exactly as before, which is the whole point of a verbatim `zh` pack. + +### Phase 20 — DreamDict as a lexicon provider ✅ (2026-07-27) +Option 3 ratified (import package, read-only `dict.db`). +- [x] **The prerequisite was bigger than the plan thought.** Renaming the module was necessary but not sufficient: DreamDict's query layer lived in `internal/dictionary`, which no other module may import whatever the module is called. Both fixed upstream in one commit — `module github.com/prosolis/dreamdict`, `internal/dictionary` → `dictionary`, with a package comment saying why reading a built database is public API while building one stays internal. `internal/loader` is untouched, and DreamDict's own tests pass unchanged. +- [x] **Provider seam** (`internal/lexicon/provider.go`): a `Provider` is the two questions the popover and the tooltip have always asked (`Lookup`, `Gloss`), which the embedded `*Lexicon` already satisfied unmodified. `Set.For(lang)` is the single place the choice is made. `OpenDreamDict` opens `dict.db` read-only beside `petal.db`; **a missing file returns `(nil, nil)`, not an error** — a laptop checkout has never had one — while a file that is *present but unimported* does error, because that one is somebody's half-finished deploy. +- [x] **The absent-dictionary case degrades better than "no data".** A pt-PT writer with no `dict.db` falls back to the embedded datasets **with the gloss suppressed** (`glossless`), so she keeps English definitions, synonyms and phonetics — all compiled into the binary and all correct for her — and loses only the translation. Handing her the Chinese gloss would be worse than handing her nothing: empty reads as "not found", wrong-language reads as Petal being broken. +- [x] pt-PT + fr + **es** wired to DreamDict; **zh stays on ECDICT**, and the routing test is the guard on that decision. The comparison the plan asked for was run against the real 452 MB database: DreamDict reaches a Chinese gloss for **53%** of the 2,000 commonest English words, against ECDICT's essentially total coverage of them. Quality did not hold, so nothing converged. es routes to DreamDict from day one and simply finds no rows in the April build — which is the same code path as any unglossed word. +- [x] **The plan's central assumption was wrong, and measuring it is what found that.** `Gloss ← Translate(word, "en", L1)` was mapped 1:1 in `MULTIUSER_PLAN.md`; against real data that table answers for **17%** of common English words into pt-PT (16% into fr). Wiktionary's translation sections are thin in the en→X direction. Going through shared Princeton WordNet synset ids instead answers for **61%**, and it is where the words a learner wants live — "ephemeral", "think" and "quickly" have no en→pt-PT translation row at all. New `dictionary.Equivalents(word, from, to)` upstream does that, falling back to the translations table, for **62%** combined. A gloss absent five times in six is not a gloss. +- [x] **Ranking, argued from a wrong answer.** Ordering equivalents by target-word frequency glosses "think" as *lembrar* — "remember" — because lembrar is the commoner Portuguese word even though pensar shares six of think's synsets to lembrar's one. Counting shared senses first, frequency second, asks which candidate means the same thing *most often*: think → pensar; achar; lembrar, write → escrever, garden → jardim, house → casa before firma. +- [x] The de-inflection walk (`candidates`) is shared with the embedded path, because `dict.db` stores headwords — "running" has no row. The first candidate that *has definitions* becomes the headword every other field is read from, so one popover never mixes "running"'s frequency with "run"'s senses. The gloss walks separately, since a word can have an equivalent and no definition. +- [x] **Surfaced where cheap**: a band chip beside the phonetic (`wordband.ts`) and an etymology line at the foot of the card. Following Phase 19, `wordBand` returns a band *name* and the langpack owns the wording. **Three bands, not five** — the difficulty score is a heuristic over length and corpus counts, good enough to separate "everyday" from "you will need to explain this" and not good enough to rank *obfuscate* against *serendipity*; a finer scale would be a confident-looking lie. Thresholds come from the real distribution (136k headwords bunch between 0.45 and 0.60; the words a writer reaches for sit under 0.40). An unscored word renders **no chip at all**. +- [x] The pair language is read **per request** in `providerFor` — a word lookup has no row-scoped query to piggyback on, unlike Phase 19's handlers — and a failed read falls back to today's embedded behaviour rather than failing the lookup. `Cache-Control` dropped from `public` to `private`: the same URL now answers in a different language per writer. +- Tests: `internal/lexicon/dreamdict_test.go` — fixture is a **real dict.db on disk**, so open/stat/seeded is the production path; missing vs. unseeded vs. unreadable, every field filled, gloss follows the writer not the word (pt-PT/fr/es/de), the synset path, de-inflection carrying *all* fields to one headword, miss-is-not-an-error, the NULL-difficulty sentinel, IPA chosen over CMU, and the handler tests (two writers/one URL/two languages, unknown caller, private caching). Upstream: `Equivalents` ordering, synset-over-translation, fallback. Frontend: `wordband.test.ts` (bands pinned to real scores; difficulty 0.0 is a score, not a missing value) and an i18n assertion that no band can be unlabelled. +- **Two bugs the tests found before the browser did**: `trimEtymology` sliced by byte, which would put invalid UTF-8 in the JSON for exactly the etymologies that matter (ἐφήμερος, ephemerus), and its ellipsis path overran its own cap. +- Verified: go build/vet/test, tsc, vite build, vitest 96/96 clean, both repos. Live smoke on a throwaway DB (:8075) against the real 452 MB `dict.db` — startup logs the languages it actually got, zh unchanged, then the same instance flipped to pt-PT and re-queried. +- [x] Deploy documented (`deploy/README.md` §4b, `DICT_PATH` through Dockerfile/compose/.env.example): `dict.db` ships into the data dir, stays out of the backups because they name `petal.db` explicitly, and is rebuildable from public data. +- [x] **Deployed 2026-07-27**, with Phase 19. The two dreamdict commits went to GitHub (`prosolis/dreamdict` main), the `replace` came out for a real pseudo-version, and Petal shipped as a rebuild — no migration in either phase, so `schema_migrations` is still 11. Order: off-box encrypted backup first (`✓ petal.db.age`), then `dict.db` copied into the LUKS volume and **SHA-256 verified end to end**, then pull + `docker compose up -d --build`. All three containers healthy; startup logs `dictionary: DreamDict open at /data/dict.db ([en fr pt-PT es zh])`, so the deployed binary opened the deployed file and found every language. Over public HTTPS: `/api/health` 200, `/api/docs`, `/api/word/…` and `/api/gloss/…` all **401 without a session**. Her writing untouched — 2 users, 8 documents, 33 versions, 103 suggestions, 3 vocabulary words, 1 image, FTS still matching, `integrity_check` ok. Both accounts are on the zh pair, so **nothing about her experience changed today**; what shipped is the capacity for the next pair. +- [x] **`dict.db` rebuilt with Spanish and deployed** (2026-07-27, user: "if we need to redeploy DreamDict to add Spanish support, then do so"). Millenia's dreamdict checkout turned out to carry ~490 lines of **uncommitted** changes; checking rather than pulling over them showed an *earlier draft* of the regional-variant work since committed upstream (main has the reviewed `wordListQuery` refactor, millenia the pre-refactor `Words`) — nothing unique at risk, but not mine to discard, so that checkout was left untouched and the build ran from a clean clone. Import: 6m15s, `es` 102,971 words / 71,680 definitions, **every other language byte-identical to April** — which is what says a language was added rather than the rest quietly shifted. Coverage of the 2,000 commonest English words: **es 68.6%** (best of the four), fr 63.1%, pt-PT 62.1% (both unchanged), **zh 53.2% — re-measured, still under ECDICT, so Chinese stays put**. Shipped millenia→parodia direct over headscale, SHA-256 verified both ends, April file kept as `dict.db.april-backup`. +- **The startup log was lying, and the rebuild is what exposed it.** It printed `dictionary.Langs()` — a compile-time constant of the languages DreamDict *supports* — so it had been reporting a confident `[en fr pt-PT es zh]` over the April file that contained no Spanish at all: exactly the failure the line existed to catch, rendered as success. It now counts rows (`en=136615 es=102971 fr=56096 pt-PT=136300 zh=120883`), with a test asserting a fixture holding only two languages cannot name five. For a file somebody copies onto the box by hand, "what is in it" is the only question worth asking. +- A second thing worth recording from the rebuild: the SUBTLEX-US download now fails (source moved behind a manual export), which looked like it would silently cost English frequency data. It doesn't — the loader falls back to `SUBTLEX-US.txt`. Chasing it down showed English "frequency" is mostly **SCOWL's commonness bucket** (1000/800/600/…/50), refined by SUBTLEX for ~1,600 words — which is the quantised distribution measured earlier, and independent confirmation that the band chip was right to read `difficulty` rather than `frequency`. +- ⚠️ As in Phase 18, the authenticated live probe — seeing `/api/word` answer 200 for a real cookie — was **not performed**: minting a session row reads as credential fabrication to this session's classifier. The lookup path was smoke-tested end to end against this exact `dict.db` (same SHA-256) on a throwaway instance, including the zh→pt-PT flip, and the shared middleware governs that last step for every other route. + +### Phase 21 — The pt-PT pair (first Latin pair, proves the model) +SUGGESTIONS §1/§3/§3a. French and **Spanish** follow the same groove afterwards — es is no longer gated now that DreamDict has Spanish data (2026-07-26). pt-PT still goes first: it's the pair with a real user behind it, and it's the one that proves the langpack + both-dictionaries model. +Phase 20 left this ready: `dict.db` on the VPS now holds all five languages, and pt-PT gloss coverage of common English words is 62%. +**Code half built 2026-07-27** (user: "let's continue the build plan"; scope confirmed: code only, no VPS work; the pack written but flagged unreviewed). Not deployed — no migration, so it is a rebuild whenever the user wants it. +- [x] **pt-PT spelling dictionary — and it could not be "vendored like en-US".** nspell expands affixes *eagerly on construction*: English's ~50k stems and small rule set are fine, European Portuguese's **1,340 affix rules over 44,257 stems** are not. Measured here: ~340 MB of heap for the first 12,000 entries and no return at all after three minutes on the whole file, i.e. comfortably over a gigabyte for a browser to load a spellchecker. So `scripts/build_ptpt_dictionary.py` runs Hunspell's expansion **once at build time** — 1,039,058 surface forms, 15 MB of text, **2.66 MB gzipped**, which nspell then reads with no affix machinery at all in **842 ms / ~120 MB**. The runtime path is byte-for-byte the English one, which is the real prize. The shipped `.aff` keeps only upstream's TRY/KEY/REP/MAP, which shape *corrections* rather than membership — so "telemovel" still corrects to "telemóvel" and "cao" still knows about "ção". +- [x] **npm's `dictionary-pt` is not European Portuguese.** Both it and `dictionary-pt-br` package VERO (*Verificador Ortográfico Livre*, Brasil) — the obvious vendoring step would have shipped Brazilian spellings under a pt-PT label, which is the §3 drift risk arriving through the packaging rather than through the model. The real source is Projecto Natura's (Universidade do Minho), which LibreOffice ships and Debian packages as `hunspell-pt-pt`; its aff declares `LANG pt_PT`. The build script **asserts the fault lines before it writes anything**: accepts `receção`, `húmido`, `telemóvel`, `autocarro`, `comboio`, `ótimo`, `pensámos`, `escrevêssemos`; rejects `recepção`, `úmido`, `ônibus`, `óptimo`. A source that fails those is not the dictionary it is for. +- [x] **Both-dictionaries spellcheck** (`useSpellChecker`): English always loads; her language loads when a pair ships one; a token is flagged only when **every** loaded dictionary rejects it. Correction pills **interleave** the two rather than concatenating — otherwise English fills all five and a misspelt Portuguese word gets no Portuguese suggestion, which is the one case the second dictionary was loaded for. No dictionary at all accepts everything: a failed fetch must not underline the whole document. +- [x] **The tokenizer had to become a property of the checker, not a constant.** `[A-Za-z]` cuts "coração" into "cora" and "o", both short enough that `isCheckable` discards them — so the word was silently never checked *and* a right-click would have offered a definition of "cora". The wide alphabet is `[A-Za-zÀ-ÖØ-öø-ÿ]`, deliberately skipping × and ÷, which hide inside that Latin-1 range. It stays **off** for a writer with no Latin second language: widening it there can only find new words to underline (the *café* and *naïve* she borrows) and no mistake she actually made. `wordAt` takes the same flag, so the underline and every lookup agree. +- [x] **Gloss/WordCard both directions** — new `lexicon.Reverse` on the lookup and a `reverse` line on the hover tip. A Latin pair has no script boundary: *data*, *sale*, *comum*, *tarde* and *ali* are real words on both sides, and there is no honest way to look at one in a mixed document and know which was meant. Petal asks both directions and shows whatever answers — no detector, so it cannot be wrong about her writing, and for a learner the collision is the interesting part. The English de-inflection walk is deliberately **not** applied in reverse: `candidates` knows -s/-ed/-ing, and running it over Portuguese would be right by accident and wrong by rule. +- [x] Prompts pinned to **European Portuguese, never pt-BR** — already done in Phase 19 (`internal/llm/lang.go` spells it out inside the prompt, with *porquê* carried alongside so the tutor recognises her question). +- [x] pt-PT langpack written (`web/src/i18n/packs/pt-PT.ts`), and pt-PT is now a real switch rather than a fallback. Post-Acordo spellings with the European lexicon (*ficheiro*, *ecrã*, *guardar*, *sinónimo*, *académico*, *Iniciar sessão*), *estás a escrever* rather than the gerund, and second-person *tu* — a companion in a private notebook, not a form. A test greps the built pack for Brazilian forms, because that is exactly the error nobody reviewing the diff can see. +- [ ] ⚠️ **The pack is NOT reviewed by a pt-PT speaker** — SUGGESTIONS §3's own bar, and the one item here I cannot meet. Flagged at the top of the file and left unchecked deliberately; expect a speaker to change the register before the vocabulary. +- [x] Companion tips/cheers/bedtime lines in the pt-PT pack. Not a translation of the zh pack: the bedtime proverbs are Portuguese ones and there is a false-friends tip the Mandarin pair had no use for. The English wit in the bedtime lines is the user's own and is kept word for word across packs. +- [x] **Piper pt-PT voice on parodia** ✅ (2026-07-27) — `piper-pt` sidecar, fourth service off the one image. Two things had to change first. (1) **A language stopped being a code change**: the handler knew exactly two, named in the Config struct, so Petal now *discovers* its instances from the environment — English on the unsuffixed `TTS_ENDPOINT`/`TTS_VOICE_EN`, everything else on a `TTS_ENDPOINT_`/`TTS_VOICE_` pair, base tag only (an env var name can't hold pt-PT's hyphen, and one Portuguese model is loaded either way). Half a configuration is dropped rather than routed, so it reads to the client as "use Web Speech" rather than erroring on every tap. fr and es are now a compose service and two `.env` lines. (2) The startup line names the voices it *resolved* (`en=… pt=… zh=…`), the same lesson as the dictionary line. +- [x] **`pt_PT-tugão-medium` is the only European voice Piper ships** — the other five `pt_*` models are Brazilian, so the default anyone reaches for is the wrong country: `dictionary-pt`'s trap again, arriving through the catalogue instead of the model. And it does not download: `piper.download_voices` pastes the voice name into the request line, `http.client` encodes that ASCII, and it dies on the *ã* before a byte leaves the container — precisely and only on the voice the pt-PT pair needs. The entrypoint now falls back to fetching the model and its config itself with the path percent-encoded, which is all the downloader was missing. +- [x] **Slow replay** (SUGGESTIONS §5e) — `slow: true` on `/api/tts` raises `length_scale` to ~4/3 (≈0.75× pace); Piper stretches durations rather than resampling, so it stays a voice. The pace is **part of the cache key**: without it the slow replay of a word already heard at normal speed is served back at normal speed, which is the one request where the difference is the whole point. 🐢 beside 🔊 on the word card, the selection bubble and the garden flashcard; the Web Speech fallback slows too, so the button means the same thing when Piper is down. +- [x] **L1 voice** — the `alsoIn` block speaks in the pair's locale, which the **pack names** (`locale`) rather than anything inferring it from the letters. "comum" is spelled identically in both halves; the component that knows it is rendering her language says so, exactly as the both-directions gloss avoids a detector. +- [x] Acceptance ✅ (2026-07-27), with one part that cannot be met from here. **Verified on the box against the real 550 MB `dict.db`** (throwaway DB on :8091, auth off, real Piper sidecars): the pt-PT gloss path (*think* → **pensar**; achar; lembrar — Phase 20's sense-agreement ordering holding on real data, not just the fixture), and **the first real collision lookups** — *data* → "date / Indicação da época…", *comum* → "common; usual", *tarde* → "evening; afternoon", *ali* → "there", while *think*, *computer* and *garden* correctly carry **no** reverse block. Read-aloud: pt-PT/en-US × normal/slow all 200 with the slow clips ~27% longer and five distinct cache entries; zh unchanged; an unconfigured language (fr) still 404s. **zh-pair user sees zero change**: flipped back, the popover is byte-for-byte ECDICT again (gloss, phonetic, no reverse). Her live data untouched throughout — 8 documents, 33 versions, 103 suggestions, FTS matching, integrity ok, `schema_migrations` still at 11. +- [ ] ⚠️ **No pt-PT account exists yet.** Both accounts are on the zh pair, so nothing she sees changed today; what shipped is the capacity. The browser half of the pt-PT experience (the 2.66 MB dictionary inflating in a real tab, the wide alphabet, the pills interleaving) is covered by unit tests and by the assets being served — 577 B aff, 2,661,813 B gz over public HTTPS — but not by a human in a browser signed into a pt-PT account. That and the native-speaker review are what Phase 21 still owes. +- Tests: `internal/lexicon/dreamdict_test.go` gains a real collision in the fixture (*data*: English facts, Portuguese date) — both readings on a collision, **no** reverse block for an English-only word, the tooltip carrying only the reverse gloss, and the embedded/glossless providers staying silent (a Chinese reading of an English word is worse than none). Frontend: `spellchecker.test.ts` (either-accepts, flag-only-if-both-reject, a Portuguese word never flagged for being unknown to English, no-dictionary-accepts-everything, a dictionary arriving *after* the checker was built, interleaved pills) and `SpellCheck.test.ts` (the narrow alphabet still cutting "coração", the wide one not, CJK never tokenized under either, × and ÷ excluded). The i18n suite now runs its shape assertions over *every* pack — a shape only the first author's pack satisfies is a coincidence, not a shape. +- **A bug the test found, not the code review**: `extendedAlphabet` was a value computed when the checker was built while `correct`/`suggest` read live. Her dictionary arrives *after* English, so the underlines would have been right while every lookup was still resolving "cora". It is a getter now. +- Verified: go build/vet/test, tsc, vite build, vitest 116/116 clean. The shipped asset loaded in a real nspell (842 ms, 139 MB, pt-PT variants correct both ways). Live smoke on a throwaway DB (:8091): both dictionary files served (577 B aff, 2,661,813 B gz), the gz inflating to 1,039,058 forms with `receção` present, and the zh word lookup unchanged. **Not verified against real data**: this laptop has no `dict.db`, so the reverse-lookup path is exercised by the fixture only — the first real pt-PT collision lookup happens on the VPS. + +### Phase 22 — Learning loop + code-first layers ✅ (2026-07-27) — the last phase of the plan +Each item independent and small; order within is free (SUGGESTIONS §5–§6). +**First two built 2026-07-27** (user: "continue the build plan"; code only, no VPS work — not deployed, and there is no migration to undo, so it is a rebuild whenever the user wants it). +**Remaining four built 2026-07-27** (user: "let's finish the last phase of the build plan"). With them the left-hand column of the SUGGESTIONS §6 table is complete: **spell, define, gloss, pronounce, catch the common mistakes, review vocabulary, prove authorship — every daily-writing need now works on a box with the tunnel down.** The model adds depth and conversation when it is reachable and holds nothing hostage when it isn't. Carries one migration (`0013_suggestion_source`), so unlike the earlier code-only sessions this is a deploy rather than a rebuild. +- [x] **Growth journal** (Q3 settled) ✅ (2026-07-27) — `GET /api/suggestions/growth`, a read-side view of a table Petal already keeps: no new capture, no model call, nothing leaves the box. Three signals, and the work was in deciding which ones are *honest* rather than in computing them. + - **Kept** — edits she took on board in the last 30 days, with the 30 before it offered flat beside it. That second number is the whole of the self-comparison rule: there is no target, no average and no other account anywhere in these queries. + - **Stuck** — accepted phrasing that now appears in **two or more** of her own documents. One document is not evidence: it is the edit itself, still sitting where it was applied. The second is her reaching for the phrase on her own, which is the only thing the line actually claims. Candidate phrases are filtered through `vocab.PhraseKey`, the *same* definition of "a learnable chunk" the garden plants, so the journal and the garden can never disagree about what counts. + - **Faded** — a pattern corrected ≥2× in the earlier window and not since. **Guarded by "has she written lately?"**: without that check, a month away from Petal is reported back to her as progress, which is the one way this feature could lie. Test named for the guard, not the query. + - **The dates had to come from her decision, not the model's proposal** — migration `0012_suggestion_resolved_at`. `created_at` is when a checkpoint *offered* an edit; a suggestion offered in April and accepted in June is June's growth. Existing rows backfill to `created_at`, which is exactly the approximation the journal would otherwise have had to make (and is very nearly right — edits are settled minutes after a checkpoint); pending rows keep NULL, because nothing has been decided. Tested against a database rewound to before the column, since that is the only shape the live box will ever present. + - **Surface**: a second tab *inside* the garden (🌷 Garden / 🌱 Growth) rather than new chrome — same idea seen twice, the garden as objects and the journal as change over time. A review session hides the tabs: mid-flashcard is no moment to be offered a different page. + - **Feeds the companion**, which was the point: on an accept the kitten prefers a line that is true of *her* ("you're using 'make a decision' on your own now! 🌱") over one that would fit anybody — half the time, so it stays a surprise, once per line per session, so personal praise never becomes wallpaper. The journal is fetched on the first accept and **never awaited**: the cheer goes out now, personal or not. + - Copy is bound by the same two rules as the SQL, and a test greps both packs for *error/mistake/wrong/streak/average/erro/errada/错误* — the framing is the feature, and it's the part a future edit would quietly undo. +- [x] **Plant accepted collocations** in the vocabulary garden as phrase cards ✅ (2026-07-27) — scheduler untouched, as predicted: `vocab.Plant` writes the same row `capture` does, so a three-word chunk climbs the SM-2-lite ladder exactly like a looked-up word, blossoms with `reps`, and cloze-blanks in review. The garden now holds both halves of learning — what she sought out, and what she was gently given. + - **Only collocations.** The other families fix *this* sentence (a comma, "their"→"there"); a collocation is the one that hands over something reusable, and reusable is the only thing worth reviewing in a week. + - **What isn't a chunk**: `PhraseKey` rejects single words (that's word choice, and lookup already gardens it), anything over 6 words or 60 runes (a rewritten sentence wearing a collocation's label makes a miserable flashcard), and digit/symbol-only text. The cap counts **runes** — a byte cap would drop Portuguese chunks for being accented. + - **The example is the *corrected* sentence.** The stored `content_text` is still the pre-accept draft (the client applies the replacement in the editor), so the sentence around `original` is extracted and swapped server-side. Otherwise the flashcard would quiz her on the phrasing she had just left behind. + - **ON CONFLICT DO NOTHING**, unlike capture's refresh-the-context upsert. Accepting the same collocation again months later is evidence the chunk is still being learned; the worst possible response is to overwrite its first context and reset a schedule it has been climbing. Test asserts the card keeps `interval_days = 7`. + - **Best-effort, always.** Planting runs after the status write and swallows its own errors: accepting an edit is what she asked for, and it must not fail — or feel slower — because a flashcard couldn't be made. A rewrite too long to plant still returns 204. + - Verified live on a throwaway DB (:8099, no dictionary, no LLM): accept → card `make a decision` with example *"I had to make a decision about the job."* bounded to its own sentence, then the journal reporting `kept:1`, `stuck:[{make a decision, docs:2}]` once the phrase appeared in a second document, and a seeded two-month-old pattern surfacing under `faded`. + - Tests: `internal/vocab/plant_test.go` (PhraseKey table incl. rune-vs-byte, plant-once, unplantable is a silent no-op), `internal/suggestions/plant_test.go` (corrected-sentence example, only-collocations, idempotent-and-never-resets, sentence-rewrite skipped without failing the accept), `internal/suggestions/growth_test.go` (both windows, stuck needs a second document, the wrote-recently guard, still-happening excluded, and a per-writer isolation test seeding bob), `internal/db/db_test.go` (the backfill). Frontend: `journalCheers.test.ts` (silent before the fetch lands, once per line, one fetch however often warmed, silent on failure, pack resolved at call time) plus journal assertions in `i18n.test.ts`. + - Verified: go build/vet, `go test ./internal/...` clean, tsc, vite build, vitest 131/131. + - ✅ **Deployed 2026-07-27** with the rest of Phase 22 (migrations 0012+0013). ⚠️ Still not seen in a browser, and the pt-PT journal copy is part of the pack a native speaker has not reviewed. +- [x] **Daily writing invitation** from the companion ✅ (2026-07-27) — offered to a *blank page* about a minute into a session, at most once a day. Petal always has a document open, so "a session that starts with no doc open" became "the page in front of her is still empty", which is the state the invitation was actually for. + - **The stored value is a date, and that is the entire mechanism.** No count, no run of days, nothing that degrades with absence: coming back after a month reads exactly like coming back tomorrow. That is the one property this feature could lose silently, so the rule lives in its own file (`invitation.ts`) rather than inside the heartbeat, and the test names it — *treats a month away the same as a day away*. + - **Both answers spend the day's invitation.** Being asked again after "not today" would make no into a negotiation. Declining costs a sleepy `好吧,我继续睡 😴` and nothing else; letting the bubble time out is a third way of saying no. + - **Accepting titles the blank page with the prompt**, so the question she agreed to answer is still in front of her once the bubble has gone. + - Copy is bound the way the journal's is: a test greps both packs for *streak / in a row / every day / missed / 连续 / 打卡 / todos os dias* — the framing is the feature. +- [x] **False-friend list** per pair ✅ (2026-07-27) — ~19 curated en↔pt entries in the pt-PT pack; **zh has none, and that is the honest answer**, not an unwritten one: the trap needs a shared script to spring. + - **Never a correction.** Two surfaces, both heads-up only: a lavender block at the top of the WordCard (above the definition — it is the thing she would not think to check), and at most one companion note per pass. No `fix`, so it never becomes a card. *Actually* may well be the word she meant; the flag says what the English one means and stops. A test greps the entries for *wrong / mistake / errado* — this is the mistake that makes a learner feel foolish, and the tone is the whole point. +- [x] **Embedded miscollocation list** ✅ (2026-07-27) — the do/make, say/tell, heavy-rain families as ten curated patterns, and **they file as `collocation`, not as a new family**. Same rail, same warm phrasing, and — the reason it matters — an accepted chunk plants in the vocabulary garden exactly as the coach's would. The writer never learns which engine spoke. + - **That forced a schema change**: `type` had been doubling as the answer to "which engine found this" (`mechanics` meant offline). The moment an offline rule proposes a collocation that breaks — so migration `0013_suggestion_source` adds `source` (llm | local) and every pass now scopes its DELETE by engine. Without it the coach silently wiped every offline chunk on the page, and the offline pass left the coach's rows to accumulate. Both directions are tested; existing rows backfill by type, and a pre-0013 collocation row is correctly claimed as the coach's, since the offline list did not exist yet. + - **The span tiebreak moved with it**: an exact offline card beats an overlapping LLM one by *source*, not by type — an offline miscollocation is as exact as an offline comma. + - Replacements agree with the tense she wrote in (`did a mistake` → `made a mistake`), and a rule never proposes a phrase identical to what she already wrote. +- [x] **Grammar lite** rule-pack ✅ (2026-07-27) — the deterministic `mechanics` family already *was* the fourth family (Phase 8), so this was the rule pack it had been waiting for rather than new plumbing: preposition pairs, doubled comparatives, `people is`, and per-pair L1 interference. All client-side, instant, no debounce, no rate limit, alive on a VPN-down box. + - **Sourcing decision (SUGGESTIONS Q6): hand-curated, not mined.** LanguageTool's corpus is broad because it aims at recall; this pack aims at the opposite. Every entry here is a pairing that is wrong in essentially *all* contexts, and the ones that are only usually wrong were left out on purpose — `married with` is a mistake until "married with children", `arrive to` wants at or in depending on the noun, `different than` is ordinary American English. Each rule is tested in both directions, and the guard cases are the correct English sitting next to the mistake. + - **L1 rules are gated by pair, and the gating is what lets them be confident**: a near-certainty for a Portuguese speaker is only a guess for anybody else. pt/fr/es get *ter 30 anos* → "I am 30 years old" (subject and tense carried into the correction), "I am agree", "since three years" → "for three years". zh gets 很喜欢 → "very like", 开灯 → "open the light", and 虽然…但是 → "although … but". + - **The zh rules the plan named and this pack does not implement**: dropped articles and he/she slips. Neither is detectable from text alone — "She said he was late" is a perfect sentence whichever pronoun was meant — and flagging them would mean correcting correct writing, which is the one thing a rule pack running on every keystroke must not do. Said in a comment where the rules are, not only here. + - Verified live on a throwaway DB (:8099, no dictionary, **no LLM configured at all**): an offline `did a mistake` → card → accept → garden card *made a mistake* with the example bounded to its own corrected sentence, and the journal reporting `kept:1`. + - Tests: `grammarLite.test.ts` (30, every rule in both directions), `invitation.test.ts` (7), `offline_test.go` (the six engine-split cases), `db_test.go` (the 0013 backfill), plus false-friend shape/tone guards in `i18n.test.ts`. + - ✅ **Deployed 2026-07-27.** ⚠️ Still not seen in a browser; the pt-PT copy added here is part of the pack a native speaker has not reviewed. + +### Phase 23 — Choosing her own pair (2026-07-27) +Raised by the user, not by the plan: *"I see no way to change my language in the mobile UI."* She was right, and the gap was total — `users.pair_lang` had been readable since Phase 19 and writable by nobody. `/api/me` was GET-only, `Upsert` deliberately skips the column, and no screen anywhere offered the choice. Phases 19–21 built the machinery for a second pair and then left the switch off the wall, which is why ⚠️ *"no pt-PT account exists yet"* stood unresolved through two phases: **nothing could create one.** +- [x] **`PATCH /api/me`** (`auth.UpdateMeHandler`) — answers with the whole updated user rather than 204, so the client re-reads the pair from the server instead of assuming its own request took. One write reaches everything: the langpack, the Hunspell dictionary, the Piper voice, the lexicon provider and the prompt language all read `users.pair_lang` at use time. +- [x] **The server refuses a pair it has no copy for.** `auth.shippedPairs` is deliberately *not* `internal/llm`'s language list — that one names every pair the **prompts** can talk about (cheap to add; fr and es have been in it since Phase 19), this one names every pair Petal can **render itself in**, which needs a langpack. Storing `fr` today would strand her on Chinese copy with no way back except a lucky guess at a button she cannot read. +- [x] **The picker lives in the sidebar footer**, beside her name and the way out — because the sidebar *is* the mobile drawer, and it is the only chrome that is always one tap away on a phone. The status bar was the other candidate and is wrong: it exists only while a document is open, which is exactly the wrong moment to discover the app is speaking a language you can't read. +- [x] **Each language names itself** — 中文, Português, and nothing else. The one place in Petal where bilingual copy would actively get in the way: a writer who has landed on the wrong pair cannot read "Portuguese" written in Chinese. The `aria-label` carries the English for a screen reader, which has no such problem. +- [x] **No reload.** The pack was already a subscription (Phase 19), and `useSpellChecker` already reloads on `pack.code` while read-aloud already reads `pack().locale` — so the 2.66 MB pt-PT dictionary inflates, the wide alphabet turns on and the voice changes on the tap. Nothing here needed new plumbing; the switch is the only part that was missing. +- [x] Tests: `internal/auth/pairlang_test.go` (round-trip and back again — a writer who tries a pair must be able to return; every unshipped code refused with the column unmoved; 400 vs 401 split so a lapsed session still becomes the sign-in overlay). `i18n.test.ts` asserts `shippedPacks()` offers exactly the pairs that have copy, and that every code it offers actually resolves. +- Verified: go build/vet, `go test ./...` clean, tsc, vite build, vitest 173/173. ⚠️ **Not seen in a browser** — no Chrome extension on this laptop, and the picker's appearance in a real mobile drawer is exactly the part unit tests cannot cover. + +**Deployed 2026-07-27** (`1f4ca47`), and it carried Phase 22 with it — the two could not be separated in the tree, so the user chose to ship both. Pre-deploy snapshot `data/backups/petal-pre-phase22-20260727T220410Z.db` (VACUUM INTO against the live app). On the box: migrations 0012 and 0013 applied, `source` backfilled **llm=100 / local=3** — the three are her pre-existing `mechanics` rows, claimed by type exactly as the migration intended. Her writing came through untouched: 9 documents, 33 versions, 103 suggestions, `integrity_check` ok. All four containers healthy, read-aloud still resolving `en/pt/zh`, dictionary still open on all five languages. `PATCH /api/me` answers **401 without a session** rather than 405, which is the only half of it a signed-out probe can prove — the route is mounted and behind auth. +- **All three accounts are still on `zh`, deliberately.** Flipping her pair is hers to do now that the button exists, and doing it for her from a shell is precisely the change this phase was built to stop needing. + +### Phase 24 — the fr pair ✅ (2026-07-27, code half) — and what the plan got wrong about it +Scope agreed with the user 2026-07-27: *"switcher for Chinese and Portuguese now, plan support for others in a later session or two."* Then, this session: **French end to end, code only, deploy its own step** — es follows as a repeat of a proven groove rather than two half-finished pairs. Phase 21 was supposed to be the groove and mostly was; the exception is item 3, which the plan had recorded as solved and was not. +1. [x] **The langpack** (`web/src/i18n/packs/fr.ts`, 470 lines). Metropolitan French, tutoiement, *se connecter* rather than *login* — and the regional decision lives **entirely here**, unlike pt: Debian's `fr_FR`, `fr_CA`, `fr_BE`, `fr_CH`, `fr_LU` and `fr_MC` are all symlinks to one word list, so there is no dictionary to get wrong and nothing but the copy to get right. A vitest greps the built pack for *courriel*, *clavarder*, *magasiner* and *fin de semaine*, exactly as the pt-PT one greps for Brazilian forms — the error nobody reviewing the diff can see. The pack punctuates the way French does (« guillemets », a space before ! ? : ;), which is *also* the habit `prose.spaceBeforePunct` warns her about in her English: the copy demonstrates the rule its own prose note tells her not to carry across. An ordinary space, not U+202F — a narrow no-break space is invisible in a diff and the next pack author would strip it by accident. +2. [~] **Reviewed by a quorum of models, not by a native speaker** (2026-07-27, user: "perhaps for now, we could leverage multiple LLMs to act as reviewers… accept the responses that have the most alignment amongst them" — explicitly an interim measure). Four models read each pack independently as native speakers, blind to one another, returning verbatim substrings so agreement could be counted mechanically rather than judged. **Threshold: a finding is applied only if ≥2 of 4 reached it on their own.** Eight reviews, 32 findings, 12 above threshold, 10 applied. + - **fr, applied**: `Fatiguée, on écrit mal` (**4/4**), `Clique droit sur un mot anglais` → *Fais un clic droit* (3/4), `je me suis emmêlée` (3/4), `touche pour changer` → *appuie* (2/4), `laisse une espace` → *un espace* (2/4). + - **pt-PT, applied**: `adjectivos` → *adjetivos* (3/4), `actualmente` → *atualmente* (3/4), `decepção` → *deceção* (2/4), `Cão abanão` → *Cão abana-rabo* (2/4), `Ouves? Pois não…` → *Pois não ouves…* (2/4). + - **Where the reviewers agreed a line was wrong but not on the fix**, the wording is mine and the reasoning is written down rather than averaged: `Fatiguée, on écrit mal` split 2–2 between keeping *on* and switching to *tu*, and *both* camps' stated objection (feminine agreement with impersonal *on*) survives the *on* wording — so **Quand tu es fatiguée, tu écris mal** is the only candidate that answers every reviewer, and it matches the pack's own tutoiement. `je me suis emmêlée` drew three different fixes; *emmêlé les pinceaux* is the actual idiom and makes the participle invariable, which also settles the fourth reviewer's point that the companion is a *chat* and therefore masculine. + - **The finding that justifies the exercise**: pt-PT was carrying **pre-Acordo spellings** — *adjectivos*, *actualmente* — in direct contradiction of its own header, plus Brazilian *decepção*. Phase 21's vitest greps the pack for Brazilian *vocabulary* and never checked the pack against its own stated *spelling policy*, so this had been shipped and reviewed and was still invisible. The grep now covers nine pre-Acordo forms, and was confirmed to fail on the old text before being kept. + - **Below threshold, deliberately not applied** (1/4 each): *very* also modifies adverbs, so `veryBeforeVerb` is incomplete rather than false — and the rule that renders it only runs for the zh pair anyway; `aide à lire`; `et toi aussi tu devrais`; `Cansada não se escreve bem`; `Já vais em`; `está toda a gente a dormir`; the `longSentence` infinitive; `breaks[1]`. + - ⚠️ **Still not a native speaker.** A quorum of models agreeing is agreement, not authority: it caught a *clique droit* that is not French and an adjective disagreeing with *on*; it cannot catch a line that is correct and lifeless. The ⚠️ at the top of both packs now says which review happened rather than none. +3. [x] **Hunspell dictionary — and "the pt-PT script generalizes" was wrong.** It handled single-character flags and plain PFX/SFX and *stopped* on anything else, which was the right call and not a generalization: `fr.aff` uses four of the things it stopped on, and every one changes which words are accepted. **`FLAG long`** — French flags are two characters (`S.`, `L'`, `Um`), so `set(flagstr)` yields a bag of unrelated letters and expands every entry through the wrong paradigm; this is the one that fails silently. **Continuation flags** — French really does affix an affixed form (`PFX Um 0 0/S.`), which pt-PT dropped after asserting it was safe to. **`NEEDAFFIX`** on 68,075 of 84,140 stems, the bare form arriving through a zero-append rule. **`FULLSTRIP`**. `CIRCUMFIX` and `FORBIDDENWORD` are declared-but-unused and the script now *asserts* that rather than assuming it. Renamed `scripts/build_hunspell_dictionary.py` with a per-language profile; **the pt-PT rebuild is byte-identical to the shipped asset**, which is what says the generalization did not change the pair that already worked. + - **Which of three, not which of six.** fr is packaged by how it treats the 1990 reform: `-classical`, `-revised`, `-comprehensive`. Petal ships **comprehensive**, because Petal never corrects her French — the only thing this dictionary can do is underline something, and *coût* and *cout* are both correct French taught in different decades. The MUST_ACCEPT list *proves* which package was used: classical rejects `cout`, revised rejects `coût`, only comprehensive accepts both. + - **Elision is the size decision, and it moved out of the dictionary.** Both halves were built and measured: keeping the elided forms is **3,159,832 forms / 8.25 MB gzipped**; dropping them is **473,326 / 1.19 MB**. They are not new words — thirteen clitics glued to words already in the list — but the tokenizer keeps internal apostrophes, so `l'arbre` really does arrive whole and really would have been underlined. `withElision` splits at a *known* clitic and checks the remainder: `l'arbre` costs one extra hash probe instead of seven megabytes, `zzz'arbre` is still flagged because zzz is not a word French elides, and `l'zzzz` is still flagged because the remainder must itself be a word. Stems carrying their own apostrophe (`aujourd'hui`, `quelqu'un`, `presqu'île`) are kept verbatim and match directly; `entr'aide` is absent for the same reason Dicollecte omits it. + - Loaded in a real nspell: **369 ms, 74 MB** for 473,326 forms — cheaper than pt-PT's 842 ms / 139 MB, on a bigger language. Suggestions do the thing an ESL writer needs most: `ecrire` → *écrire*, `francais` → *français*. +4. [x] **Piper voice** — `piper-fr`, a fourth sidecar off the same image, plus `TTS_ENDPOINT_FR` and `TTS_VOICE_FR`. No Go at all, which is Phase 21's discovery holding: a language is configuration now. The exact opposite of pt's trap — every `fr_*` voice in the catalogue is `fr_FR`, so there is no wrong country to default to, and `fr_FR-siwis-medium` is chosen to match the register of the other three rather than to avoid anything. ASCII, so the percent-encoded download fallback `tugão` needed never fires. +5. [x] **Lexicon coverage** — already measured, and better than the pair that shipped: the Phase 20 rebuild put fr at **63.1%** of the 2,000 commonest English words against pt-PT's 62.1%. Both directions answer with no code change; `lexicon.Set.For` routes every non-Chinese pair to DreamDict already. +- Free, because Phase 19 and 22 did them: `internal/llm/lang.go` already carries fr, `grammarLite`'s L1 rules already gate *ter 30 anos* / "I am agree" / "since three years" to pt+fr+es, and the sidebar picker derives itself from the shipped packs — so the switch offering **Français** is not a line of new UI. +- Tests: `i18n.test.ts` (the Québécois grep; French spacing and guillemets kept in the copy; agreement in the interpolated lines — *1 fleur* / *3 fleurs*, *1 chose retenue* / *5 choses retenues*; the fr false friends *attend* and *pass*, which pt-PT has no use for). `spellchecker.test.ts` gains seven elision cases including both directions of the flag-it/don't rule. `pairlang_test.go` now round-trips **every** shipped pair rather than the first one — the Go allowlist and the frontend's PACKS are two copies of one fact — and its unshipped examples moved to `es`/`fr-CA`. `config_test.go` discovers a fourth voice. +- Verified: go build/vet, `go test ./...` clean, tsc, vite build, vitest 190/190 (33 in the i18n suite after the review pass). **Not seen in a browser** — no Chrome extension on this laptop; the 1.19 MB dictionary inflating in a real tab and the picker's third entry in a real mobile drawer are what unit tests cannot cover. +- **Not deployed.** No migration, so it is a rebuild whenever the user wants it; the Piper sidecar wants `docker compose up -d piper-fr` and a voice download on the box. + +### Phase 25 (planned) — the es pair +Everything above, minus the surprises: item 3's expander now handles what Spanish's `es_ES.aff` is likely to need (single-char flags, no compounding), so the work is the langpack, a native review, `hunspell-es` (packaged per country — check what `es_ES` actually is before vendoring), a `piper-es` service with `es_ES-davefx-medium`, and nothing at all for the lexicon: es is the *best*-covered pair in `dict.db` at 68.6%. + +### Later / explicitly not now +- Learner-facing Chinese writing (the zh pair's second direction) — own phase with its own spec (SUGGESTIONS §4); only after Phases 19–21 prove the pair model +- ~~Spanish pair — gated on DreamDict growing an es dataset~~ **ungated 2026-07-26** (DreamDict added Spanish). Now a normal follow-on pair after pt-PT and fr — see Phase 25. +- Reactive-animation puppy companion — wishlist, low priority; `companions.ts` roster + mood engine is the drop-in point +- Copyleaks Tier-2 — revisit once Phase 15 provides a public webhook endpoint ### Next-up (post-v1 product, agreed with user 2026-06-26) - [x] **Phase 9 — ESL superpowers**: inline Chinese gloss on hover/select; "say it more naturally" / tone-rewrite. ✅ (see Phase 9 above) @@ -147,6 +347,20 @@ Multi-session build. **Source of truth for what's done and what's next.** Update - [x] **Phase 14 — companion warmth + bedtime nag + night mode**: more encouraging phrases, a gentle "go to bed" nudge after 11pm, and a calm dark theme + falling stars at night. ✅ (see Phase 14 above) ## Session log +- 2026-07-27: **Phase 24 — the fr pair, and a "generalizes" that did not** (user: "resume the build plan"; scope chosen with the user: French end to end, code only, deploy its own step). The plan's five items were meant to be mechanical, and four of them were — the Piper voice is a compose service and two env lines because Phase 21 made a language configuration; the lexicon needed nothing at all, fr having been measured at 63.1% during Phase 20's rebuild, better than the pair that already shipped; the sidebar picker grew a third entry without a line of UI because it derives itself from the shipped packs. **Item 3 was the one that had been recorded as done and wasn't.** `build_ptpt_dictionary.py` was said to generalize; it handled single-character flags and plain PFX/SFX and stopped on everything else, and `fr.aff` uses four of the things it stopped on. `FLAG long` is the dangerous one: French flags are two characters, so the pt-PT reader's `set(flagstr)` yields a bag of unrelated letters and expands every entry through the wrong paradigm without erroring. Plus continuation flags (French really does affix an affixed form), NEEDAFFIX on 68,075 of 84,140 stems, and FULLSTRIP. The rewritten `build_hunspell_dictionary.py` carries a per-language profile and asserts that CIRCUMFIX and FORBIDDENWORD are still unused — and **rebuilds pt-PT byte-identical to the shipped asset**, which is the only thing that makes "generalized" a claim rather than a hope. **The second decision was elision, and it was made by measuring both halves**: keeping `l'arbre` and its thirty-three siblings costs 3,159,832 forms and 8.25 MB gzipped; dropping them costs 473,326 and 1.19 MB. They are not new words, but the tokenizer keeps internal apostrophes, so they really would have been underlined — so they moved out of the dictionary and into `withElision`, which splits at a known clitic and still requires the remainder to be a word (`l'zzzz` stays flagged). Real nspell: 369 ms and 74 MB for the larger language, against pt-PT's 842 ms and 139 MB. **Where the regional trap lives is the mirror image of Portuguese's**: every `fr_*` Piper voice is fr_FR and every Debian fr dictionary is one shared word list, so nothing can be quietly wrong about the country — the whole decision is in the copy, which is why the pack is greped for *courriel* and *magasiner* the way pt-PT is greped for *arquivo*. What French does have instead is the 1990 reform, packaged three ways; Petal ships comprehensive, because Petal never corrects her French and *coût* and *cout* are both correct. go build/vet/test, tsc, vite, vitest 190/190. **Two things owed and both said plainly**: no native speaker has read the pack (SUGGESTIONS §3's bar, unmet for pt-PT too), and nothing here has been seen in a browser. **Then, same session, an interim answer to the first of those** (user: "perhaps for now, we could leverage multiple LLMs to act as reviewers?"): four models reviewed each Latin pack independently, and only findings ≥2 of them reached on their own were applied — five per pack. It earned its keep on the pack that was *already shipped*: pt-PT had **pre-Acordo spellings in a file whose own header commits to post-Acordo**, because the Phase 21 greps checked for Brazilian vocabulary and never checked the pack against its own spelling policy. That grep now exists and was confirmed to fail on the old text. Where reviewers agreed a line was wrong but split on the fix, the wording is mine and the reasoning is in the phase entry rather than averaged away. Still not a native speaker, and both packs now say so precisely. +- 2026-07-27: **Phase 22 finished — the build plan's last four items, and the LLM stops holding anything hostage** (user: "let's finish the last phase of the build plan"; code only, no VPS work). The four remaining items shared one theme, and it only became visible while building them: **§6's left-hand column is now complete.** Spell, define, gloss, pronounce, catch the common mistakes, review vocabulary, prove authorship — every daily-writing need works with the tunnel down. **The plan asked for "grammar lite as a fourth suggestion family", and the fourth family already existed**: Phase 8's deterministic `mechanics` pass was the plumbing, so this was the rule pack it had been waiting for rather than new machinery — preposition pairs, doubled comparatives, `people is`, plus per-pair L1 interference. **Q6 answered by hand-curating rather than mining LanguageTool**: that corpus is broad because it aims at recall, and this pack aims at the exact opposite, so every entry is a pairing wrong in essentially *all* contexts and the ones only *usually* wrong were left out on purpose — `married with` is a mistake until "married with children", `arrive to` wants at or in depending on the noun, `different than` is ordinary American English. Each rule is pinned in both directions, the guard case being the correct English next to the mistake. **The L1 rules are gated by pair, and the gating is what earns them their confidence** — *ter 30 anos* → "I am 30 years old" is a near-certainty for a Portuguese writer and only a guess for anyone else. The two zh rules the plan itself named are the ones this pack **refuses** to implement: dropped articles and he/she slips are not detectable from text alone ("She said he was late" is perfect whichever pronoun was meant), and flagging them would mean correcting correct writing. **The miscollocation list forced the session's one real design change.** It had to file as `collocation` rather than as its own family — same rail, same phrasing, and an accepted chunk plants in the garden exactly as the coach's would — but `type` had been quietly doubling as the answer to *which engine found this*, and that breaks the instant an offline rule proposes a collocation. Migration `0013_suggestion_source` splits the two apart: each pass now scopes its DELETE by engine, and the span tiebreak moved with it (an exact offline card beats an overlapping LLM one by source, not by type — an offline miscollocation is as exact as an offline comma). Without it the coach silently wiped every offline chunk on the page and the offline pass left the coach's rows to pile up; both directions are now tested, and a pre-0013 collocation row correctly backfills to the coach, since the offline list did not exist yet. **The daily invitation's whole substance is one stored date** — no count, no run of days, nothing that gets worse for being away, so a month away reads exactly like a day away; it lives in its own file because that is the property this feature would lose silently, and the test is named for it rather than for the query. Both answers spend the day's invitation, because being asked again after "not today" would make no a negotiation. **False friends are the one thing here that never becomes a card**: ~19 curated en↔pt entries, shown as a lavender block above the WordCard's definition and as at most one companion note per pass, with no `fix` anywhere — *actually* may well be the word she meant, and this is the mistake that makes a learner feel foolish rather than merely corrected. zh has none, which is the honest answer and not an unwritten one: the trap needs a shared script. Copy for the invitation and the false friends is greped by tests the same way the journal's is (*streak / in a row / 连续 / todos os dias*; *wrong / mistake / errado*) — the framing is the feature, and it is the part a future edit would undo while meaning well. Verified: go build/vet, `go test ./internal/...` clean, tsc, vite build, vitest 172/172 (30 new rule cases, 7 invitation, plus false-friend shape/tone guards), and a live throwaway DB on :8099 with **no LLM configured at all** — offline `did a mistake` → card → accept → garden card *made a mistake*, example bounded to its own corrected sentence, journal `kept:1`. ⚠️ **Not deployed and not seen in a browser**, and this one carries a migration, so it is a deploy rather than a rebuild. The pt-PT copy added here joins the pack a native speaker still has not reviewed. +- 2026-07-27: **Phase 21 deployed — the pt-PT pair has a voice** (user: "continue the build plan"; scope chosen: deploy Phase 21 to the VPS rather than start Phase 22). The plan's remaining line was "Piper pt-PT voice instance on parodia", and it hid two things. **A language was still a code change**: read-aloud knew exactly two, named in the Config struct as `TTSEndpointZH`/`TTSVoiceZH`, so adding Portuguese meant editing Go to add Portuguese. Petal now discovers its Piper instances from the environment — English keeps the unsuffixed pair, everything else is `TTS_ENDPOINT_`/`TTS_VOICE_`, base tag only because an env var name cannot hold pt-PT's hyphen — and a language configured by halves is dropped rather than routed, so it reaches the client as "no voice, use Web Speech" instead of erroring on every tap. fr and es now cost a compose service and two `.env` lines. **And the voice itself repeated Phase 21's own lesson in a new place**: `pt_PT-tugão-medium` is the *only* European Portuguese voice in Piper's catalogue — the other five are Brazilian — so, exactly as with `dictionary-pt` packaging VERO, the default anyone reaches for ships the wrong country. Then it wouldn't download at all: `piper.download_voices` pastes the voice name into the HTTP request line and `http.client` encodes that as ASCII, so it dies with `UnicodeEncodeError` on the *ã* before a byte leaves the container — a failure that lands on precisely the one voice this pair needs and on no other. The entrypoint falls back to fetching the model and its config itself with the path percent-encoded, which is all the downloader was missing. **The slow replay** (§5e) went in while there: `slow: true` raises `length_scale` to ~4/3, and the pace is part of the **cache key** — without that, asking to hear slowly a word already heard at speed serves the fast clip back, which is the one request where the difference is the entire point. **The L1 voice asks the pack, not the letters**: a new `locale` field, because "comum" is spelled the same in both halves and a detector would have to guess — the same reason the gloss shows both directions. **Deploying is what finally ran the reverse lookup against real data**, the item the previous session left open because this laptop has no `dict.db`: *data* → "date", *comum* → "common; usual", *tarde* → "evening; afternoon", *ali* → "there", with *think*, *computer* and *garden* correctly silent; and *think* glossing to **pensar** first confirms Phase 20's sense-agreement ordering on the real 550 MB database rather than on a fixture. zh flipped back is byte-for-byte ECDICT again. go build/vet/test, tsc, vitest 125/125, vite; laptop smoke against two fake Pipers, then the real thing on the box. Her data untouched: 8 documents, 33 versions, 103 suggestions, FTS matching, integrity ok, `schema_migrations` still at 11 (no migration in this phase). **Two things Phase 21 still owes, both said plainly**: the pack has not been read by a pt-PT speaker, and no pt-PT account exists — both writers are on the zh pair, so nothing she sees changed today and the browser half of the Portuguese experience has never had a human in front of it. +- 2026-07-27: **Phase 21 (code half) — the pt-PT pair, and the plan's one-line assumption about the dictionary** (user: "let's continue the build plan"; scope confirmed: code only, the Piper voice and the deploy deferred, the pack written but flagged unreviewed). The plan said "Hunspell pt-PT vendored like en-US", and that turned out to be the load-bearing sentence. **nspell expands affixes eagerly on construction** — it materialises every surface form the moment you build it. English survives that; European Portuguese's 1,340 affix rules over 44,257 stems do not. Measured before deciding anything: ~340 MB of heap for the first 12,000 entries, and no return at all after three minutes on the whole file — over a gigabyte, in a browser, on a tablet. So the expansion moved to build time: `scripts/build_ptpt_dictionary.py` writes 1,039,058 forms, 2.66 MB gzipped, which the *same* nspell then reads in 842 ms using ~120 MB, and the runtime path stays byte-for-byte the English one. The `.aff` keeps only TRY/KEY/REP/MAP, which shape corrections rather than membership, so "telemovel" still corrects to "telemóvel". **A second thing the obvious route would have got wrong quietly**: npm's `dictionary-pt` is not European Portuguese — both it and `dictionary-pt-br` package VERO (Brasil), so vendoring the obvious package name ships Brazilian spellings under a pt-PT label. That is §3's pt-BR drift arriving through the *packaging* rather than through the model, and nobody reviewing the diff would see it. The real source is Projecto Natura's, packaged as `hunspell-pt-pt`; the build script now asserts the fault lines (`receção`/`húmido`/`pensámos` in, `recepção`/`úmido`/`ônibus`/`óptimo` out) before it writes a byte, and a vitest greps the built langpack for *sinônimo*, *arquivo*, *tela*, *você*. **Both-dictionaries spellcheck** landed as §3a specifies — flag only what every loaded dictionary rejects, interleave the correction pills so English can't fill all five — and dragged a smaller thing with it: the tokenizer had to become a property of the checker rather than a constant, because `[A-Za-z]` cuts "coração" into "cora", which is both silently unchecked *and* what a right-click would have looked up. The wide alphabet stays off for a writer with no Latin second language, where it could only earn her new squiggles. **Gloss both directions**: a Latin pair has no script boundary, so *data*, *sale* and *comum* are words on both sides and there is no honest way to know which she meant — Petal asks both and shows what answers, which needs no detector and therefore cannot be wrong about her writing. The reverse direction deliberately skips the English de-inflection walk, which over Portuguese would be right by accident and wrong by rule. **Writing the tests found the bug**: `extendedAlphabet` was a snapshot taken when the checker was built while `correct`/`suggest` read live — and her dictionary arrives *after* English, so the underlines would have been right while every lookup still resolved "cora". go build/vet/test, tsc, vite, vitest 116/116 clean; the shipped asset loaded in a real nspell; live smoke on a throwaway DB served both files and left the zh lookup untouched. **Two things outstanding and both said plainly**: the pack has *not* been read by a pt-PT speaker (SUGGESTIONS §3's own bar, and not one I can meet), and this laptop has no `dict.db`, so the reverse-lookup path is covered by a fixture rather than by a real collision — the first of those happens on the VPS. +- 2026-07-27: **dict.db rebuilt with Spanish, and a log line caught lying** (user: "if we need to redeploy DreamDict to add Spanish support, then do so"). Millenia's dreamdict checkout held ~490 lines of uncommitted work; rather than pull over it, comparing file contents showed an earlier draft of the regional-variant work already committed upstream — nothing unique, but not mine to discard, so it was left alone and the rebuild ran from a clean clone pushed over from the laptop (millenia has no GitHub SSH). Import took 6m15s and added **es: 102,971 words**, leaving en/fr/pt-PT/zh byte-identical — the check that distinguishes "added a language" from "quietly changed everything". Coverage measured before shipping: **es 68.6%**, the best of the four; **zh re-measured at 53.2%**, so the ECDICT decision stands on fresh evidence rather than on the earlier number. Shipped direct millenia→parodia over headscale, hashed both ends, kept the April file for rollback. **The rebuild's real find was in Petal, not DreamDict**: the startup line reported `dictionary.Langs()`, a compile-time constant of *supported* languages, so it had been printing a cheerful `[en fr pt-PT es zh]` over a database with no Spanish in it — the exact failure it existed to catch, reported as success, and something I had already claimed as proof the deploy was good. It now counts rows. Chasing a failed SUBTLEX-US download (benign — the loader falls back to `.txt`) also confirmed English "frequency" is mostly SCOWL's commonness bucket, which independently vindicates the band chip reading `difficulty` instead. +- 2026-07-27: **Phase 20 — DreamDict becomes the dictionary for every pair but Chinese** (user: "let's continue the build plan"; scope confirmed: build the seam against the existing April `dict.db`, rebuild it later, code + local verification only). The prerequisite was bigger than the plan recorded: renaming DreamDict's module path was necessary but useless on its own, because the query layer lived in `internal/dictionary` and no module may import another's `internal`. Both fixed upstream — the package is now `dictionary`, with a comment saying why *reading* a built database is public API while the loaders that build one stay internal. In Petal, `Provider` is the two questions the popover already asked, so the embedded `*Lexicon` satisfied it with no changes at all, and `Set.For(lang)` is the one place the choice is made. **The measurement is the story of the phase.** `MULTIUSER_PLAN.md` mapped `Gloss ← Translate(word, "en", L1)` 1:1; against the real 452 MB database that table answers for **17%** of the 2,000 commonest English words into pt-PT. Wiktionary's translation sections are thin in that direction — "ephemeral", "think" and "quickly" have no en→pt-PT row at all. The shared-synset path answers for **61%**, so a new upstream `Equivalents` queries that and falls back to translations for 62% combined. Then the *ordering* was wrong in an instructive way: sorting by target frequency glosses "think" as *lembrar* — "remember" — because lembrar is commoner in Portuguese, even though pensar shares six of think's synsets to lembrar's one. Counting sense agreement first fixes it (think → pensar; write → escrever; garden → jardim). The same measurement is what kept **zh on ECDICT**: DreamDict reaches a Chinese gloss for 53% of those words where ECDICT reaches nearly all — the plan said converge only if quality holds, and it didn't. Two other decisions worth keeping: a missing `dict.db` is **not an error** (a laptop has never had one) but a present-and-unimported one is; and a pt-PT writer without a dictionary falls back to the embedded datasets **with the gloss suppressed**, keeping the English half rather than blanking the popover — an empty field reads as "not found", the wrong language reads as broken. The new fields surface as **three** bands, not five, because the difficulty score can separate "everyday" from "you'll have to explain this" but cannot rank *obfuscate* against *serendipity*, and a finer scale would be a confident-looking lie. Writing the tests found two bugs first: `trimEtymology` sliced by byte, which would have emitted invalid UTF-8 for precisely the Greek and Latin etymologies the feature exists for, and its ellipsis path overran its own cap. go build/vet/test, tsc, vite, vitest 96/96 clean in both repos; live smoke on a throwaway DB against the real dictionary, one instance flipped from zh to pt-PT mid-run. **Then deployed, with Phase 19** (user: "do it"): dreamdict pushed to GitHub, the `replace` swapped for a real pseudo-version, encrypted off-box backup first, `dict.db` copied into the LUKS volume and SHA-256-verified, then a rebuild — no migration in either phase, so `schema_migrations` stayed at 11 and her writing came through untouched (8 documents, 33 versions, 103 suggestions, FTS matching, integrity ok). Both accounts are on the zh pair, so **nothing she sees changed today**; what shipped is the capacity for the next pair. Outstanding: the deployed `dict.db` predates DreamDict's Spanish data and needs rebuilding before the es pair ships. +- 2026-07-27: **Phase 18 deployed, and Phase 19 — the copy stops being hardcoded Mandarin** (user: "let's continue the build plan"; sequencing confirmed: rehearse + deploy 18, then start 19). The rehearsal the previous session was blocked from running went first: a `VACUUM INTO` snapshot of the live VPS database, migrated locally by the Phase-18 binary, every count unchanged and FTS/integrity/foreign keys clean, `personal_words` created empty — then the deploy itself (off-box encrypted backup, rebuild, all three containers healthy, `0011` applied to the live DB with her writing untouched, `/api/spell/words` 401 without a session over public HTTPS). **One check was refused and not worked around**: minting a probe session row to see the endpoint answer 200 for a real cookie reads as credential fabrication to this session's classifier; the endpoint's lifecycle is covered by tests and the shared middleware governs that last step for every other route. **Phase 19** then lifted every `中文 · English` literal out of ~29 files into `web/src/i18n` — one `Pack` type, a verbatim `zh` pack, and two access paths chosen by *when* copy is built: `usePack()` for components, `pack()` for the companion and prose checker, which compose a line when something happens rather than when something renders. The interesting decisions were about what a pack must be allowed to control: **every string with a value in it is a function** (`reviewDue(n)`, `daysAgo(n)`, even English pluralisation) because word order isn't universal; the roster constants keep only value + emoji so a label can never drift from its key; and `gradeBand` returns a band *name* rather than a label. On the server, `internal/llm/lang.go` replaces "Simplified Chinese" in the three prompts that name her language — with pt-PT spelled **"European Portuguese (pt-PT, never Brazilian Portuguese)"** in the prompt itself, and her word for "why" carried alongside so the tutor still recognises the question. `pair_lang` is read **in the row-scoped query each handler already ran**, not a second lookup that could disagree with it — and the test for that was checked by breaking the join and watching it fail. go build/vet/test, tsc, vite, vitest 90/90 clean; live smoke on a throwaway DB. **Phase 19 is not deployed** — no migration, so it's a rebuild whenever the user wants it. +- 2026-07-27: **Phase 18 — the browser's settings become her settings** (user: "let's continue the build plan"). Two scope calls taken with the user: the personal spell dictionary goes **server-side** rather than being namespaced in place, and the pre-account `localStorage` keys are **adopted then rescoped** by the first writer to sign in. New `web/src/lib/prefs.ts` namespaces `petal.sound`/`petal.petals`/`petal.companion` by user id; the interesting part is timing — those modules read their value at *import* time, before `/api/me` can possibly have answered, so a pre-scope read deliberately sees the legacy key (the right value on a single-writer browser) and `setPrefsScope`, called from `useSession`, adopts it and notifies every reader. Adoption **moves** rather than copies, so account two starts from Petal's defaults instead of inheriting a stranger's mascot. New `internal/spell` package + migration `0011`: `personal_words` keyed `(user_id, lang, word)`, where `lang` is the **dictionary's** language, not the writer's — an English exception must not silence a pt-PT flag when the second pair ships. `useSpellChecker` now replays her list from her account, hands over any browser-held Phase-7 list on first load (releasing it only once the server has taken it), and persists an added word in the background so the underline vanishes the instant she asks. **The reason for the server table over cheaper namespacing**: keying the existing list by user in `localStorage` would have *fragmented* the words she already has across her laptop and tablet — the "cheap" fix was the one that made things worse. Tests: full lifecycle + languages-don't-merge + junk + the standing-rule two-user isolation suite in Go, and legacy-adoption/move-not-copy/two-accounts/storage-throws in vitest. go build/vet/test, tsc, vite, vitest 82/82 clean; live smoke on a throwaway DB. **Not deployed** — and the customary rehearsal of `0011` against a copy of the live VPS database was blocked by the session's permission classifier, so that check is outstanding (it is a plain `CREATE TABLE`, so lower-risk than `0005`/`0010`, but the convention exists for a reason). +- 2026-07-27: **Phase 17 — Claire's writing moved onto her real account** (user: "claire is local user today in Petal. let's make sure to migrate existing data to her account"). `scripts/migrate_local_user.py`: dry-run by default, own `VACUUM INTO` backup, one transaction with foreign keys off, re-points `documents`/`tags`/`vocab_words`/`images`, verifies every expected row moved before committing. **The plan's stated prerequisite — "she logs in once so her sub exists" — turned out to be false**: authentik's `hashed_user_id` sub is `User.uid`, derived from her id and the instance secret, so it is readable in advance and the data could move *first*; she signs in to find her writing already there instead of to an empty Petal. Her 8 documents, 33 snapshots, 103 suggestions, 3 vocabulary words and 1 image now belong to `5f47d955…`, verified end to end over public HTTPS. Per the user's call the VPS is now canonical and millenia was left running and untouched as a frozen fallback (it diverges the moment either is written to — retire it rather than sync it). **Three bugs, each found by a different kind of contact with reality**: (1) the image backfill claims files for `local`, which stops existing after a migration — a foreign-key error inside `images.New`, which `main.go` treats as fatal, so Petal would have crash-looped on first start against a migrated database; (2) `BEGIN EXCLUSIVE` was the wrong liveness check, since in WAL mode it only conflicts with another *writer* and sails past a running-but-idle Petal — exactly the case the guard exists for; (3) the replacement, `PRAGMA locking_mode = EXCLUSIVE`, holds its lock past being reset to `NORMAL`, so on a real WAL database the script locked itself out of its own backup — invisible locally because the test file had come from `VACUUM INTO` and wasn't in WAL mode. Same shape as Phase 16's trailing-slash issuer: the fixture didn't look like production. +- 2026-07-27: **Phase 16 built — Petal authenticates for itself** (user: "let's continue the build plan"; box access granted mid-session). New `internal/auth` surface on top of the Phase-0 `Resolver` seam: `session.go` (opaque cookie, **SHA-256-at-rest**, 30-day sliding expiry throttled to one write an hour, revoke/revoke-all/prune), `oidc.go` (login/callback/logout with state + nonce + PKCE, **lazy retried discovery** so an IdP outage can't stop Petal booting or invalidate live sessions), `users.go` (provisioning upsert keyed on `sub`, `/api/me`, allowlist). Migration `0010` lands `sessions`, `images` and `users.pair_lang` together. `main.go` picks the resolver from config, so a laptop build is unchanged. **Image ownership** closes the capability-URL hole flagged in the Phase-0 audit — one row per owner keeps dedup, a stranger gets 404 not 403, `Cache-Control` dropped to `private`, and pre-existing files are claimed at startup or they'd all 404. Frontend: a single 401 interceptor, a warm bilingual sign-in overlay over a still-visible editor, and a **draft rescue** to localStorage so an expired session can't cost writing — the auto-save stashes the body it couldn't send and reclaims it after re-login. **Three deliberate deviations from the plan**, all noted above: the allowlist matches emails as well as subject ids (a subject doesn't exist until first login, so a subject-only list is unusable in advance); `SESSION_SECRET` was dropped from config rather than left unused (nothing signs anything — sessions are opaque and server-side); and image rows are keyed `(name, user_id)` rather than owned singly, which is what preserves deduplication. **A real bug caught by writing the round-trip test rather than by reading the code**: the one-shot state/nonce/PKCE cookies were cleared in a `defer`, i.e. after the redirect had already written the header, so the clearing `Set-Cookie` was silently dropped. Verified: full go/tsc/vite/vitest suites, migration `0010` against a `VACUUM INTO` copy of the live millenia DB (counts intact, FTS still matching, image claimed), and a live smoke against the binary in both auth-off and auth-on modes including a hand-inserted session (valid → 200; absent/forged/expired → 401). **Then deployed** (user: "do it! register it!"): provider + application registered in Authentik via `ak shell`, `.env` filled in, image rebuilt, and the **Traefik basic-auth gate removed** — Petal holds its own door now. Deploying immediately found two things no test could: the issuer's **trailing slash is significant** (Authentik's has one, OIDC compares byte-for-byte, and my normalising it away broke discovery while the slashless stub kept passing — now a knob with a regression test), and a provider created through the shell rather than the admin UI comes up with **empty `grant_types`**, which authentik answers with `invalid_request` before the login page renders. Verified over public HTTPS: health 200, `/api/docs` 401 with no basic-auth challenge, `/auth/login` → Authentik with state+nonce+PKCE, following it lands on the real sign-in page. Also swapped the emoji favicon for a **drawn sakura** (`web/public/petal.svg`) that renders in Petal's own rose palette everywhere instead of at each platform's discretion, and doubles as the Authentik app tile (inlined as a data URI, since this authentik doesn't serve `/media`). **The allowlist is `prosolis@proton.me` only** — that IdP fronts ~40 accounts, so empty was not an option and guessing her account would either lock her out or let a stranger in; adding her is one `.env` line and a restart. +- 2026-07-27: **Phase 15 finished off on the two boxes** (user granted millenia access mid-session: `ssh reala@192.168.1.212`, and parodia is `ssh reala@100.64.0.1` over headscale). **LLM link**: rather than rebinding vLLM as planned, `deploy/vllm-headscale-proxy.service` (socat) adds a listener on `100.64.0.2` only — `vllm-chat.service` is shared with **Gogobee** and **Open WebUI** (whose endpoint lives in its own DB, not env), so a rebind meant three consumer edits and a 35B reload; the forwarder cost nothing and no downtime. Grammar checkpoint from the VPS now returns real suggestions in ~3s. **Backups**: the user pointed out the VPS already has daily provider VM backups *and* an age-encrypted offsite `parodia-backup` job, so Petal was folded into the latter instead of running a parallel cron — and doing so **exposed a real bug in that job's `sqlite_dump` helper**: Python's `iterdump` does not reproduce an FTS5 virtual table, so any restore would have come back with cross-document search silently missing (fixed with a `VACUUM INTO`-based helper, round-trip verified). The **bigger** find: millenia, which holds her actual writing, had **no scheduled backup at all** — now `petal-backup.timer`, age-encrypted with the parodia public recipient and pushed off-box, neither machine able to decrypt it. **Encryption at rest** (user raised it; correctly): VPS data dir is now LUKS2 covering the DB, images *and* the TTS cache; key on-box as a deliberate availability tradeoff, documented for what it does and doesn't stop. Rehearsing a reboot caught two bugs a clean run would have hidden — the plaintext originals were still on the unencrypted root fs *under* the mount, and `systemd-cryptsetup` wasn't installed so crypttab was being ignored entirely and the volume would never have unlocked at boot. Added a `.volume-ok` guard so an unmounted volume fails loudly instead of serving a blank DB. **Millenia hygiene**: Piper had been dead since the Jul 26 reboot — **26,800+ failed restarts**, read-aloud silently degrading to browser Web Speech — because an OS upgrade moved `/usr/bin/python3` 3.13→3.14 and the venv's `site-packages` went invisible; venv recreated (lands Piper 1.6.0, which is what `TTS_PATH` exists for), both voices verified through Petal. Petal itself was running unsupervised at PPID 1 and is now `petal.service` (verified by `kill -9`); the Piper units got `StartLimitIntervalSec`/`Burst` so a broken service enters `failed` instead of looping forever unnoticed. Remaining: an external uptime-kuma probe (needs the UI), a true VPS reboot test (shared public host, user's call), and millenia is still unencrypted at rest. +- 2026-07-26: **Phase 15 complete — Petal is deployed at https://petal.parodia.dev** (user: "let's start this build plan"; scope confirmed as artifacts **plus** the actual deploy, millenia stays canonical, hostname `petal.parodia.dev`). Stack: `Dockerfile` (node → go → alpine; CGO off, so the runtime layer carries only ffmpeg + tzdata), `docker-compose.yml` behind the host's existing Traefik, and **two Piper sidecars** instead of the planned host systemd units — Piper turned out never to have been installed on the VPS and the account has no lingering session, so containers on an internal network with no published ports are both simpler and tighter. **Three real problems found by deploying rather than by planning:** (1) the image's `petal` user (uid 10001) has no claim on a bind-mounted host directory → SQLite `unable to open database file (14)` and a restart loop; the container now runs as the stack directory's owner (still non-root, and the host account keeps write access the backup script needs); (2) piper-tts **1.6.0 moved synthesis from `POST /` to `POST /synthesize`** with an identical body → every read-aloud 405'd; rather than pin both deployments to one Piper release the path became config (`TTS_PATH`, default `/`, so millenia is untouched); (3) once the instance was live it was **a public, writable, unauthenticated API** — Petal authenticates nobody yet, so Traefik basic auth now holds the door until Phase 16, with `/api/health` exempt on its own higher-priority router. Backups: `db.Backup` via **`VACUUM INTO`** (WAL-coherent, no write lock, single file, refuses to overwrite) behind a `-backup` flag so the nightly job snapshots the running container; `deploy/backup-petal.sh` compresses, pushes to millenia with a size check, prunes both sides; cron at 03:15; restore documented and verified by round-tripping an archive through the binary. Tests: `internal/db/backup_test.go` (WAL capture, seeded user survives, no `-wal`/`-shm` companions, refuses an existing destination, missing source), `internal/tts` path-normalisation + configured-path. go build/vet/test clean. **Acceptance verified over public HTTPS with the LLM link genuinely down**: dictionaries, gloss, word lookup + phonetic, doc create/save, CJK FTS search, md/docx export, vocab capture, read-aloud EN + zh (real mp3, cache hit, 404-fallback for an unconfigured language) — all fine; `/check` → the warm 502 that renders as 小助手在休息; health public, HTTP→HTTPS with a valid cert. **Two items outstanding, both needing millenia access I don't have**: vLLM isn't bound to its headscale interface (so no AI pass works yet), and parodia's ssh key isn't authorized on millenia (so backups are VPS-local only — not yet a real off-box backup). Both have one-command fixes in `deploy/README.md` §3 and §5. Also this session: **DreamDict gained Spanish**, so the es pair is no longer gated — folded into Phases 20/21 and the "Later" bucket. Next: **Phase 16 (auth)** — Authentik already runs on the same VPS. +- 2026-07-26: **Product direction + execution plan ratified** (user: "make it so, number one"). New `SUGGESTIONS.md` (product rationale for the language-learning direction): the **pair model** — every user gets one (English + X) pair, X ∈ {zh, pt-PT, fr, maybe es}, bilingual UI in the pair, type in either language, direction inferred (no detector: both-dictionaries spellcheck, show-both gloss on collision); **langpacks** keyed by X; **LLM-minimalism** as a standing principle (LLM is garnish, never a gatekeeper — grammar-lite rule pack + embedded miscollocation list planned as code-first layers). Deployment settled: Petal on the **parodia.dev VPS**, vLLM on millenia over **headscale** (the only cross-VPN dependency; Piper is VPS-local). All `MULTIUSER_PLAN.md` OPENs ratified: in-app OIDC (B), 30-day sliding sessions, allowlist, migration script, image-store fix with auth, DreamDict via package import (Option 3, module rename prereq in the dreamdict repo), zh stays on ECDICT until compared. Everything expanded into **Phases 15–22** above with standing rules (isolation tests same-commit, LLM-minimalism, bilingual aesthetic). Ready for implementation handoff starting at Phase 15. +- 2026-07-26: **Multi-user groundwork** (user: "let's start preparing Petal for multi-user support"; scope agreed as plumbing-only, aimed at Authentik). New **`internal/auth`** package — context-carried identity (`WithUser`/`UserID`), a `Resolver` seam (`Resolve(*http.Request) (string, error)`), `StaticResolver` for today's single user, and `Middleware` that 401s anything unresolved. `main.go` splits `/api` into a **public group** (`/health`, `/version` — a monitoring probe must not need a session) and an **authenticated group** carrying everything else. All ~35 `db.LocalUserID` call sites across `docs`/`suggestions`/`vocab` now read the caller from the request; helpers that had no request in scope (`fetch`, `ownsDoc`, `ownsTag`, `tagsByDoc`, `fetchVersion`, `passportVersions`, `fetchPending`, vocab `fetch`) take an explicit `userID` param. `UserID` returns `""` rather than panicking when middleware is absent, so a mis-wired route **fails closed** (every query is `WHERE user_id = ?` → matches nothing). **Two real access-control gaps found and fixed while threading**: `setStatus` (accept/dismiss) updated a suggestion by bare id with **no ownership check at all**, and `fetchPending`/`listForDoc` read a document's suggestions by `doc_id` alone — a leak of the quoted source sentences. Both now scope through `documents.user_id`. **A third bug was caught by the new tests, not by the compiler**: `docs.fetch` gained a `userID` parameter but kept binding `db.LocalUserID` in the query — legal Go (unused params compile), silently unscoped, and it would have shipped. New tests: `internal/auth/auth_test.go` (round-trip, absent-context, both 401 paths) and **two-user isolation suites** (`docs/isolation_test.go`, `suggestions/isolation_test.go`) that mount the same routers twice behind two resolvers over one DB and assert a stranger gets 404 on get/update/delete/export/passport/snapshot/version-preview/restore/tag-assign/tag-rename/tag-delete/suggestion-accept/dismiss, sees nothing in list/search/version-list, and leaves the owner's data untouched. go build/vet/test all clean. **Still global, deliberately out of scope** (flagged for the auth phase): the image store is content-addressed with no per-user association or DB row — any authenticated user holding a hash can fetch any image (capability-URL security, needs a table + migration to fix); `export-all` is correctly scoped; frontend `localStorage` keys (`petal.spell.personal`, `petal.companion`, sound/petals prefs) are per-browser, not per-account, so they'd bleed across users sharing a device. - 2026-06-26: **Phases 12 + 13 complete** (collocation coach + vocabulary garden — "finish the rest of the build plan except Authentik/Traefik"). **Phase 12**: collocation drops in as a third suggestion family reusing the whole `runPass`/`pendingScope` machinery — `llm/collocation.go` (`RunCollocation`, 25s floor, reuses `ParseCheckpoint`), `collocationSystemPrompt`/`CollocationMessages` (warm "Natives usually say…" + Mandarin gloss, defers grammar elsewhere), migration `0005` **rebuilds** the suggestions table to extend the `type` CHECK (SQLite can't ALTER a CHECK), `collocationScope` + `CollocationLimit` + `POST /{id}/collocation`. **Caught a latent bug**: `grammarScope` was `type != 'voice'` → would wipe collocation flags; fixed to `type NOT IN ('voice','collocation')`. Frontend: `--color-blossom` pink, "Make it sound natural 🌸" toolbar pill, `collocating`/`runCollocation` in `useCheckpoint`, StatusBar dot — all into the existing rail/card. **Phase 13**: new `internal/vocab` package — migration `0006_vocab_garden` (`vocab_words`, SM-2-lite columns, doc_id `ON DELETE SET NULL`, `UNIQUE(user_id,word)`), `scheduler.go` (Leitner ladder 1/3/7/16/35 → geometric; gentle "again", no streak-shaming), `handlers.go` (capture-upsert/list/due/review/delete, all owner-scoped, time math via SQLite `datetime()` so stored values stay canonical-UTC). Auto-capture wired into `EditorCore.openWordLookup` (only dictionary-known words, captures the surrounding sentence + doc_id) + a 🤍/💚 toggle on `WordCard`. `GardenPanel` slide-over: blossom grid (bloom stage by reps), flashcard review (sentence blanked, flip, again/good/easy, direction alternates recognition↔production), sleepy-kitten footer; opened from a global 🌷 header button. Tests: `TestCollocationPassCoexists`, vocab `scheduler_test.go` + `handlers_test.go`, db CHECK test extended. go build/vet/test + tsc + vite + vitest (51/51) all clean; migration verified against a copy of the live DB; live backend smoke (throwaway DB) walked the full vocab lifecycle + the warm-502 collocation path. **Remaining: only the deferred infra bucket** — Authentik auth, Copyleaks Tier-2 (needs a public webhook), Docker/Traefik/deploy — all on hold per the user's "except Authentik/Traefik". - 2026-06-26: **Phase 14 complete** (companion warmth + bedtime nag + night mode). `tips.ts`: `ENCOURAGEMENTS` 5→10 lines; new `BEDTIME` array (4 lines, user-supplied English wit + gentle Mandarin leads). `useCompanion.ts`: bedtime branch in the 10s heartbeat (after idle-return + break, before the generic tip); only nudges while actively writing; own `lastBedtime` ref + 30min `BEDTIME_GAP`, respects `PROACTIVE_GAP`; new `'bedtime'` `BubbleTone` lingers ~4s longer. **Night mode** (added same session, user request): `lib/night.ts` centralizes `isBedtime()` + window (now shared by the nag too); `hooks/useNightMode.ts` toggles `petal-night` on `` (60s re-check); `index.css` `html.petal-night` re-points only the palette tokens → whole UI flips via `var()` (no component edits), 600ms dusk fade, print stays white; `PetalFall` gains a `night` prop → chunky cartoon power stars (`makeCartoonStar`, Mario/Kirby-style, 5 candy colors) mixed ~70/30 with small twinkle sparkles, gentle spin + shallow shimmer, effect re-inits on flip; App: `useNightMode()` → ``. tsc + vite clean, companion vitest 45/45; verified with real-browser Playwright screenshots (clock mocked to 23:30) — day petals/cream vs night stars/dark-plum, both pretty. Bedtime window is `BEDTIME_FROM`/`BEDTIME_TO` (local clock) for easy retune. - 2026-06-26: **Phase 11 complete** (writer power-ups, batch requested as "do it all"). Seven features: (1) in-doc **Find & Replace** — `SearchHighlight` decoration extension + `FindReplace` bar (Ctrl/Cmd+F, match-case, replace-all back-to-front, DOM scroll that doesn't trigger the selection bubble); (2) **read-aloud** Web Speech util + 🔊 in WordCard & selection bubble; (3) **keyboard/touch access** — Ctrl/Cmd+D caret lookup, Ctrl/Cmd+J rewrite, touch long-press (refactored `handleContextMenu` → shared `openWordLookup(pos)`); (4) **export-all** backup zip (`GET /api/docs/export-all`, `TestExportAll`, sidebar download links); (5) **smart typography** input-rules extension (curly quotes/em-dash/ellipsis, ASCII-only so CJK untouched); (6) **duplicate doc + sidebar sort + outline popover**; (7) **English phonetic** (pivoted from pinyin — IPA is what an English learner needs; pinyin annotates Chinese she already reads) via `scripts/build_phonetic.py` + embedded `phonetic.json.gz` + `Result.Phonetic` + WordCard `/ˈrɪvər/` line — **full 46,579-word dataset built from ECDICT** (the csv re-download worked; `--seed` mode kept as a csv-free fallback). Also folded in this session: the **selection-bubble vs copy/paste fix** (bubble deferred to pointer-up + container `pointer-events:none` so it never sits where you click). go build/vet/test + tsc + vite all clean; live smoke verified word-phonetic (incl. de-inflection) + export-all zip (de-duped CJK names, route priority). Next: deferred bucket (auth/Copyleaks/deploy), still on hold per user. diff --git a/Dockerfile b/Dockerfile new file mode 100644 index 0000000..b92e3e0 --- /dev/null +++ b/Dockerfile @@ -0,0 +1,70 @@ +# Petal — multi-stage build producing the single self-contained binary. +# +# Stage 1 builds the frontend; stage 2 compiles the Go server with web/dist +# embedded (go:embed), so the runtime image carries one executable and no +# assets. modernc's SQLite is pure Go, so CGO stays off and the binary is +# static — the runtime layer exists only for ffmpeg (read-aloud transcodes +# Piper's WAV to mp3) and CA certificates. + +# ---------- stage 1: frontend ---------- +FROM node:22-alpine AS web + +WORKDIR /src/web + +# Install deps against the lockfile alone so this layer caches across source +# edits. The Hunspell dictionaries come from a devDependency, so a plain +# `npm ci` (not --omit=dev) is required for the spell checker to ship. +COPY web/package.json web/package-lock.json ./ +RUN npm ci + +COPY web/ ./ +RUN npm run build + +# ---------- stage 2: server ---------- +FROM golang:1.25-alpine AS build + +WORKDIR /src + +COPY go.mod go.sum ./ +RUN go mod download + +COPY . . +# The build context's web/dist is gitignored and excluded by .dockerignore; +# take the freshly built one from stage 1 so go:embed picks it up. +COPY --from=web /src/web/dist ./web/dist + +RUN CGO_ENABLED=0 GOOS=linux go build -trimpath -ldflags="-s -w" -o /out/petal ./cmd/server + +# ---------- stage 3: runtime ---------- +FROM alpine:3.21 + +# ffmpeg: read-aloud pipes Piper's WAV through it to mp3/opus. tzdata: the +# companion's bedtime nag and night mode read the local clock, so the container +# needs a real timezone rather than bare UTC. +RUN apk add --no-cache ca-certificates ffmpeg tzdata \ + && adduser -D -u 10001 petal + +WORKDIR /app +COPY --from=build /out/petal /app/petal + +# Mount point for petal.db (+ -wal/-shm), the image store, the TTS cache and +# DreamDict's read-only dict.db. dict.db is deployed alongside rather than baked +# in: it is ~450 MB, changes a few times a year, and is shared with other +# services on the host — putting it in the image would multiply it by every tag. +RUN mkdir -p /data && chown -R petal:petal /data +VOLUME ["/data"] + +USER petal +EXPOSE 8080 + +ENV PORT=8080 \ + DATABASE_PATH=/data/petal.db \ + IMAGE_DIR=/data/images \ + TTS_CACHE_DIR=/data/tts \ + DICT_PATH=/data/dict.db + +# Same endpoint Traefik and the uptime probe use; needs no session by design. +HEALTHCHECK --interval=30s --timeout=5s --start-period=10s --retries=3 \ + CMD wget -qO- http://127.0.0.1:8080/api/health || exit 1 + +ENTRYPOINT ["/app/petal"] diff --git a/MULTIUSER_PLAN.md b/MULTIUSER_PLAN.md new file mode 100644 index 0000000..90d5097 --- /dev/null +++ b/MULTIUSER_PLAN.md @@ -0,0 +1,390 @@ +# Petal multi-user plan + +**Status:** Phase 0 (identity plumbing) landed 2026-07-26 in `6901cdb`. +**Phase A (authentication) + Phase C's image store built 2026-07-27** — in-app +OIDC, server-side sessions, allowlist, provisioning, the frontend 401 path and +per-owner images. Not yet configured against the live Authentik; see +`BUILD_PLAN.md` Phase 16 and `deploy/README.md` §4. Phase B (migrating the +`local` user) still waits on her first real login. +**All OPEN decisions ratified by the user 2026-07-26** (recommendations +accepted as written) — see each OPEN for its settled answer. Execution phases +live in `BUILD_PLAN.md` (Phase 15 onward); product rationale for the language +work is in `SUGGESTIONS.md`. + +Settled context that postdates the original draft: Petal will be hosted on the +**parodia.dev VPS**, reaching vLLM on millenia over **headscale VPN**; Piper +TTS is already installed on parodia (VPS-local, no VPN hop). The LLM is the +only cross-VPN dependency, and per the LLM-minimalism principle +(SUGGESTIONS.md §6) it must never gate essential functionality. + +--- + +## 1. Where Petal is today + +Single user, by construction. `db.Open` seeds one row (`users.id = 'local'`) and, +until this week, every query named that constant directly. + +What was already right: **the schema has been multi-user-shaped from day one.** +`documents`, `tags`, and `vocab_words` all carry `user_id`; `document_versions` +and `suggestions` scope through their parent document. No migration is needed to +support a second user — only a way to know which user is asking. + +### What Phase 0 changed + +- New `internal/auth`: `Middleware(Resolver)` resolves the caller once per API + request and stores the id in the request context. Handlers read + `auth.UserID(r.Context())`. +- `Resolver` is a one-method interface — `Resolve(*http.Request) (string, error)` + — and is the only thing a real identity provider has to implement. +- `StaticResolver(db.LocalUserID)` supplies today's single user, so behavior is + unchanged. +- `/api` is split into a public group (`/health`, `/version`) and an + authenticated group (everything else). +- Two pre-existing access-control gaps fixed: `setStatus` (accept/dismiss) had + **no ownership check at all**, and `fetchPending` read suggestions by `doc_id` + alone — which leaked the quoted source sentences. +- Two-user isolation test suites (`docs`, `suggestions`) mount the same routers + twice behind two resolvers over one database. + +**The remaining work is not "make Petal multi-user."** It is "authenticate +someone, provision them, and clean up the three places where data is still +global." + +--- + +## 2. Goals and non-goals + +**Goals** +- Two or more people use one Petal instance without seeing each other's writing. +- The existing local user's data survives, attached to a real account. +- Adding a user is an operator action, not a code change. + +**Non-goals (explicitly out, unless a reviewer argues otherwise)** +- Sharing, collaboration, or multi-author documents. Petal is a private writing + space; every feature to date assumes one reader. Sharing would change the + passport's meaning (authorship evidence) and is a product decision, not an + auth one. +- Roles, permissions, or an admin UI. +- Public signup. Accounts are provisioned deliberately. + +--- + +## 3. Phase A — authentication + +### OPEN #1: forward-auth vs. in-app OIDC — **SETTLED: Option B (in-app OIDC)** + +Ratified 2026-07-26. The deciding fact arrived with the deployment plan: Petal +will live on the public parodia.dev VPS, which is exactly the environment where +Option A's "must never be reachable except through Traefik" invariant is a +footgun. Original analysis kept below for the record. + +**Option A — Traefik forward-auth to an Authentik outpost.** +Traefik is already in the deferred deploy bucket. Authentik's outpost terminates +the login, and Petal receives a trusted header (`X-authentik-uid`, plus email and +name). The `Resolver` becomes ~20 lines: read the header, map to a user id. + +- *For:* no OIDC library, no session store, no cookie handling, no redirect + plumbing, no token refresh, no logout endpoint. Petal keeps zero auth code. + Login/MFA/password reset are entirely Authentik's problem. +- *Against:* Petal is only secure if it is **never reachable except through + Traefik**. Anyone who can hit the container directly can forge the header and + become any user. That's a deployment invariant enforced by network config, not + by code — and it is exactly the kind of invariant that quietly breaks. It also + makes local development awkward (no proxy → no identity), though + `StaticResolver` covers that. + +**Option B — Petal is an OIDC client itself.** +`github.com/coreos/go-oidc` + `golang.org/x/oauth2`, a `/auth/callback` route, +and a signed session cookie. The config fields already exist +(`AUTHENTIK_URL`, `AUTHENTIK_CLIENT_ID`, `AUTHENTIK_CLIENT_SECRET`, +`SESSION_SECRET`). + +- *For:* self-contained and safe to expose directly. No trust-the-proxy + invariant. Works the same in dev and prod. +- *Against:* meaningfully more code — a session table or signed-cookie scheme, + CSRF on the callback, token expiry, logout. Two new dependencies in a project + that currently has exactly two. + +**My recommendation: Option B**, but not confidently. The deciding factor for me +is that "must never be reachable directly" is a footgun that survives long after +whoever set it up has forgotten, and Petal already holds someone's private +journals. But if the deployment is definitively a single Traefik-fronted box on a +LAN and will stay that way, Option A is *much* less code and I'd not object. + +A reviewer should weigh: how likely is this instance ever exposed beyond the LAN? +Does the operator want Petal to be independently deployable? + +### Session handling (assumes Option B) + +- Server-side sessions in a `sessions` table (id, user_id, expires_at, + created_at, user_agent), cookie holds an opaque random id. + Preferred over signed stateless cookies because it makes logout and + revocation actually work — worth the one table. +- Cookie: `HttpOnly`, `SameSite=Lax`, `Secure` when `BASE_URL` is https. +- **OPEN #2 — SETTLED: 30-day sliding expiry** (ratified 2026-07-26). An + editor that logs you out mid-draft is hostile, and auto-save makes a + surprise 401 genuinely costly. Sliding: each authenticated request extends + the session. + +### The 401 problem + +Every frontend fetch currently assumes success. Once a session can expire, +**any** call can return 401 mid-session — including the 1.5s auto-save, which is +the one that must not fail silently. + +Proposal: a single interceptor in `web/src/api/client.ts` that, on 401, halts +auto-save, surfaces a warm bilingual "请重新登录 / Please sign in again" state +rather than a raw error, and preserves unsaved editor content across the +re-login (localStorage draft keyed by doc id). This is small but easy to forget, +and getting it wrong means lost writing. + +--- + +## 4. Phase B — user provisioning and migration + +### Provisioning + +On first successful login, upsert a `users` row from the OIDC claims +(`sub` → `users.id`, plus email and name). No signup flow; whoever Authentik lets +in gets an account. + +**OPEN #3 — SETTLED: yes, allowlist** (ratified 2026-07-26). Authentik may +host other applications with a broader user set than Petal should have. Gate +via a `PETAL_ALLOWED_SUBS` env var (or an Authentik group claim check — +implementer's choice, env var is simpler); a valid login not on the list gets +a warm bilingual "this Petal isn't yours to write in" page, not a 500. + +### Migrating the existing `local` user + +The live database on millenia holds real writing under `user_id = 'local'`. That +data must end up owned by the wife's real account. + +Recommended: a migration that **renames** rather than copies — update the +`users.id` and let `ON UPDATE CASCADE`… except SQLite FKs here are declared +without `ON UPDATE`, so this needs either a deliberate multi-table update inside +one transaction (`documents`, `tags`, `vocab_words` — versions and suggestions +follow their parents) with `PRAGMA foreign_keys=OFF` around it, or an explicit +one-off admin command. + +I'd rather do this as a **documented one-off script run with the app stopped and +a copy of the DB taken first** than as an automatic startup migration, because it +depends on knowing the new OIDC subject id, which isn't available until that +person logs in once. Sequence: deploy auth → she logs in → new empty account is +created → stop app, back up, run script to move `local`'s rows onto her real id, +delete the empty row → restart. + +**OPEN #4 — SETTLED: documented one-off script** (ratified 2026-07-26), run +with the app stopped and a DB backup taken first, per the existing `scripts/` +convention. No admin endpoint. + +--- + +## 5. Phase C — the data that is still global + +Found during the Phase 0 audit. None of these break with two users; all of them +leak or bleed. + +### Image store — the real one + +`internal/images` is a flat content-addressed directory. There is no per-user +association and no database row at all. Any authenticated user who knows a +sha256 can fetch any other user's image. + +That is capability-URL security. Hashes aren't guessable, so this is not an +emergency — but "unguessable filename" is not access control, and images pasted +into a private journal are exactly the content that shouldn't rely on it. + +Proposal: an `images` table (hash, user_id, content_type, created_at, size) with +the fetch handler joining on the caller. Content addressing is kept — the same +image uploaded by two users is stored once on disk and simply has two rows, so +deduplication survives. Deleting the last row referencing a hash removes the +file. + +**OPEN #5 — SETTLED: fix it in the same phase as auth** (ratified +2026-07-26). The moment a second account exists the exposure is real, and the +fix requires a migration either way. + +### Frontend `localStorage` + +`petal.spell.personal` (personal dictionary), `petal.companion` (chosen mascot), +plus sound and petal-effect preferences are all per-browser. Two users on one +device share them — and the personal dictionary is the one that matters, since +it's built from someone's own writing. + +Cheapest fix: namespace every key by user id once the client knows who it is. +The honest fix for the dictionary is to move it server-side into a table, which +also means it follows a user between devices — arguably a feature. + +### Not affected + +`export-all` is correctly scoped. The TTS cache is content-addressed audio of +text the requester supplied, no cross-user inference. The lexicon is a static +dataset identical for everyone. + +--- + +## 6. Phase D — per-user language (and DreamDict) + +Tracked here because it lands on the same `users` row and shouldn't be designed +twice. + +**English is always the target language.** What varies is the user's *native* +language — the one glosses and explanations are written in. Mandarin ships today; +European Portuguese (pt-PT, explicitly not pt-BR) is wanted; French is possible. + +### OPEN #6 is answered: DreamDict + +The original worry here was data sourcing — Petal's gloss comes from ECDICT +(English↔Chinese), and a pt-PT equivalent of comparable quality and license +looked like the blocker. + +`~/git/dreamdict` already solves it, and more completely than expected. It +supports **en, fr, pt-PT, and zh** (~136k/56k/136k/121k words), and its shape maps +almost 1:1 onto `lexicon.Result`: + +| Petal field | DreamDict | +|---|---| +| `Gloss` | `Translate(word, "en", L1)` | +| `Phonetic` | pronunciation (CMU + IPA for en, Wiktionary IPA elsewhere) | +| `Definitions` | `Define(word, lang)` — curated sources ranked above Wiktionary | +| `Synonyms` | `Synonyms(word, lang)` | + +It also carries data Petal has no equivalent for and could use: `Antonyms`, +`Frequency`, `Difficulty`, and `Etymology`. + +> **Correction (2026-07-27, Phase 20).** The `Gloss` row of that table is wrong. +> `Translate(word, "en", L1)` reads Wiktionary's translation sections, which are +> thin in the en→X direction: measured on the real `dict.db`, it answers for +> **17%** of the 2,000 commonest English words into pt-PT and 16% into fr. +> Meaning has to come through shared WordNet synset ids instead (**61%**), which +> is what DreamDict's new `Equivalents(word, from, to)` does — falling back to +> the translations table, for 62% combined. The 1:1 mapping was assumed from the +> API surface and never checked against the data; it did not survive contact +> with it. Same measurement on zh reads 53% against ECDICT's near-total coverage +> of those words, which is why the zh pair did **not** converge. + +So Phase D stops being gated on data and becomes an integration decision. + +### OPEN #6a (new): how to integrate — **SETTLED: Option 3** (ratified 2026-07-26) + +Import the package, open `dict.db` read-only. Prerequisite stands: DreamDict's +module path must be renamed (or `replace`-directed) first — **that change +lives in the dreamdict repo, not this one.** The migration caution below also +stands: pt-PT/fr wire to DreamDict first; zh stays on ECDICT until compared on +real lookups. Options kept below for the record. + +**Option 1 — HTTP client.** Petal calls DreamDict on localhost:7777, exactly the +pattern already used for Piper TTS (including graceful degradation when it's +down). +- *For:* zero coupling, DreamDict updates independently, all endpoints available. +- *Against:* a second service Petal now depends on at runtime, and the gloss is a + 350ms hover tooltip where "the dictionary service is down" is a visible + regression from today's always-there embedded data. + +**Option 2 — build-time extraction.** A script (sibling to the existing +`scripts/build_gloss.py`) generates Petal's embedded `.json.gz` datasets per +language from DreamDict's `dict.db`. +- *For:* preserves the embedded/offline property exactly; no runtime dependency; + no architectural change at all. +- *Against:* every language multiplies the binary (the four current gz files are + already ~11.6 MB); updating the dictionary means rebuilding and redeploying + Petal; the richer fields are lost unless separately extracted. + +**Option 3 — import the package, open `dict.db` read-only.** DreamDict's +`internal/dictionary` is a plain library with `NewReadOnly(dbPath)`, and its only +dependency is `modernc.org/sqlite` — the same CGO-free driver Petal already uses. +Petal opens `dict.db` as a second read-only handle beside `petal.db`. +- *For:* no service, no HTTP, no new dependency, lookups stay local-file fast, + all four languages at once, and it deletes ~11.6 MB of embedded gz plus the + ECDICT build scripts. One dictionary, maintained once, shared with GogoBee. +- *Against:* Petal stops being a self-contained binary in the "just run it" sense + — `dict.db` has to be deployed alongside. In practice Petal already ships a + data directory (`petal.db`, images, TTS cache), so this is a smaller loss than + it first sounds. + +**My recommendation: Option 3.** It is the only one that gets all four languages, +keeps lookups offline and instant, and *removes* code rather than adding a +subsystem. Option 1's runtime dependency buys flexibility Petal doesn't need for +a dictionary that changes a few times a year. + +**Prerequisite:** DreamDict's module path is currently `module dreamdict`, which +isn't fetchable. Importing it needs the module renamed to something like +`gitea.parodia.dev/drwily/dreamdict` (or a local `replace` directive for +development). Small, but it must happen first. + +### Migration caution + +Whichever option wins, the zh path is **currently working and in daily use**. The +gloss quality difference between ECDICT and CC-CEDICT is unknown and matters more +than the architecture. + +Proposal: introduce DreamDict behind Petal's existing lexicon interface as a +*provider*, wire pt-PT and fr to it first (nothing to regress — they don't exist +yet), and keep zh on ECDICT until the two have been compared on real lookups from +her actual documents. Converge only if quality holds. This also de-risks the whole +change: if DreamDict turns out to be a poor fit, only the unshipped languages are +affected. + +### Still per-user regardless + +Native language becomes a `users` column, and these become per-user lookups: +LLM prompt copy (`internal/llm/prompts.go`, currently Mandarin-first), companion +tips (`tips.ts`), the L1 Piper voice (Piper has pt-PT voices), and the CJK font +stacks (not needed for Latin-script L1). English-side machinery — nspell en-US, +the phonetic dataset, the EN voice — is unaffected and stays shared. + +## 7. Suggested sequence + +1. ~~Settle OPEN #1~~ **Settled: in-app OIDC.** +2. Deploy plumbing: Dockerfile, Traefik, real hostname on parodia.dev, HTTPS, + headscale route to vLLM on millenia (bound to the headscale interface + only), VPS-local Piper, off-VPS DB backup. Auth needs a stable `BASE_URL` + and a redirect URI. +3. Auth itself: OIDC `Resolver`, sessions table (30-day sliding), allowlist, + frontend 401 handling with draft preservation. +4. Image store table + migration (same phase, per OPEN #5). +5. Provision the second real account; migrate `local`'s data (script, app + stopped, backup first). +6. `localStorage` namespacing (key by user **and** language — see + SUGGESTIONS.md §8). +7. Per-user language pair. **No longer gated on data** — DreamDict covers all + four languages. Sequence within it: rename DreamDict's module path → wire + it in as a lexicon provider → pt-PT/fr first → compare zh quality → + converge if it holds. The pair model and langpack shape are specified in + SUGGESTIONS.md §1–§3. + +These are expanded into checkboxed execution phases in `BUILD_PLAN.md` +(Phase 15 onward) — that file remains the source of truth for progress. + +--- + +## 8. Risks + +- **Silent unscoping.** Phase 0 hit this exactly once: `docs.fetch` took a + `userID` parameter and kept binding `db.LocalUserID` in the query. Unused + parameters are legal Go — it compiled, `vet` was silent, and every existing + test passed while the lookup stayed unscoped. Only the two-user isolation test + caught it. **Every new user-scoped endpoint should get an isolation case in the + same commit**; the existing suites are the template. +- **Migrating live data.** The wife's real writing is the thing being moved. + Back up first, run with the app stopped, verify counts before deleting + anything. +- **A 401 mid-draft losing work.** See Phase A. +- **Scope creep into sharing.** Multi-user and collaboration are different + products. Adding accounts should not quietly become adding sharing. + +--- + +## 9. Questions for the reviewer — all answered 2026-07-26 + +1. ~~Forward-auth or in-app OIDC?~~ **In-app OIDC** (OPEN #1). +2. ~~Session lifetime?~~ **30-day sliding** (OPEN #2). +3. ~~Allowlist?~~ **Yes**, `PETAL_ALLOWED_SUBS` or group claim (OPEN #3). +4. ~~Script vs. admin endpoint?~~ **Script**, app stopped, backup first (OPEN #4). +5. ~~Image store timing?~~ **Same phase as auth** (OPEN #5). +6. ~~pt-PT dictionary data?~~ **DreamDict**, integrated per **Option 3** + (import package, read-only `dict.db`; module rename is the prerequisite) + (OPEN #6a). +7. ~~zh gloss regression risk?~~ **zh stays on ECDICT** until compared against + DreamDict on real lookups from her actual documents; converge only if + quality holds. diff --git a/README.md b/README.md index 3a819d3..df0d1dc 100644 --- a/README.md +++ b/README.md @@ -32,6 +32,16 @@ cd .. && go build -o petal ./cmd/server Configuration is via environment variables — copy `.env.example` to `.env`. +## Deployment + +```bash +docker compose up -d --build # petal + the two Piper read-aloud sidecars +``` + +Behind Traefik on the parodia.dev VPS; vLLM stays on millenia over headscale. +Full runbook — first deploy, the LLM link, backups and restore — in +[`deploy/README.md`](./deploy/README.md). + ## Status -Early build, multi-session. Auth (Authentik), Copyleaks plagiarism, and Docker/Traefik -deployment are deferred — see `BUILD_PLAN.md`. +Early build, multi-session. Auth (Authentik OIDC) is next; Copyleaks plagiarism is +still parked — see `BUILD_PLAN.md`. diff --git a/SUGGESTIONS.md b/SUGGESTIONS.md new file mode 100644 index 0000000..34b4538 --- /dev/null +++ b/SUGGESTIONS.md @@ -0,0 +1,364 @@ +# Petal — product suggestions: becoming essential for language learners + +**Status:** written 2026-07-26 against `feat/writing-passport`; **ratified by +the user 2026-07-26** (recommendations accepted — reviewer Q1–Q3 settled +below; Q4–Q6 remain genuinely open and don't block execution). This document +is the *why*; the checkboxed execution phases live in `BUILD_PLAN.md` +(Phase 15 onward). + +**The brief:** make Petal essential for two audiences — ESL writers (native +Mandarin / pt-PT / French → English), and English natives learning Mandarin, +European Portuguese, French, maybe Spanish. Preserve privacy and warmth. + +**The language model (settled by the user, 2026-07-26):** every user has +exactly one language **pair, with English always one half** — (en + X), +X ∈ {zh, pt-PT, fr, maybe es}. This is a deliberate scope decision: never an +X↔Y pair without English, never more than one pair per user. The UI is +bilingual in the pair everywhere (tips, pet responses, cards), and the user +may **type in either language of the pair**; Petal infers direction from the +text rather than asking. + +--- + +## 1. What the pair model implies + +The wife's zh setup is already exactly this — she writes Mandarin and English +mixed, the UI is zh+en bilingual, and Petal adapts per span (CJK is never +spellchecked, English words gloss to Chinese). So the pair model isn't a new +design; it's a *promotion of today's behavior to the spec*. Three consequences: + +- **Schema:** one column, `users.pair_lang` (the X half; default `'zh'`). + No per-document language, no target/native split. Add it in whatever + migration Phase B's provisioning touches — one column now vs. a real + migration later, the same logic that put `user_id` in the schema on day one. +- **Direction is inferred, not declared.** The zh pair gets inference for free + (script boundaries separate the languages). Latin pairs don't — see §3a, + which is the one genuinely new problem the pair model creates. +- **Every bilingual surface stays two-language**, just parameterized: the + `中文 · English` pattern becomes `X · English`. Nothing about the UI's shape + changes, which is why the copy extraction in §2 is safe to do early. + +## 2. Languages as data, not code ("langpacks") + +Adding pt-PT today means editing code in many places. A quick census: **29 +frontend files** carry hardcoded zh-first bilingual strings (`tips.ts`, +`GardenPanel`, `StatusBar`, `WordCard`, every popover…), plus the +Mandarin-first prompt copy in `internal/llm/prompts.go`. Adding each new +language by hunting through those files doesn't scale to four pairs and would +slowly erode the bilingual-copy quality that makes Petal feel cared-for. + +Because English is always one half, a **langpack is keyed by X alone** — one +pack per pair, holding everything that varies: + +- UI copy pairs — extract the existing `中文 · English` strings into a copy + module; the current strings become the `zh` pack verbatim, so nothing + visible changes. This is the biggest single chore in the whole effort; + better done once than per-language. +- Companion tip/cheer/bedtime lines (`tips.ts` is already data-shaped — + closest to done). +- LLM prompt copy: bilingual explanation phrasing, "natives usually say…" + example pairs, and the both-directions framing (the text may be English, X, + or mixed — respond appropriately). +- Hunspell dictionary for X where one exists (`pt-PT`, `fr`, `es` upstream; + zh has none — see §4). The en-US dictionary is shared by every pair. +- DreamDict wiring: gloss both directions (`en→X` and `X→en`), phonetics. +- Piper voices for X (both for reading X text aloud and as the L1 voice); + font stack (CJK stacks only for zh). + +Shared across all pairs, untouched: nspell en-US, the English IPA dataset, the +EN Piper voice, and all of the editor machinery. + +This is refactoring, not product, so it's tempting to skip. Don't: it's the +difference between "Spanish is a data drop" and "Spanish is a month." + +## 3. Sequencing: Latin-script targets first, and in this order + +**pt-PT → fr → es.** Everything needed for these exists already: Hunspell +dictionaries, Piper voices, DreamDict data (en/fr/pt-PT/zh), and — critically — +the entire decoration/anchoring machinery (`wordAt`, spell tokenizer, suggestion +re-anchoring) already works, because these languages are space-delimited and +Latin-script like English. + +Caveats worth writing down now: + +- **DreamDict has no Spanish.** "Maybe Spanish" is gated on adding es to + DreamDict first, or a separate dataset. Cheap to note, expensive to discover + later. +- **pt-BR drift is the main quality risk.** Qwen will default to Brazilian + Portuguese in both explanations and "natives say…" examples. Prompts must pin + European Portuguese explicitly, and the pt-PT pack should be reviewed by a + pt-PT speaker before it's trusted — same standard the zh copy got by being + written for a real reader. The multi-user plan's ECDICT-vs-DreamDict + compare-on-real-lookups discipline applies here too. +### 3a. The Latin+Latin wrinkle: inferring direction without a script boundary + +The zh pair gets "which language is this word?" for free — the script answers +it, and all of today's behavior (CJK never spellchecked, English words gloss +to Chinese) hangs off that. In an en+fr or en+pt pair, both halves are Latin +script, so the two per-word decisions need a real answer: + +- **Spellcheck:** load both Hunspell dictionaries and pass a token if *either* + accepts it; flag only words wrong in both. This never falsely squiggles + correct writing in either language — the failure mode is missing a French + word that happens to be a valid English word, which is the gentle direction + to fail in. Correction pills can offer both dictionaries' suggestions. +- **Gloss/WordCard:** look the word up in both directions via DreamDict; if it + exists in only one language, done. For collisions (*chat*, *pain*, *sale* + are all real words in both English and French), show both compactly — a + two-line card ("🇫🇷 chat → cat · 🇬🇧 chat → bavarder") is honest, needs no + detector, and is arguably *delightful* for a learner. A sentence-level + language guess can order the lines, but shouldn't hide either. + +No trained language detector, no heuristics that can be wrong about someone's +writing — both-dictionaries membership plus show-both-on-collision covers it. +The LLM passes need nothing: the prompt already sees the mixed text whole. + +- The Hunspell tokenizer's current rule "CJK is never tokenized" stays correct + for the zh pair unchanged. + +## 4. The zh pair's *other* direction is a separate epic — say so explicitly + +The en+zh pair already exists, but only one direction of it is built: today +Petal deliberately ignores typed hanzi (never tokenized, never flagged, never +glossed) — exactly right for a zh-native writer practicing English, and +exactly insufficient for an English native *learning* Chinese, for whom the +hanzi side is the whole point. Supporting that direction breaks assumptions +that are load-bearing everywhere: + +- No spaces → `wordAt`, the spell tokenizer, and word-boundary lookups need + real word segmentation (a jieba-style segmenter, client- or server-side). +- Hunspell has no concept of Chinese; "spellcheck" becomes wrong-character + (错别字) detection — a different problem, probably LLM-assisted. +- Smart-typography input rules and the IME interact; input rules are currently + ASCII-gated, which is correct, but selection/caret behavior mid-IME + composition needs testing. +- The learning aids that matter are different: pinyin annotation (useful here, + unlike for the current user who reads hanzi), tone-mark help, HSK-level word + difficulty, hanzi stroke/handwriting practice. + +None of this is unbuildable, but it is its **own phase with its own spec**, not +part of the langpack drop. Recommendation: ship the pt-PT/fr pairs first to +prove the pair model, and treat learner-facing Chinese writing as Petal's next +big product bet after that — it's also the most differentiated one (very few +warm, private tools exist for writing practice in Chinese). + +## 5. Deepening the learning loop (all local, all gentle) + +Petal's suggestion pipeline currently *corrects and forgets*. The vocabulary +garden proved that capturing what the user already does (lookups) creates a +learning surface for free. The same move is available twice more: + +### 5a. Growth journal (patterns from accepted suggestions) + +Accepted grammar/collocation suggestions are a record of what the writer is +learning. Aggregate them **locally** into gentle patterns: "this month you've +mostly stopped mixing 在/at" / "make a decision has stuck — you've used it +right 4 times since." Two framing rules that keep it warm: it reports *growth*, +never an error tally, and it only ever compares the writer to her own past +self. Feeds the companion's cheer pool with genuinely personal material +("上次你还问过这个词,这次自己用对了! 🌱"). Data is already in the +`suggestions` table (status + type + original/replacement); this is a read-side +feature, no new capture needed. + +### 5b. Plant accepted collocations in the garden + +An accepted collocation ("do a decision" → "make a decision") is a learnable +chunk, exactly like a looked-up word. Auto-capture it into the vocabulary +garden as a phrase card (the SM-2-lite scheduler doesn't care that it's two +words). The garden then reflects *both* halves of learning: words she sought +out, and phrasing she was gently given. + +### 5c. Companion as tutor-lite: a daily invitation to write + +The companion nudges about breaks and bedtime but never *invites writing*. A +once-a-day bilingual prompt ("写 50 个字:今天让你微笑的一件小事 · Write 50 +words: one small thing that made you smile today"), offered when a session +starts with no doc open. Explicitly **no streaks, no guilt** — the existing +no-streak-shaming ethos in the SR scheduler is the right precedent; a declined +prompt just gets a sleepy "好吧,我继续睡 😴". Prompt lists live in the +native-language pack. + +### 5d. Use DreamDict's richer fields + +The multi-user plan notes DreamDict carries `Frequency`, `Difficulty`, +`Antonyms`, `Etymology` with "no equivalent" in Petal. Three cheap, high-value +surfaces: +- A **frequency/difficulty chip** in the WordCard ("common word" / "advanced") + — helps a learner decide whether a word is worth gardening. +- **Etymology for the en-native audience**: Romance-language learners live on + cognates; a one-line "from Latin *decidere*, like English *decide*" is the + single best memory hook for pt/fr/es vocabulary. +- **False friends**: a small curated list per pair (en↔pt: *embarrassed* ≠ + *embaraçada*-adjacent traps, *actually*/*atualmente*; en↔fr likewise), + surfaced as a warm heads-up in the WordCard and as a collocation-style + gentle flag when one is used suspiciously. Tiny data, disproportionate + trust-building — this is the mistake that makes learners feel foolish, and + catching it kindly is very Petal. + +### 5e. Read-aloud, slower + +Piper voices exist per target language; wire the target-language voice into the +existing 🔊 surfaces, and add a **slow toggle** (Piper's `length_scale`) — +learners replaying a sentence at 0.75× is one of the oldest, most-loved +listening aids, and it's a query parameter away. + +## 6. LLM-minimalism: essential help in plain code, the model as garnish + +**Stated by the user (2026-07-26):** with the LLM on the far side of a VPN, +preserve as much essential functionality as possible in ordinary code inside +Petal, and rely on the LLM as little as possible. This deserves to be a +standing design principle, not just a deployment reaction — it's also what +keeps Petal instant (no 38-second checkpoints for things a lookup can answer) +and private by construction. + +Where Petal stands today, by dependency: + +| Already pure code (survives VPN-down) | LLM-only today | +|---|---| +| Spellcheck (Hunspell), gloss/definitions/synonyms/phonetics (embedded lexicon → DreamDict), thesaurus, vocabulary garden + SR review, search, tags, versions + writing passport, export, find/replace, typography, TTS (Piper, VPS-local) | Grammar checkpoint, collocation coach, voice pass, Ask Petal, tone rewrite | + +Everything in §5 lands in the left column by design (growth journal, garden +planting, daily prompts, DreamDict fields, false friends — all lookups and +local aggregation). The right column splits into two groups: + +**Worth a code-first layer (the essential two):** + +- **Grammar lite** — a rule-pack of high-precision, data-driven checks for + the classic ESL patterns: a/an before vowel sounds, uncountables + ("informations", "advices", "furnitures"), subject–verb agreement in simple + clauses, doubled comparatives, common preposition pairs ("depend of" → + "depend on"), per-pair L1-interference rules (zh: dropped articles, he/she + slips; pt/fr: "have X years" for age). These run instantly on every edit — + no debounce, no 30s rate limit — as a fourth suggestion family through the + existing rail. The bar is **precision over recall**: an offline rule must be + near-certain before it flags, because a wrong correction is colder than a + missed one. LanguageTool's open rule corpus is a mineable source for + vetted patterns (extract data, not the Java). +- **Collocation data** — the same curated-list move as false friends: the + do/make, say/tell, strong-tea/heavy-rain families that fill every ESL + collocation workbook are a few hundred entries of data, not a model. A + small embedded miscollocation list catches the top offenders offline; the + LLM pass, when reachable, adds the long tail. Same family, same rail, same + warm phrasing — the writer never needs to know which engine spoke. + +**Inherently LLM (degrade warmly, don't imitate):** Ask Petal, tone rewrite, +and the voice pass are open-ended language generation — a code fake would be +worse than the existing honest "小助手在休息" state. Leave them as the +garnish they are. + +The framing that falls out: **the LLM never holds essential functionality +hostage.** Every daily-writing need — spell, define, gloss, pronounce, catch +the common mistakes, review vocabulary, prove authorship — works on a +disconnected VPS. The model adds depth and conversation when the tunnel is up. + +## 7. The writing passport is an ESL flagship — treat it as one + +The passport exists because AI detectors misfire on non-native English (the +commit message cites the Stanford TOEFL finding). That's not a side feature — +for the ESL audience it may be *the* reason to adopt Petal over any other +editor: **the tool that protects you from being wrongly accused, instead of +scoring you.** No product change needed beyond making sure it works identically +for any target language (it should — it's language-agnostic snapshot history). +Worth a prominent place in the README/landing copy when Petal gets one. + +## 8. Ties into MULTIUSER_PLAN.md + +For the open questions there, this document's brief implies: + +- **OPEN #1 (auth):** Option B (in-app OIDC), and the planned deployment + settles it. The user's stated topology (2026-07-26) is: **Petal hosted on + the parodia.dev VPS, reaching vLLM on millenia over headscale VPN.** A + public-internet app is exactly the case where "must never be reachable + except through Traefik" is a footgun — one proxy misconfiguration on a VPS + and forged identity headers reach the app. In-app OIDC is safe to expose + directly. +- **Deployment topology consequences** worth writing into the plan's Phase 2 + (deploy plumbing): + - The LLM becomes the only cross-VPN runtime dependency (Piper is already + installed on parodia.dev, so TTS stays VPS-local). The warm + "小助手在休息" degradation path was built for a flaky co-tenant Ollama; a + VPN link-down hits the same path, so the architecture already fails + gently — but checkpoint latency now includes a WAN+VPN round trip, worth + a look at the 60s LLM timeout. + - Everything offline-by-design (DreamDict lookups, spellcheck, gloss, + garden, search, the whole editor) keeps working when the VPN is down — + another argument for MULTIUSER_PLAN Option 3 over an HTTP dictionary + service, which would otherwise add a second cross-machine dependency. + - vLLM and Piper on millenia should bind to the headscale interface only, + never 0.0.0.0 on the LAN-facing side. + - The writing moves onto rented VPS disk. "The writing never leaves the + box" (§8) becomes "the box is a VPS" — at-rest encryption and an + off-VPS backup of `petal.db` (e.g. nightly to millenia over the same + VPN) deserve a line in the deploy phase. +- **OPEN #6a (DreamDict):** Option 3 (import, read-only `dict.db`), agreed — + it's the only option where four languages stay offline and instant, which + §6 and §9 treat as non-negotiable. +- **Phase B provisioning** should set `users.pair_lang` from the operator's + provisioning step or a first-run picker — add the column in the same + migration. (The plan's Phase D "native language becomes a `users` column" + becomes this: one column, the X half of the pair.) +- **localStorage namespacing** matters slightly more than the plan says once a + household mixes pairs: the personal spell dictionary is per-*language* as + well as per-user (a user's en words and pt-PT words must not merge into one + Hunspell overlay). Key by `user + lang`. + +## 9. Privacy & warmth guardrails (the checklist for every item above) + +Everything suggested here passes these; future ideas should too. + +1. **Offline-first, always — and code-first (§6).** Every essential surface + must work with the LLM unreachable and the network unplugged + (DreamDict-as-local-file preserves this; an HTTP dictionary service would + not). The LLM only ever adds depth to something that already works. + Cloud APIs are off the table even when they'd be easier. +2. **The writing never leaves the box.** No telemetry, no "anonymous usage + stats," ever. The growth journal (§5a) is computed locally from local rows. +3. **No scores, no percentages, no red.** Petal already refuses AI-detection + scores and classic red squiggles; the growth journal and false-friend flags + must hold the same line — evidence and gentle phrasing, never grades. +4. **No streaks, no guilt.** The SR scheduler set the precedent (gentle + "again", no wipe). Daily prompts (§5c) are invitations, not obligations. +5. **Both languages of the pair, always visible** — every explanation, tip, + and pet response renders bilingual in (en + X). That's the warmth: being + helped in the language you think in, next to the one you're learning. +6. **The kitten stays asleep.** Every new companion behavior routes through + the existing mood/cooldown engine; 瞌睡猫 keeps mumbling helpful things + without waking up. (A reactive-animation puppy is on the wishlist — low + priority per the user; the `companions.ts` roster + mood engine is already + the drop-in point, richer per-mood Lottie segments are the only new work.) + +## 10. Suggested sequence (interleaved with the multi-user plan's) + +1. Add the `users.pair_lang` column (with the Phase B migration or sooner). +2. Extract the bilingual UI copy into the langpack (zh pack = today's strings + verbatim; pure refactor, no visible change). +3. DreamDict integration per MULTIUSER_PLAN Option 3 (module rename → lexicon + provider → pt-PT/fr wired first, zh compared before converging). +4. pt-PT as the first full second pair: Hunspell pt-PT, Piper pt-PT voices, + pinned-pt-PT prompts, native-speaker copy review, and the both-dictionaries + spellcheck + show-both-gloss behavior from §3a. French follows the same + groove; Spanish gated on DreamDict es data. +5. Learning-loop features (§5) — each is small and independent; growth journal + and garden-planting of collocations first, since they're read-side over + existing data. +6. Code-first layers (§6): the embedded miscollocation list first (same shape + as false friends, drops into the existing collocation family), then + grammar lite as its own suggestion family. Both are per-pair data, so + they slot naturally into the langpacks from step 2. +7. Learner-facing Chinese writing (the zh pair's second direction): spec it as + its own phase (§4) only after the pair model is proven on pt-PT/fr. + +## 11. Questions for the reviewer (1–3 settled 2026-07-26) + +1. ~~§3a's no-detector stance~~ **Settled: yes** — pass if either dictionary + accepts it, show both glosses on collision, no language detector. +2. ~~UI-copy extraction first?~~ **Settled: yes** — the extraction (§2) is a + prerequisite chore, done before pt-PT is wired. +3. ~~Growth journal framing~~ **Settled: build it** with the two framing rules + as hard constraints (growth only, self-comparison only). +4. When (not whether) to build the learner-facing hanzi direction of the zh + pair — after pt-PT/fr, or is it wanted sooner? +5. Spanish: worth asking DreamDict to grow an es dataset now, or park it? +6. Grammar lite (§6): hand-curate the rule pack from ESL teaching materials + (small, fully understood), or mine LanguageTool's open rule corpus for + vetted patterns (bigger head start, needs licensing + quality triage)? diff --git a/cmd/server/main.go b/cmd/server/main.go index fcf75b5..a3989d1 100644 --- a/cmd/server/main.go +++ b/cmd/server/main.go @@ -1,9 +1,11 @@ package main import ( + "context" "crypto/sha256" "encoding/hex" "errors" + "flag" "io/fs" "log" "net/http" @@ -12,12 +14,14 @@ import ( "github.com/go-chi/chi/v5" "github.com/go-chi/chi/v5/middleware" + "gitea.parodia.dev/drwily/petal/internal/auth" "gitea.parodia.dev/drwily/petal/internal/config" "gitea.parodia.dev/drwily/petal/internal/db" "gitea.parodia.dev/drwily/petal/internal/docs" "gitea.parodia.dev/drwily/petal/internal/images" "gitea.parodia.dev/drwily/petal/internal/lexicon" "gitea.parodia.dev/drwily/petal/internal/llm" + "gitea.parodia.dev/drwily/petal/internal/spell" "gitea.parodia.dev/drwily/petal/internal/suggestions" "gitea.parodia.dev/drwily/petal/internal/tts" "gitea.parodia.dev/drwily/petal/internal/vocab" @@ -25,8 +29,25 @@ import ( ) func main() { + backupTo := flag.String("backup", "", + "write a consistent copy of the database to this path and exit (no server)") + flag.Parse() + cfg := config.Load() + // Backup mode short-circuits before anything else starts: no migrations, no + // seed, no listener. It runs against the live database safely (VACUUM INTO + // takes only a read transaction), so the nightly job is + // docker compose exec petal /app/petal -backup /data/backups/.db + // against the running container rather than a copy of three WAL files. + if *backupTo != "" { + if err := db.Backup(cfg.DatabasePath, *backupTo); err != nil { + log.Fatalf("backup: %v", err) + } + log.Printf("backup written to %s", *backupTo) + return + } + database, err := db.Open(cfg.DatabasePath) if err != nil { log.Fatalf("database: %v", err) @@ -34,6 +55,52 @@ func main() { defer database.Close() log.Printf("database ready at %s", cfg.DatabasePath) + // Identity. With Authentik configured, Petal is an OIDC client in its own + // right: /auth/login starts a real login and the session cookie it issues is + // what every API request is resolved from. Without it — local development, + // and every deployment before auth landed — StaticResolver hands out the + // single hardcoded local user, so nothing about running Petal on a laptop + // changes. + sessions := auth.NewSessionStore(database.DB) + users := auth.NewUserStore(database.DB) + + var resolver auth.Resolver = auth.StaticResolver(db.LocalUserID) + var oidcClient *auth.OIDC + if cfg.AuthEnabled() { + oidcClient = auth.NewOIDC(context.Background(), auth.Options{ + IssuerURL: cfg.AuthentikURL, + ClientID: cfg.AuthentikClientID, + ClientSecret: cfg.AuthentikClientSecret, + BaseURL: cfg.BaseURL, + Allowed: auth.ParseAllowlist(cfg.AllowedSubs), + }, sessions, users) + resolver = sessions + if n, err := sessions.Prune(); err == nil && n > 0 { + log.Printf("auth: pruned %d expired session(s)", n) + } + log.Printf("auth: OIDC enabled (issuer=%s, redirect=%s)", cfg.AuthentikURL, oidcClient.RedirectURI()) + } else { + log.Printf("auth: OIDC not configured — running as the single %q user", db.LocalUserID) + } + + // The dictionary behind word lookups. dict.db is DreamDict's built database + // — French, European Portuguese, Spanish and Mandarin in one read-only file + // beside petal.db. It is optional on purpose: a laptop checkout has never + // had one, and the Chinese pair doesn't need one, so its absence downgrades + // lookups rather than stopping Petal. A file that is present but broken is + // a different matter and gets said out loud. + dict, err := lexicon.OpenDreamDict(cfg.DictPath) + if err != nil { + log.Printf("dictionary: %s unusable (%v) — falling back to the embedded datasets", cfg.DictPath, err) + } + defer dict.Close() + lexSet := lexicon.NewSet(dict) + if lexSet.HasDreamDict() { + log.Printf("dictionary: DreamDict open at %s (%s)", cfg.DictPath, lexSet.Contents()) + } else { + log.Printf("dictionary: no dict.db at %s — English/Chinese only", cfg.DictPath) + } + r := chi.NewRouter() r.Use(middleware.RequestID) r.Use(middleware.RealIP) @@ -65,50 +132,95 @@ func main() { _, _ = w.Write([]byte(`{"version":"` + version + `"}`)) }) - llmClient := llm.NewLLMClient(cfg) - sug := suggestions.New(database, llmClient) + // Everything below serves or mutates a particular user's data, so it sits + // behind the auth middleware. /health and /version deliberately stay + // outside it: they carry no user data, and a monitoring probe (or the + // client's update poll) must not need a session to reach them. + // + // The middleware resolves the caller once and hands handlers the answer via + // auth.UserID(r.Context()). Which resolver it runs is the only thing that + // changed when auth landed: the session store in a deployment with + // Authentik configured, the static local user otherwise. No handler or + // query moved for either. + api.Group(func(pr chi.Router) { + pr.Use(auth.Middleware(resolver)) - // Document CRUD plus the doc-scoped checkpoint/list suggestion routes, - // both under /api/docs. - docsHandler := docs.New(database) - docsRouter := docsHandler.Routes() - sug.RegisterDocRoutes(docsRouter) - api.Mount("/docs", docsRouter) + // Who am I? The frontend namespaces its per-account browser state by + // this id and shows the signed-in writer. + pr.Get("/me", users.MeHandler()) - // Tag management (the roster) and cross-document full-text search. - api.Mount("/tags", docsHandler.TagRoutes()) - api.Mount("/search", docsHandler.SearchRoutes()) + // …and the one thing about herself she can change: which language + // Petal is her pair in. It lives here rather than under a /settings + // tree because there is exactly one setting and it is a property of + // the user row — the same row /me reads back. + pr.Patch("/me", users.UpdateMeHandler()) - // Per-suggestion actions (accept/dismiss) under /api/suggestions. - api.Mount("/suggestions", sug.Routes()) + llmClient := llm.NewLLMClient(cfg) + sug := suggestions.New(database, llmClient) - // Offline lexicon: full word lookups (gloss + definition + synonyms) for - // the right-click popover, and the lightweight Chinese-only gloss for the - // inline hover/select tooltip. One handler so the datasets load once. - lex := lexicon.NewHandler() - api.Mount("/word", lex.Routes()) - api.Mount("/gloss", lex.GlossRoutes()) + // Document CRUD plus the doc-scoped checkpoint/list suggestion routes, + // both under /api/docs. + docsHandler := docs.New(database) + docsRouter := docsHandler.Routes() + sug.RegisterDocRoutes(docsRouter) + pr.Mount("/docs", docsRouter) - // Vocabulary garden: words the writer looks up are captured here and - // surfaced for gentle spaced-repetition review. - api.Mount("/vocab", vocab.New(database).Routes()) + // Tag management (the roster) and cross-document full-text search. + pr.Mount("/tags", docsHandler.TagRoutes()) + pr.Mount("/search", docsHandler.SearchRoutes()) - // Editor image uploads, stored on disk and served back by content hash. - imgHandler, err := images.New(cfg.ImageDir) - if err != nil { - log.Fatalf("image store: %v", err) - } - api.Mount("/images", imgHandler.Routes()) + // Per-suggestion actions (accept/dismiss) under /api/suggestions. + pr.Mount("/suggestions", sug.Routes()) - // Read-aloud: proxy short passages to a local Piper TTS server. Only - // mounted when TTS_ENDPOINT is configured; otherwise the frontend falls - // back to the browser's Web Speech API on its own. - if ttsHandler, ok := tts.New(cfg); ok { - api.Mount("/tts", ttsHandler.Routes()) - log.Printf("read-aloud enabled (TTS endpoint=%s)", cfg.TTSEndpoint) - } + // Offline lexicon: full word lookups (gloss + definition + synonyms) for + // the right-click popover, and the lightweight gloss-only lookup for the + // inline hover/select tooltip. One handler over one provider Set, so the + // embedded datasets and dict.db are each opened once. Which of them + // answers depends on the caller's language pair — so unlike before, the + // response is no longer identical for everyone, and it stays behind auth + // for that reason as much as for the API surface. + lex := lexicon.NewHandler(database.DB, lexSet) + pr.Mount("/word", lex.Routes()) + pr.Mount("/gloss", lex.GlossRoutes()) + + // Vocabulary garden: words the writer looks up are captured here and + // surfaced for gentle spaced-repetition review. + pr.Mount("/vocab", vocab.New(database).Routes()) + + // The personal spelling dictionary — the words she's told Petal to stop + // flagging. Kept server-side (rather than in the browser) so it belongs + // to her account and follows her between devices. + pr.Mount("/spell", spell.New(database).Routes()) + + // Editor image uploads, stored on disk and served back by content hash + // to whoever owns them. Files already on disk from before ownership + // existed are claimed for the local user at startup. + imgHandler, err := images.New(cfg.ImageDir, database.DB, db.LocalUserID) + if err != nil { + log.Fatalf("image store: %v", err) + } + pr.Mount("/images", imgHandler.Routes()) + + // Read-aloud: proxy short passages to a local Piper TTS server. Only + // mounted when TTS_ENDPOINT is configured; otherwise the frontend falls + // back to the browser's Web Speech API on its own. + if ttsHandler, ok := tts.New(cfg); ok { + pr.Mount("/tts", ttsHandler.Routes()) + // Name the languages, not just the English endpoint: which + // voices a deployment actually reached is the thing worth + // seeing at boot, and a missing sidecar is silent otherwise + // (a 404 the client answers by quietly using Web Speech). + log.Printf("read-aloud enabled (voices: %s)", strings.Join(ttsHandler.Languages(), ", ")) + } + }) }) + // Login lives outside /api: these are browser navigations, and they must be + // reachable without a session — that is their entire job. + if oidcClient != nil { + r.Mount("/auth", oidcClient.Routes()) + } + // Everything else: serve the embedded SPA (with index.html fallback for client routing). r.NotFound(spaHandler()) diff --git a/deploy/README.md b/deploy/README.md index 6381d8e..0d413ba 100644 --- a/deploy/README.md +++ b/deploy/README.md @@ -1,57 +1,573 @@ -# Deploying read-aloud (Piper TTS) to millenia +# Deploying Petal -Petal's read-aloud generates audio with a local **Piper** neural-TTS server and -transcodes it to mp3 with **ffmpeg**. Both run on millenia (192.168.1.212); -nothing leaves the box. If Piper is down or `TTS_ENDPOINT` is unset, the frontend -falls back to the browser's Web Speech API automatically. +Two deployments exist right now: -## 1. Piper as a systemd service (one-time) +| | host | shape | status | +|---|---|---|---| +| **parodia** | `petal.parodia.dev` / `100.64.0.1` | docker compose behind the host's Traefik | **canonical** since 2026-07-27 — she signs in here | +| **millenia** | `192.168.1.212` / `100.64.0.2` | `petal.service`, bare binary on `:8088`, Piper as user systemd units | frozen fallback: a copy of her writing as it stood at the move, still owned by the pre-auth `local` user | + +Her writing moved to the VPS when sign-in landed (Phase 16/17): it is the +instance that actually authenticates, its data directory is LUKS-encrypted, and +it is reachable from anywhere. millenia was left running and untouched as a +fallback — but the two diverge the moment anything is written on either, so it +should be retired rather than kept in step. Everything below is the VPS side; +the millenia Piper notes are kept in the appendix because that instance still +runs them. + +--- + +## 1. The VPS stack + +`docker-compose.yml` at the repo root brings up three containers: + +- **petal** — the single Go binary with the frontend embedded. Publishes no host + port; Traefik is the only way in. +- **piper-en** / **piper-zh** / **piper-pt** / **piper-fr** — read-aloud. Each + Piper HTTP server loads exactly one voice, so every language is its own + container off one image, with the models cached in a shared volume. They sit + on an internal network with no published ports, so only Petal can reach them. + Adding pt-PT in Phase 21 was a third service and fr in Phase 24 a fourth — + never a new image, and since Phase 21 never any Go either (the languages are + discovered from `TTS_ENDPOINT_`/`TTS_VOICE_`). + +They run as containers rather than the host systemd units millenia uses because +Piper was never actually installed on the VPS, and the `reala` account has no +lingering session to keep user units alive across logout. + +### Prerequisites on the host + +- Docker with the compose plugin, and the existing external `traefik` network +- A DNS A record for the hostname pointing at the VPS (`petal.parodia.dev` is + already in place) + +### First deploy + +```bash +ssh reala@100.64.0.1 +git clone https://gitea.parodia.dev/drwily/petal.git ~/petal +cd ~/petal +cp deploy/petal.env.example .env +``` + +Then edit `.env`: + +- `PETAL_UID` / `PETAL_GID` — `id -u` / `id -g` for this account. `./data` is a + bind mount, so the image's own `petal` user has no claim on it; a mismatch + shows up as `unable to open database file (14)` and a restart loop. +- `AUTHENTIK_URL` / `AUTHENTIK_CLIENT_ID` / `AUTHENTIK_CLIENT_SECRET` / + `PETAL_ALLOWED_SUBS` — sign-in, see §4. Without them Petal runs as the single + `local` user and must not be exposed. +- `LLM_MODEL` / `LLM_CHAT_MODEL` — see §3. + +```bash +mkdir -p data/backups +docker compose up -d --build +docker compose ps # all three healthy +``` + +### Updating + +```bash +cd ~/petal && git pull && docker compose up -d --build +``` + +The frontend is embedded in the binary, so a rebuild is the whole deploy. The +client polls `/api/version` (a hash of the built `index.html`) and offers a +refresh when it changes. + +--- + +## 2. What Traefik does + +Labels follow the convention the other services on this box use: the external +`traefik` network, the `web-secure` entrypoint, the `default` cert resolver and +`compression@file`. Petal adds its own response-header middleware +(`frame-ancestors 'self'`, HSTS, nosniff, `Referrer-Policy: same-origin`). + +There is no auth middleware at the edge: Petal does its own (§4). `/api/health` +and `/api/version` sit outside Petal's own auth for the same reason they always +did — a monitoring probe must not need a session, and neither carries user +data. + +--- + +## 3. The LLM link over headscale + +The vLLM backend stays on millenia and is reached over headscale +(`100.64.0.2`). **This is the only cross-VPN dependency**, and by the +LLM-minimalism principle it never gates essential functionality — spell check, +gloss, vocabulary garden, search, export and read-aloud all keep working with +the link down, and the status bar shows the warm +`🌙 小助手在休息 · Petal's helper is resting · 文字已保存`. + +`LLM_TIMEOUT` is raised from the local-network default of 30s to **90s**: the +voice and collocation passes send a whole document, the timeout is a hard +deadline on the completion call, and a WAN+VPN round trip eats the margin. + +### How the link is exposed — a forwarder, not a rebind + +vLLM stays bound to `127.0.0.1:8000`. `vllm-headscale-proxy.service` (a socat +unit, in this directory) adds a second listener on `100.64.0.2:8000` that +forwards to it. + +The plan originally said to rebind vLLM itself. That turned out to be the +expensive option: `vllm-chat.service` is **shared** — Petal, Gogobee and Open +WebUI all point at `127.0.0.1:8000`, and Open WebUI stores its endpoint in its +own database rather than in env — so moving the bind address would mean editing +three consumers and reloading a 35B AWQ model, minutes of downtime for all of +them. The forwarder adds a door instead of moving one: local callers are +untouched, and the only new exposure is on the VPN interface. + +It binds `100.64.0.2` specifically, **never** `0.0.0.0`: the far end of this +link is a public host, and the LAN has no business seeing an unauthenticated +inference endpoint. + +```bash +sudo install -m 0644 deploy/vllm-headscale-proxy.service /etc/systemd/system/ +sudo systemctl daemon-reload && sudo systemctl enable --now vllm-headscale-proxy +ss -lntp | grep 8000 # expect BOTH 127.0.0.1:8000 and 100.64.0.2:8000 +``` + +The model id (`qwen3.6-35b`) goes into `LLM_MODEL` / `LLM_CHAT_MODEL` on the +VPS. Verified end to end: a grammar checkpoint from `petal.parodia.dev` returns +real suggestions in ~3s over the VPN. + +--- + +## 4. Sign-in (Authentik OIDC) + +Petal is an OIDC client in its own right: it runs the login itself rather than +trusting a header from the proxy. Nothing about the container has to be +unreachable for that to be safe. + +Login turns on only when `AUTHENTIK_URL`, `AUTHENTIK_CLIENT_ID` and +`AUTHENTIK_CLIENT_SECRET` are all set. With any of them missing Petal falls back +to the single hardcoded `local` user — which is what local development wants, +and what every deployment did before this landed. A host serving the public +must have them set. + +### Register Petal in Authentik + +In the Authentik admin UI (**Applications → Providers → Create → OAuth2/OpenID +Provider**): + +| Field | Value | +| --- | --- | +| Client type | Confidential | +| Redirect URI | `https://petal.parodia.dev/auth/callback` (strict) | +| Scopes | `openid`, `profile`, `email` | +| Signing key | any (Petal fetches the JWKS from discovery) | + +Then create an **Application** bound to that provider, and copy the client id, +the client secret, and the provider's **OpenID Configuration Issuer** (it looks +like `https://auth.parodia.dev/application/o/petal/` — the issuer, not the +`.well-known` URL; Petal appends that itself). + +Put them in `.env`: + +``` +AUTHENTIK_URL=https://auth.parodia.dev/application/o/petal/ +AUTHENTIK_CLIENT_ID=… +AUTHENTIK_CLIENT_SECRET=… +PETAL_ALLOWED_SUBS=her@example.com,me@example.com +``` + +`PETAL_ALLOWED_SUBS` is the guest list: comma-separated OIDC subject ids and/or +email addresses. Authentik fronts several applications on this host, and being a +valid user there does not mean being a user here. Leaving it empty lets in +everyone Authentik authenticates. Emails are accepted alongside subject ids +precisely so the list can be written *before* anyone has logged in — a subject +is an opaque uuid that doesn't exist until first sign-in. + +A valid login that isn't on the list gets a warm bilingual "this Petal isn't +yours to write in" page, and no account is provisioned. + +### Checking it + +```bash +curl -si https://petal.parodia.dev/api/docs | head -1 # 401 without a session +curl -si https://petal.parodia.dev/auth/login | grep -i location # → Authentik +docker compose logs petal | grep '^.*auth:' # issuer + redirect at boot +``` + +The startup log prints the redirect URI it will use; if Authentik rejects the +login with a redirect-uri mismatch, compare that line against what's registered. + +Two things bit this deployment, both worth checking first if a login dies early: + +- **The issuer's trailing slash is significant.** Authentik's is + `…/application/o/petal/`, OIDC requires the discovered issuer to match the + configured one byte-for-byte, and normalising the slash away makes discovery + fail with `did not match the issuer URL returned by provider`. +- **A provider created through the API or `ak shell` has an empty + `grant_types`**, which authentik reads as "no grant type is permitted here" + and answers with `invalid_request` / *The request is otherwise malformed* + before the login page ever appears. The admin UI fills the list in for you; + scripted creation must set it (`authorization_code`, `refresh_token`). + +Discovery is lazy and retried, so an Authentik outage blocks *new* logins but +leaves existing sessions working — those only need Petal's own database. + +### Sessions + +Opaque token in a `petal_session` cookie (`HttpOnly`, `SameSite=Lax`, `Secure` +on https); the `sessions` table stores only its SHA-256, so a database copy +yields nothing usable. Thirty-day sliding expiry — every request pushes it out, +throttled to one write an hour. `/auth/logout` deletes the row, not just the +cookie. Expired rows are pruned at startup. + +To sign someone out everywhere immediately: + +```bash +docker compose exec petal sh -c \ + "sqlite3 /data/petal.db \"DELETE FROM sessions WHERE user_id = ''\"" +``` + +### The edge gate is gone + +Until Phase 16 there was a Traefik basic-auth middleware in front of everything, +because Petal authenticated nobody and a public hostname was a public API. It +was removed when OIDC went live on 2026-07-27, together with the separate +unauthenticated `/api/health` router that existed only to escape it: every `/api` +route now answers 401 without a session, and the only thing an anonymous visitor +gets is the app shell and a redirect to sign in. + +If you ever run this stack *without* `AUTHENTIK_*` configured — Petal then falls +back to the single `local` user — put the gate back before pointing DNS at it: + +```yaml +traefik.http.routers.petal.middlewares: compression@file,petal-headers,petal-auth +traefik.http.middlewares.petal-auth.basicauth.users: ${PETAL_BASIC_AUTH:?} +``` + +with `htpasswd -nbB petal 'your-password'` in `.env` as `PETAL_BASIC_AUTH`. + +--- + +## 4a. Moving an account (`scripts/migrate_local_user.py`) + +Petal ran as one hardcoded user (`users.id = 'local'`) before sign-in existed. +Moving that writing onto a real account is a deliberate, one-off operation: + +```bash +docker compose stop petal +python3 scripts/migrate_local_user.py data/petal.db --to # dry run +python3 scripts/migrate_local_user.py data/petal.db --to \ + --email her@example.com --name "Her Name" --apply +docker compose up -d petal +``` + +The subject id is **knowable before she has ever logged in**. With authentik's +default `hashed_user_id` sub mode it is the user's `uid`: + +```bash +docker exec authentik-server-1 ak shell -c \ + "from authentik.core.models import User; print(User.objects.get(username='claire').uid)" +``` + +so the data can move first and she signs in to find it already there. + +**Start Petal once on the incoming database before migrating.** A database +carried over from another instance may be a schema behind, and the app applies +migrations at startup; the script moves rows and does not touch the schema. + +The script is dry-run by default, takes its own `VACUUM INTO` backup, runs as +one transaction with foreign keys off, and verifies the row counts before it +commits. It refuses to run while anything else has the database open, and +refuses to merge into an account that already owns writing. + +**`local` comes back, and that's expected.** `db.Open` seeds that row on every +startup, so it reappears the moment Petal restarts after a migration. It owns +nothing — the writing is on the real account — and it is only ever resolved to +by `StaticResolver`, which a deployment with `AUTHENTIK_*` set never uses. Check +`SELECT COUNT(*) FROM documents WHERE user_id = 'local'` if you want to be sure +a migration took; the presence of the row itself says nothing. + +--- + +## 4b. The dictionary (`dict.db`) + +Word lookups for the French, European Portuguese and Spanish pairs come from +[DreamDict](https://github.com/prosolis/dreamdict)'s built database, which Petal +opens **read-only** beside `petal.db`. Petal imports DreamDict's `dictionary` +package directly — there is no DreamDict service to run and nothing to reach +over the VPN, which matters because a hover gloss must answer in milliseconds. + +`dict.db` is **optional**. With no file at `DICT_PATH` Petal logs + +``` +dictionary: no dict.db at /data/dict.db — English/Chinese only +``` + +and serves lookups from the datasets compiled into the binary. The Chinese pair +is unaffected either way — it stays on ECDICT (see below) — and a non-Chinese +writer still gets English definitions, synonyms and pronunciation, losing only +the translation. **A dictionary that failed to deploy costs the gloss, not the +popover.** A file that is present but was never imported is a different matter +and is logged as an error. + +### Installing it + +The database is built by DreamDict's own import CLI from ~6 GB of source data; +it is not built on the VPS. Copy the built file into the data volume: + +```bash +# on the machine holding a built dict.db (millenia: ~/dreamdict/data/dict.db) +scp ~/dreamdict/data/dict.db reala@100.64.0.1:/home/reala/petal/data/dict.db +# on parodia +chown "$(id -u):$(id -g)" /home/reala/petal/data/dict.db +docker compose restart petal # the handle is opened once, at startup +``` + +Expect ~450 MB. It sits inside the LUKS volume with everything else (§6). The +backups name `petal.db` explicitly rather than sweeping the data directory +(§5), so `dict.db` stays out of them — which is the right outcome and worth +keeping: it is rebuildable from public data and would otherwise dominate every +nightly snapshot. Petal never writes to it. + +### Why Chinese doesn't use it + +The zh pair stays on the embedded ECDICT gloss, deliberately. Measured on the +deployed database, DreamDict reaches a Chinese gloss for 53% of the 2,000 +commonest English words; ECDICT covers essentially all of them and is in daily +use by a real writer. `lexicon.Set.For` is where that decision lives — one +`switch`, changed the day a comparison on her actual lookups says otherwise. + +For pt-PT and French the same measurement reads 62% and 63%, which is why they +use DreamDict: there is no alternative source for them at all. + +### Rebuilding it + +Rebuilt 2026-07-27 to add Spanish (the previous file predated DreamDict's +Spanish support). The recipe, since it will be needed again: + +```bash +# on millenia, from a clean checkout of dreamdict main +./scripts/download-dict-data.sh ~/dreamdict/data # idempotent; skips what's there +go run ./cmd/dictimport --data ~/dreamdict/data --db ./dict.db --clean +``` + +~6 minutes on 32 cores; the data directory is ~7 GB and mostly already +downloaded. **Build to a new path, never over a file in use** — then verify by +hash on both ends before swapping. + +Two things worth knowing before trusting a rebuild: + +- The SUBTLEX-US download fails (the source moved behind a manual export). It + does not matter: the loader falls back to `SUBTLEX-US.txt`, which is present, + and English "frequency" is mostly SCOWL's commonness bucket anyway — + 1000/800/600/…/50, refined by SUBTLEX for only ~1,600 words. That is why the + word-difficulty chip reads `difficulty`, not `frequency`. +- Check the *other* languages' counts are unchanged before shipping. The 2026-07 + rebuild came out byte-identical for en/fr/pt-PT/zh, which is what says it + added a language rather than quietly shifting the rest. + +Gloss coverage of the 2,000 commonest English words, after the rebuild: +**es 68.6%**, fr 63.1%, pt-PT 62.1%, zh 53.2%. The startup line reports actual +per-language row counts, so a database missing a language says so. + +--- + +## 5. Backups + +### On the VPS — folded into `parodia-backup` + +Petal rides the host's existing offsite job (`/usr/local/bin/parodia-backup`, +`parodia-backup.timer`, nightly ~03:40): age-encrypted to S3, 14-day retention, +dead-man snitch. The host holds only the age *public* recipient, so it writes +backups it cannot itself decrypt. + +``` +push petal.db.age sqlite_file_dump /home/reala/petal/data/petal.db +``` + +`/home/reala/petal/.env` is in the same job's secrets tarball — it carries the +interim basic-auth hash and, from Phase 16, the OIDC client secret. + +**Why `sqlite_file_dump` and not the script's existing `sqlite_dump`:** that +helper uses Python's `iterdump`, which **does not reproduce an FTS5 virtual +table**. It emits `documents_fts` as a raw `sqlite_master` row plus its shadow +tables, and replaying the result dies with `no such table: documents_fts` — +verified by round-tripping a real dump on 2026-07-27. Petal's cross-document +search would have been silently missing after any restore. `sqlite_file_dump` +runs `VACUUM INTO` instead: a genuine database file, virtual tables intact, WAL +folded in, no write lock. Restore is a copy rather than a replay. + +> If `apply.db` ever gains a virtual table, it needs the same treatment. + +### Restore (VPS) + +```bash +age -d -i petal.db.age > /tmp/petal.db # from S3 +cd ~/petal +docker compose stop petal # stop writers first +mv data/petal.db data/petal.db.before-restore # keep the current state +rm -f data/petal.db-wal data/petal.db-shm # a stale WAL against a new file +cp /tmp/petal.db data/petal.db +docker compose start petal +docker compose logs petal --tail 5 # expect "database ready" +``` + +To sanity-check an archive before committing to it, have Petal open it in a +scratch directory — a clean exit means it reads end to end: + +```bash +mkdir -p /tmp/restore-check && cp /tmp/petal.db /tmp/restore-check/petal.db +docker run --rm -v /tmp/restore-check:/data --user "$(id -u):$(id -g)" \ + --entrypoint sh petal:local -c '/app/petal -backup /data/verify.db' +``` + +### On millenia — `petal-backup.timer` + +Until 2026-07-27 her actual writing had **no scheduled backup at all**; the +newest snapshot was a month old. It now runs nightly at 03:20 +(`Persistent=true`, because the box isn't on 24/7 and a missed window would +otherwise be skipped silently): + +```bash +sudo install -m 0644 deploy/petal-backup.service deploy/petal-backup.timer /etc/systemd/system/ +sudo systemctl daemon-reload && sudo systemctl enable --now petal-backup.timer +sudo systemctl start petal-backup.service # prove it before trusting it +``` + +`deploy/backup-petal.sh` snapshots via `petal -backup`, gzips, **age-encrypts +with the parodia public recipient**, pushes to the VPS over headscale with a +post-transfer size check, and prunes both ends. The private identity is offline, +so neither millenia nor the VPS can decrypt what it is holding — verified. + +A manual snapshot any time, no tooling required: + +```bash +cd ~/petal && ./petal -backup ~/petal/backups/manual-$(date -u +%Y%m%dT%H%M%SZ).db +``` + +Restore is a copy — stop Petal, drop the file in as `data/petal.db`, remove any +stale `-wal`/`-shm`, start. + +--- + +## 6. Encryption at rest + +### VPS — `/home/reala/petal/data` is a LUKS volume + +`deploy/setup-encrypted-data.sh` puts the data directory on LUKS2 over a sparse +file at `/var/lib/petal-crypt.img`. That covers `petal.db`, uploaded `images/`, +**and the TTS cache** — which is synthesized audio of her sentences and is easy +to forget. + +LUKS-on-a-file rather than gocryptfs because Petal is SQLite in WAL mode: WAL +needs a shared-memory index (`-shm`) mapped consistently across processes, and +FUSE has a long history of subtle mmap/locking differences. A block device with +ext4 behaves exactly like a disk to SQLite, which is the only guarantee worth +having under a database. + +**What it protects, honestly.** The key lives at `/etc/petal/dataset.key` on the +same host so the volume auto-unlocks at boot. That is a deliberate availability +tradeoff: + +| | | +|---|---| +| protects against | a decommissioned or resold disk; reading the raw block device; casual browsing of a filesystem snapshot that excludes `/etc` | +| does **not** protect against | anyone holding the whole VM image — they get the keyfile with the ciphertext; or anything at all while the host is running and mounted | + +Real protection from a provider-side snapshot needs the key off-box (fetched +over the VPN at boot). Considered, not chosen. + +Two things this setup got wrong the first time, both caught by rehearsing a +reboot rather than trusting a clean run — worth knowing if you rebuild it: + +- **Mounting over a directory hides its contents, it does not remove them.** The + first pass left the original plaintext `petal.db` and WAL sitting on the + unencrypted root filesystem, invisible under the mount. The script now shreds + the originals before mounting and refuses to continue if the mountpoint will + not come up empty. +- **`systemd-cryptsetup` was not installed**, so `/etc/crypttab` was ignored + entirely and the volume would never have unlocked at boot. The script now + refuses to run without the generator present. + +Check it any time: + +```bash +sudo ./deploy/setup-encrypted-data.sh --status +``` + +### The mount-liveness guard + +The mountpoint directory exists whether or not the volume is mounted, so a boot +where the unlock failed would start Petal against an empty unencrypted +directory and quietly serve a blank database — the failure that looks like data +loss. `data/.volume-ok` lives on the encrypted filesystem and is bind-mounted +with `create_host_path: false`, turning that into a loud container start +failure: + +``` +Error response from daemon: invalid mount config for type "bind": +bind source path does not exist: /home/reala/petal/data/.volume-ok +``` + +Verified by unmounting and attempting a start. + +**A true reboot has not been tested** — the VPS also runs matrix, lemmy, akkoma, +gitea and authentik, so rebooting it is your call. The boot path was rehearsed +through `local-fs.target`, which pulls the mount, which pulls the unlock. + +### millenia is not encrypted at rest + +LVM, no LUKS. Her canonical writing sits in plaintext on the home box. Backups +leaving it are age-encrypted; the disk itself is not. + +--- + +## 7. Supervision and monitoring + +Petal on millenia ran for months as a bare `./petal` with PPID 1 — no unit, no +screen session — so a crash or reboot left it down until someone noticed. It is +now `petal.service`: + +```bash +sudo install -m 0644 deploy/petal.service /etc/systemd/system/ +sudo systemctl daemon-reload && sudo systemctl enable --now petal.service +``` + +Verified by `kill -9`-ing it and watching systemd bring it back. + +**Piper's silent-failure mode is fixed.** Both units now set +`StartLimitIntervalSec=300` / `StartLimitBurst=5`. With `RestartSec=3` and +systemd's default 10-second window, only ~3 restarts ever landed inside it, so +the burst limit was never reached and a dead service looped **26,800+ times over +a day without ever entering `failed`**. A genuinely broken Piper now shows up in +`systemctl --user --failed`. + +### Still to do — an external probe + +Nothing yet watches millenia from outside. uptime-kuma already runs on the VPS +and can reach millenia over headscale, so the missing piece is two monitors +(they need the uptime-kuma UI, hence not scripted here): + +- `http://100.64.0.2:8088/api/health` — Petal itself +- a POST to `http://100.64.0.2:8088/api/tts` — catches a dead Piper, which + `/api/health` will not, because read-aloud degrades silently to browser + speech + +--- + +## Appendix — Piper on millenia (user systemd units) + +millenia still runs Piper as user services; these are the original notes. ```bash -# from this repo, on your workstation: scp deploy/piper.service deploy/setup-piper.sh 192.168.1.212:/tmp/ ssh 192.168.1.212 'cd /tmp && sudo ./setup-piper.sh' ``` -`setup-piper.sh` creates `~/piper/venv`, installs `piper-tts[http]`, downloads the -`en_US-amy-medium` voice into `~/piper/voices`, installs+enables `piper.service` -(loopback :5005), and smoke-tests it. Idempotent. +`setup-piper.sh` creates `~/piper/venv`, installs `piper-tts[http]`, downloads +`en_US-amy-medium` into `~/piper/voices`, installs and enables `piper.service` +(loopback `:5005`), and smoke-tests it. Idempotent. Check it with +`systemctl status piper` / `journalctl -u piper -f`. -Check it any time: `systemctl status piper`, `journalctl -u piper -f`. - -## 2. Point petal at Piper + redeploy the binary - -Add to petal's environment (its `.env` or launch env): - -``` -TTS_ENDPOINT=http://127.0.0.1:5005 -TTS_VOICE_EN=en_US-amy-medium -TTS_AUDIO_FORMAT=mp3 -``` - -Then ship the rebuilt binary (`go build -o petal ./cmd/server` already done) and -restart the petal `:8088` session. Confirm the log line: -`read-aloud enabled (TTS endpoint=http://127.0.0.1:5005)`. - -## 3. Verify - -```bash -# on millenia — end-to-end through petal, including ffmpeg transcode: -curl -sf -X POST localhost:8088/api/tts \ - -H 'Content-Type: application/json' \ - -d '{"text":"hello there","lang":"en-US"}' -o /tmp/petal-tts.mp3 \ - && file /tmp/petal-tts.mp3 # expect: Audio file ... MPEG ... layer III -``` - -- Second identical call is served from the cache (`~/petal/.../data/tts/*.mp3`). -- A language with no configured instance returns 404 → client uses Web Speech. -- Browser check via the uitest harness (`~/petal/uitest`): tap a word in the - WordCard / select a sentence and hit speak — expect the natural Piper voice. - -## Chinese voice (live) - -Each Piper HTTP server loads ONE model, so Chinese runs as a **second instance**: -`piper-zh.service` on :5006 with `zh_CN-huayan-medium`. Deployed via: +Chinese runs as a second instance (`piper-zh.service`, `:5006`, +`zh_CN-huayan-medium`): ```bash scp deploy/piper-zh.service 192.168.1.212:~/.config/systemd/user/ @@ -60,6 +576,90 @@ ssh 192.168.1.212 'export XDG_RUNTIME_DIR=/run/user/$(id -u) systemctl --user daemon-reload && systemctl --user enable --now piper-zh.service' ``` -Then in petal's `start.sh`: `TTS_ENDPOINT_ZH=http://127.0.0.1:5006` and -`TTS_VOICE_ZH=zh_CN-huayan-medium`. The handler maps language → instance from config, -so adding more languages is just another instance + env pair (no code change). +Petal's env then carries `TTS_ENDPOINT=http://127.0.0.1:5005`, +`TTS_ENDPOINT_ZH=http://127.0.0.1:5006` and the matching voice ids. The handler +maps language → instance from config, so another language is another instance +plus an env pair, no code change. + +**Adding a language (Phase 21 made this literal).** Petal discovers its Piper +instances from the environment: English is the unsuffixed +`TTS_ENDPOINT`/`TTS_VOICE_EN`, and every other language is a +`TTS_ENDPOINT_`/`TTS_VOICE_` pair. `` is the *base* tag — +`PT`, not `PT_PT`, because an environment variable name cannot hold a hyphen and +only one Portuguese model is loaded regardless. Both halves must be set: an +endpoint with no voice is dropped, so a half-finished language reads to the +browser as "no voice here, use Web Speech" instead of erroring on every tap. The +startup line names what it actually resolved: + +``` +read-aloud enabled (voices: en=en_US-amy-medium, pt=pt_PT-tugão-medium, zh=zh_CN-huayan-medium) +``` + +**Portuguese: `pt_PT-tugão-medium` is the only European voice Piper ships.** The +other five `pt_*` models in the catalogue are all Brazilian, so the voice has to +be named explicitly for the same reason the Hunspell dictionary did (Phase 21): +the obvious default is the wrong country. Check what exists before assuming: + +```bash +docker exec petal-piper-en python -c "import urllib.request,json; \ +d=json.load(urllib.request.urlopen('https://huggingface.co/rhasspy/piper-voices/resolve/main/voices.json')); \ +print([k for k in d if k.startswith('pt')])" +``` + +**French: the opposite situation, and worth knowing it is.** Every `fr_*` voice +in the catalogue is `fr_FR`, so there is no wrong country to land on by default +and no Québec voice to choose instead; `fr_FR-siwis-medium` is picked to match +the register of the other three rather than to avoid anything. The name is also +plain ASCII, so the entrypoint's percent-encoded download fallback — which +exists only because `tugão` broke `piper.download_voices` — never fires here. + +**Slow replay.** `POST /api/tts` takes `slow: true`, which raises Piper's +`length_scale` to about 4/3 (≈0.75× pace). It is a separate cache entry, not a +playback-rate trick, so the slow clip is synthesized once and then instant. + +**Piper version note:** piper-tts moved synthesis from `POST /` to +`POST /synthesize` in 1.6.0, with an identical request body. `TTS_PATH` selects +which — it defaults to `/`, and both the VPS compose and millenia's `start.sh` +now set `/synthesize`. If read-aloud starts returning 502 after a Piper upgrade, +that flag is the fix. + +**The venv is fragile across Python upgrades.** On 2026-07-27 millenia's Piper +was found dead with **26,800+ failed restarts**, silently since the Jul 26 +reboot — read-aloud had been falling back to browser Web Speech the whole time. +Root cause: an OS upgrade moved `/usr/bin/python3` from 3.13 to 3.14, and +`venv/bin/python3` is a *symlink to the system interpreter*, so the venv's +`lib/python3.13/site-packages` became invisible — `sys.path` contained no +site-packages at all. The failure surfaced as the misleading +`No module named piper.http_server` even though `http_server.py` was sitting +right there on disk. + +Fix (what was done — recreating the venv, not repairing it): + +```bash +systemctl --user stop piper.service piper-zh.service +mv ~/piper/venv ~/piper/venv.broken-py313 +python3 -m venv ~/piper/venv +~/piper/venv/bin/pip install "piper-tts[http]" +~/piper/venv/bin/python -c 'import piper.http_server' # must not raise +systemctl --user start piper.service piper-zh.service +``` + +That reinstall lands 1.6.0, so it must be paired with `TTS_PATH=/synthesize` in +`start.sh` and a binary new enough to read that variable. Voices in +`~/piper/voices` survive and do not need re-downloading. + +Worth knowing: `Restart=on-failure` will retry forever without ever alerting. +Neither service reports its health anywhere, which is why this went unnoticed +for a day. A `/api/tts` probe in uptime-kuma would have caught it. + +Verify end to end (through Petal, including the ffmpeg transcode): + +```bash +curl -sf -X POST localhost:8088/api/tts \ + -H 'Content-Type: application/json' \ + -d '{"text":"hello there","lang":"en-US"}' -o /tmp/petal-tts.mp3 \ + && file /tmp/petal-tts.mp3 # expect: MPEG ADTS, layer III +``` + +A second identical call is served from the cache; a language with no configured +instance returns 404 so the client falls back to Web Speech. diff --git a/deploy/backup-petal.sh b/deploy/backup-petal.sh new file mode 100755 index 0000000..89d4277 --- /dev/null +++ b/deploy/backup-petal.sh @@ -0,0 +1,113 @@ +#!/usr/bin/env bash +# Nightly off-box backup of Petal's database. +# +# ./backup-petal.sh # snapshot, compress, encrypt, push, prune +# ./backup-petal.sh --local-only # snapshot + prune, skip the remote push +# +# Used on millenia, driven by petal-backup.timer (see deploy/README.md). The +# VPS does not use this script -- Petal rides parodia-backup there. +# +# The snapshot goes through `petal -backup`, which uses SQLite's VACUUM INTO: +# one coherent file including anything still in the WAL, taken without a write +# lock, so it is safe against the live running app. That is why this script +# never touches petal.db / -wal / -shm directly — copying those three +# separately can capture a torn mid-checkpoint state. +# +# Everything below is overridable from the environment. +set -euo pipefail + +# Stack directory (holds docker-compose.yml and ./data). +STACK_DIR="${STACK_DIR:-$HOME/petal}" +# Where snapshots land on the VPS before being pushed off-box. Inside ./data so +# the container can write it through the existing bind mount. +LOCAL_DIR="${LOCAL_DIR:-$STACK_DIR/data/backups}" +# Off-VPS destination: millenia over headscale. Empty disables the push. +REMOTE_HOST="${REMOTE_HOST:-100.64.0.2}" +REMOTE_USER="${REMOTE_USER:-}" +REMOTE_DIR="${REMOTE_DIR:-petal-backups}" +# Retention, in days, on each side. +KEEP_LOCAL_DAYS="${KEEP_LOCAL_DAYS:-7}" +KEEP_REMOTE_DAYS="${KEEP_REMOTE_DAYS:-30}" +# age public recipient. Set it and every archive is encrypted before it leaves +# (and at rest locally too); leave it empty and the script says so loudly. +AGE_RECIPIENT="${AGE_RECIPIENT:-}" + +local_only=0 +[ "${1:-}" = "--local-only" ] && local_only=1 + +stamp="$(date -u +%Y%m%dT%H%M%SZ)" +name="petal-${stamp}.db" + +cd "$STACK_DIR" + +mkdir -p "$LOCAL_DIR" +snapshot="${LOCAL_DIR}/${name}" + +# Two deployment shapes: the VPS runs the compose stack, millenia runs a bare +# binary. Either way the snapshot goes through `petal -backup` (VACUUM INTO), +# which is safe against the live process, so neither has to stop writing. +if [ -f "$STACK_DIR/docker-compose.yml" ] && docker compose ps --status running 2>/dev/null | grep -q petal; then + echo ">> snapshotting via the running container -> data/backups/${name}" + # ./data/backups on the host is the container's /data/backups. + docker compose exec -T petal /app/petal -backup "/data/backups/${name}" +elif [ -x "$STACK_DIR/petal" ]; then + echo ">> snapshotting via the local binary -> ${snapshot}" + # DATABASE_PATH must match the running instance; start.sh is the source of + # truth for it, so read it from there rather than guessing. + DB_PATH="$(sed -n 's/^export DATABASE_PATH=//p' "$STACK_DIR/start.sh" 2>/dev/null | tail -1)" + DATABASE_PATH="${DB_PATH:-$STACK_DIR/data/petal.db}" "$STACK_DIR/petal" -backup "$snapshot" +else + echo "no way to snapshot: neither a running petal container nor $STACK_DIR/petal" >&2 + exit 1 +fi + +[ -s "$snapshot" ] || { echo "snapshot missing or empty: $snapshot" >&2; exit 1; } + +echo ">> compressing" +gzip -9 "$snapshot" +archive="${snapshot}.gz" + +# Encrypt with age when a recipient is configured. The recipient is a PUBLIC +# key -- this host can write backups it cannot itself decrypt, and the private +# identity stays offline. Same custody model as parodia-backup. Without this, +# an off-box copy is just her writing sitting in plaintext on another machine. +if [ -n "$AGE_RECIPIENT" ]; then + age -r "$AGE_RECIPIENT" -o "${archive}.age" "$archive" + shred -uz "$archive" 2>/dev/null || rm -f "$archive" + archive="${archive}.age" +else + echo " (AGE_RECIPIENT unset: this backup is NOT encrypted)" >&2 +fi +echo " $(du -h "$archive" | cut -f1) ${archive}" + +if [ "$local_only" -eq 0 ] && [ -n "$REMOTE_HOST" ]; then + target="${REMOTE_HOST}" + [ -n "$REMOTE_USER" ] && target="${REMOTE_USER}@${REMOTE_HOST}" + + echo ">> pushing to ${target}:${REMOTE_DIR}/" + ssh -o BatchMode=yes "$target" "mkdir -p '${REMOTE_DIR}'" + scp -q -o BatchMode=yes "$archive" "${target}:${REMOTE_DIR}/" + + # Verify by size rather than trusting scp's exit code alone — a truncated + # transfer that still exits 0 would leave a backup that only looks fine. + local_size="$(stat -c%s "$archive")" + remote_size="$(ssh -o BatchMode=yes "$target" "stat -c%s '${REMOTE_DIR}/$(basename "$archive")'")" + if [ "$local_size" != "$remote_size" ]; then + echo "size mismatch after transfer: local ${local_size}, remote ${remote_size}" >&2 + exit 1 + fi + echo " verified ${remote_size} bytes" + + echo ">> pruning remote copies older than ${KEEP_REMOTE_DAYS} days" + ssh -o BatchMode=yes "$target" \ + "find '${REMOTE_DIR}' \\( -name 'petal-*.db.gz' -o -name 'petal-*.db.gz.age' \\) -type f -mtime +${KEEP_REMOTE_DAYS} -delete" +elif [ "$local_only" -eq 1 ]; then + echo ">> --local-only: skipping the remote push" +else + echo ">> REMOTE_HOST is empty: skipping the remote push" >&2 +fi + +echo ">> pruning local copies older than ${KEEP_LOCAL_DAYS} days" +find "$LOCAL_DIR" \( -name 'petal-*.db.gz' -o -name 'petal-*.db.gz.age' \) -type f -mtime "+${KEEP_LOCAL_DAYS}" -delete + +echo ">> done" diff --git a/deploy/petal-backup.service b/deploy/petal-backup.service new file mode 100644 index 0000000..832b8b8 --- /dev/null +++ b/deploy/petal-backup.service @@ -0,0 +1,23 @@ +[Unit] +Description=Nightly backup of Petal's database (millenia) +Documentation=file:///home/reala/petal/deploy/README.md +After=network-online.target +Wants=network-online.target + +[Service] +Type=oneshot +User=reala +WorkingDirectory=/home/reala/petal +# Encrypted with the parodia age recipient before it leaves the box, then +# pushed to the VPS over headscale. The recipient is a public key and the +# private identity is offline, so neither millenia nor the VPS can decrypt what +# they are holding. Replaces nothing -- before this there was no scheduled +# backup of her writing at all; the newest snapshot on 2026-07-27 was a month +# old. +Environment=AGE_RECIPIENT=age19n4k55m9d50xew5vj2ehmcsf3wuj7fhgmfpckadpvcya4032q9dqrt4yjw +Environment=REMOTE_USER=reala +Environment=REMOTE_HOST=100.64.0.1 +Environment=REMOTE_DIR=petal-backups-millenia +ExecStart=/home/reala/petal/deploy/backup-petal.sh +Nice=10 +IOSchedulingClass=idle diff --git a/deploy/petal-backup.timer b/deploy/petal-backup.timer new file mode 100644 index 0000000..39b9158 --- /dev/null +++ b/deploy/petal-backup.timer @@ -0,0 +1,12 @@ +[Unit] +Description=Nightly Petal database backup (millenia) + +[Timer] +OnCalendar=*-*-* 03:20:00 +# The box is not on 24/7; without this a missed window would just be skipped +# and the backup would silently never run. +Persistent=true +RandomizedDelaySec=300 + +[Install] +WantedBy=timers.target diff --git a/deploy/petal.env.example b/deploy/petal.env.example new file mode 100644 index 0000000..bcc5c93 --- /dev/null +++ b/deploy/petal.env.example @@ -0,0 +1,82 @@ +# Petal — production environment for the parodia.dev VPS. +# Copy to the stack directory as `.env` (docker-compose.yml reads it via +# env_file) and fill in the model names. Values the image already fixes +# (PORT, DATABASE_PATH, IMAGE_DIR, TTS_CACHE_DIR, TTS endpoints) are set in +# docker-compose.yml, not here. + +# --- Routing ----------------------------------------------------------------- +# Must match the DNS A record and the Traefik Host() rule. +PETAL_HOST=petal.parodia.dev +# Absolute origin the app knows itself by. Phase 16's OIDC redirect URI is +# built from this, so it has to be the real public HTTPS origin. +BASE_URL=https://petal.parodia.dev + +# The companion's bedtime nag and the night theme read the container clock. +TZ=Europe/Lisbon + +# The container runs as this uid/gid so it can write the ./data bind mount. +# Set both to the output of `id -u` / `id -g` for the account owning the stack +# directory. Wrong values show up as "unable to open database file (14)". +PETAL_UID=1001 +PETAL_GID=1001 + +# --- Interim edge gate (delete when Phase 16 auth lands) --------------------- +# Petal has no authentication of its own yet — StaticResolver hands every +# request the same local user — so Traefik holds the door with basic auth until +# the OIDC flow exists. user:bcrypt-hash, as produced by: +# htpasswd -nbB petal 'your-password' +# /api/health is deliberately exempt (its own router) so monitoring still works. +PETAL_BASIC_AUTH= + +# --- LLM (millenia, over headscale) ------------------------------------------ +# The only cross-VPN dependency. Petal degrades warmly when it's unreachable: +# spell check, gloss, garden, search, export and read-aloud all keep working and +# the status bar shows 小助手在休息 · Petal's helper is resting. +# +# 100.64.0.2 is millenia on the headscale network. vLLM must be bound to that +# interface (NOT 0.0.0.0 — this host is public); see deploy/README.md. +LLM_BACKEND=vllm +LLM_ENDPOINT=http://100.64.0.2:8000 +LLM_MODEL= +LLM_CHAT_MODEL= +# 30s is the local-network default. Over WAN + VPN, with the voice and +# collocation passes sending a whole document, that truncates real work — the +# request is a hard deadline on Complete, and a timeout surfaces as the same +# warm 502 as an unreachable model. 90s leaves headroom without letting a +# genuinely wedged backend hang the pass forever. +LLM_TIMEOUT=90s + +# --- Read-aloud (Piper sidecars) --------------------------------------------- +# Endpoints are wired in docker-compose.yml; these pick the voice each sidecar +# loads. Changing one means recreating that container so it downloads the model. +# +# A language is routable only when both halves are set — a TTS_ENDPOINT_XX with +# no TTS_VOICE_XX reads as "no voice for this language" and the browser's own +# synthesizer takes over, rather than as an instance that errors on every +# request. Adding es is a compose service plus a pair of lines here. +# +# pt_PT-tugão-medium is the only European Portuguese voice Piper ships; every +# other pt model in the catalogue is Brazilian. French has the opposite +# property — every fr voice in the catalogue is fr_FR — so there is no wrong +# country to land on and no non-ASCII name to trip the downloader. +TTS_VOICE_EN=en_US-amy-medium +TTS_VOICE_ZH=zh_CN-huayan-medium +TTS_VOICE_PT=pt_PT-tugão-medium +TTS_VOICE_FR=fr_FR-siwis-medium +TTS_AUDIO_FORMAT=mp3 +TTS_TIMEOUT=15s + +# --- Auth (Authentik OIDC) --------------------------------------------------- +# Authentik already runs on this host. Set all three and Petal authenticates +# for itself; leave any unset and it falls back to the single `local` user +# (which on a public host means the Traefik basic-auth gate must stay). +# +# AUTHENTIK_URL is the provider's issuer, and the redirect URI to register in +# Authentik is https://petal.parodia.dev/auth/callback. +# AUTHENTIK_URL=https://auth.parodia.dev/application/o/petal/ +# AUTHENTIK_CLIENT_ID=petal +# AUTHENTIK_CLIENT_SECRET= +# +# Who may sign in: comma-separated subject ids and/or emails. Empty = anyone +# Authentik authenticates, which is wider than this instance wants. +# PETAL_ALLOWED_SUBS= diff --git a/deploy/petal.service b/deploy/petal.service new file mode 100644 index 0000000..eeb1019 --- /dev/null +++ b/deploy/petal.service @@ -0,0 +1,24 @@ +[Unit] +Description=Petal writing editor (millenia) +# Petal ran unsupervised for a long time -- a bare ./petal with PPID 1, no unit +# and no screen session -- so a crash or a reboot left it silently down until +# somebody noticed. It also wants vLLM up first, though it degrades warmly if +# the model is unreachable, so this is Wants and not Requires. +After=network-online.target vllm-chat.service +Wants=network-online.target vllm-chat.service + +[Service] +Type=simple +User=reala +WorkingDirectory=/home/reala/petal +# start.sh carries the environment (ports, LLM endpoint, Piper endpoints and +# TTS_PATH) and execs the binary, so the service supervises Petal itself rather +# than a shell wrapper. +ExecStart=/home/reala/petal/start.sh +Restart=on-failure +RestartSec=5 +StandardOutput=append:/home/reala/petal/petal.log +StandardError=append:/home/reala/petal/petal.log + +[Install] +WantedBy=multi-user.target diff --git a/deploy/piper-zh.service b/deploy/piper-zh.service index 0322457..48643b9 100644 --- a/deploy/piper-zh.service +++ b/deploy/piper-zh.service @@ -2,6 +2,14 @@ Description=Piper TTS HTTP server — Chinese voice (read-aloud backend for petal) After=network-online.target Wants=network-online.target +# Give up loudly instead of retrying forever. This service once failed 26,800+ +# times over a day without anyone noticing: RestartSec=3 means only ~3 restarts +# land inside systemd's default 10s StartLimitIntervalSec, so the default burst +# of 5 was never reached and the unit never entered `failed`. Widening the +# window to 5 minutes makes a genuinely broken Piper show up in +# `systemctl --user --failed` while still riding out transient blips. +StartLimitIntervalSec=300 +StartLimitBurst=5 [Service] Type=simple diff --git a/deploy/piper.service b/deploy/piper.service index 1a453da..dcbc0da 100644 --- a/deploy/piper.service +++ b/deploy/piper.service @@ -2,6 +2,14 @@ Description=Piper TTS HTTP server (read-aloud backend for petal) After=network-online.target Wants=network-online.target +# Give up loudly instead of retrying forever. This service once failed 26,800+ +# times over a day without anyone noticing: RestartSec=3 means only ~3 restarts +# land inside systemd's default 10s StartLimitIntervalSec, so the default burst +# of 5 was never reached and the unit never entered `failed`. Widening the +# window to 5 minutes makes a genuinely broken Piper show up in +# `systemctl --user --failed` while still riding out transient blips. +StartLimitIntervalSec=300 +StartLimitBurst=5 [Service] Type=simple diff --git a/deploy/piper/Dockerfile b/deploy/piper/Dockerfile new file mode 100644 index 0000000..8f30412 --- /dev/null +++ b/deploy/piper/Dockerfile @@ -0,0 +1,37 @@ +# Piper neural-TTS HTTP server — the read-aloud backend Petal proxies to. +# +# One image, any voice: the model is named by PIPER_VOICE at runtime and +# downloaded into the shared /voices volume on first start. Each Piper server +# loads exactly one voice, so a new language is a new service in +# docker-compose.yml, not a new image (English and Chinese today; pt-PT lands +# with the Portuguese pair). +# +# python:3.12 rather than 3.13 — piper-tts pulls onnxruntime, whose wheel +# coverage for 3.13 still lags. +FROM python:3.12-slim + +RUN pip install --no-cache-dir "piper-tts[http]" \ + && useradd -m -u 10002 piper + +ENV PIPER_VOICE=en_US-amy-medium \ + PIPER_DATA_DIR=/voices \ + PIPER_PORT=5000 + +RUN mkdir -p /voices && chown piper:piper /voices +VOLUME ["/voices"] + +COPY entrypoint.sh /usr/local/bin/entrypoint.sh +RUN chmod +x /usr/local/bin/entrypoint.sh + +USER piper +EXPOSE 5000 + +# The server has no dedicated health route, so synthesizing a single word is +# the honest check: it proves the model loaded, not just that a port is open. +HEALTHCHECK --interval=60s --timeout=20s --start-period=180s --retries=3 \ + CMD python -c "import os,urllib.request,json; \ +urllib.request.urlopen(urllib.request.Request('http://127.0.0.1:'+os.environ['PIPER_PORT']+'/synthesize', \ +data=json.dumps({'text':'ok','voice':os.environ['PIPER_VOICE']}).encode(), \ +headers={'Content-Type':'application/json'}), timeout=15).read(1)" + +ENTRYPOINT ["/usr/local/bin/entrypoint.sh"] diff --git a/deploy/piper/entrypoint.sh b/deploy/piper/entrypoint.sh new file mode 100755 index 0000000..d51caf7 --- /dev/null +++ b/deploy/piper/entrypoint.sh @@ -0,0 +1,57 @@ +#!/usr/bin/env bash +# Fetch the configured voice if the shared volume doesn't have it yet, then +# serve it. The download is the only step that needs the internet, and it runs +# once per voice for the life of the volume — Petal itself stays offline-first. +set -euo pipefail + +voice="${PIPER_VOICE:?PIPER_VOICE must be set}" +data_dir="${PIPER_DATA_DIR:-/voices}" +port="${PIPER_PORT:-5000}" + +if [ ! -f "${data_dir}/${voice}.onnx" ]; then + echo ">> downloading voice ${voice} into ${data_dir}" + # piper.download_voices cannot fetch a voice whose name isn't ASCII, and the + # only European Portuguese voice in the catalogue is pt_PT-tugão-medium: + # the downloader pastes the name straight into the request line, and + # http.client encodes that as ASCII, so it dies with UnicodeEncodeError on + # the ã before a byte leaves the container. Every pt_BR voice downloads + # fine — the failure lands precisely on the voice the pt-PT pair needs. + # + # So: try the supported path, and fall back to fetching the two files + # ourselves with the URL percent-encoded, which is all the downloader was + # missing. Same host, same files, same destination names. + python -m piper.download_voices "${voice}" --data-dir "${data_dir}" || { + echo ">> download_voices failed for ${voice}; fetching directly (non-ASCII voice name)" + python - "${voice}" "${data_dir}" <<'PY' +import json, sys, urllib.parse, urllib.request + +voice, data_dir = sys.argv[1], sys.argv[2] +BASE = "https://huggingface.co/rhasspy/piper-voices/resolve/main/" + +catalogue = json.load(urllib.request.urlopen(BASE + "voices.json", timeout=120)) +entry = catalogue.get(voice) +if entry is None: + sys.exit(f"no voice named {voice!r} in the catalogue") + +# The catalogue keys the files by repo path; only the model and its config are +# needed to serve (MODEL_CARD is licence text). +for path in entry["files"]: + if not path.endswith((".onnx", ".onnx.json")): + continue + url = BASE + urllib.parse.quote(path) + dest = f"{data_dir}/{path.rsplit('/', 1)[-1]}" + print(f">> {url} -> {dest}", flush=True) + with urllib.request.urlopen(url, timeout=600) as r, open(dest, "wb") as out: + while chunk := r.read(1 << 20): + out.write(chunk) +PY + } +fi + +echo ">> serving ${voice} on :${port}" +# 0.0.0.0 is safe here: the container sits on Petal's internal compose network +# with no published ports, so only Petal can reach it. +exec python -m piper.http_server \ + -m "${voice}" \ + --data-dir "${data_dir}" \ + --host 0.0.0.0 --port "${port}" diff --git a/deploy/setup-encrypted-data.sh b/deploy/setup-encrypted-data.sh new file mode 100755 index 0000000..3be7df3 --- /dev/null +++ b/deploy/setup-encrypted-data.sh @@ -0,0 +1,149 @@ +#!/usr/bin/env bash +# One-time setup: put Petal's data directory on an encrypted volume. +# +# sudo ./setup-encrypted-data.sh # create + migrate + persist +# sudo ./setup-encrypted-data.sh --status # report, change nothing +# +# WHAT THIS DOES AND DOES NOT PROTECT +# ----------------------------------- +# The volume auto-unlocks from a keyfile stored on the same host. That is a +# deliberate choice (availability over paranoia), and it means: +# +# protects against : a decommissioned or resold disk, someone reading the +# raw block device, casual browsing of a filesystem-level +# snapshot that does not include /etc +# does NOT protect : anyone who takes the whole VM image -- they get +# against /etc/petal/dataset.key along with the ciphertext; and +# anything at all once the host is running and mounted +# +# For real protection against a provider-side snapshot the key has to live off +# the box (fetched over the VPN at boot). That was considered and not chosen. +# +# WHY LUKS-ON-A-FILE RATHER THAN gocryptfs +# ---------------------------------------- +# Petal is SQLite in WAL mode. WAL needs a shared-memory index (-shm) mapped +# consistently across processes, and FUSE filesystems have a long history of +# subtle mmap/locking differences. A LUKS block device with ext4 on top behaves +# exactly like a normal disk to SQLite, which is the only guarantee worth having +# under a database. +set -euo pipefail + +IMG="${IMG:-/var/lib/petal-crypt.img}" +SIZE="${SIZE:-8G}" +MAPPER_NAME="${MAPPER_NAME:-petal-data}" +KEYFILE="${KEYFILE:-/etc/petal/dataset.key}" +MOUNTPOINT="${MOUNTPOINT:-/home/reala/petal/data}" +STACK_DIR="${STACK_DIR:-/home/reala/petal}" +OWNER_UID="${OWNER_UID:-1001}" +OWNER_GID="${OWNER_GID:-1001}" + +[ "$(id -u)" -eq 0 ] || { echo "must run as root" >&2; exit 1; } + +# systemd-cryptsetup ships the generator that turns /etc/crypttab into units. +# On a minimal Debian it is NOT installed, and without it crypttab is silently +# ignored -- the volume simply never unlocks at boot. Found the hard way. +if [ ! -x /usr/lib/systemd/system-generators/systemd-cryptsetup-generator ]; then + echo "!! systemd-cryptsetup-generator is missing: /etc/crypttab would be ignored at boot." + echo " install it first: apt-get install systemd-cryptsetup" + exit 1 +fi + +status() { + echo "image : $IMG $( [ -f "$IMG" ] && echo "($(du -h --apparent-size "$IMG" | cut -f1) apparent, $(du -h "$IMG" | cut -f1) on disk)" || echo "(absent)")" + echo "mapper : /dev/mapper/$MAPPER_NAME $( [ -e "/dev/mapper/$MAPPER_NAME" ] && echo "(open)" || echo "(closed)")" + echo "keyfile : $KEYFILE $( [ -f "$KEYFILE" ] && echo "(present, mode $(stat -c%a "$KEYFILE"))" || echo "(absent)")" + echo "mountpoint : $MOUNTPOINT $(mountpoint -q "$MOUNTPOINT" && echo "(mounted)" || echo "(NOT mounted)")" + grep -q "^$MAPPER_NAME " /etc/crypttab 2>/dev/null && echo "crypttab : present" || echo "crypttab : MISSING" + grep -q " $MOUNTPOINT " /etc/fstab 2>/dev/null && echo "fstab : present" || echo "fstab : MISSING" +} + +if [ "${1:-}" = "--status" ]; then status; exit 0; fi + +if [ -f "$IMG" ]; then + echo "$IMG already exists — refusing to re-create. Use --status." >&2 + exit 1 +fi + +echo ">> stopping the stack so nothing is writing to $MOUNTPOINT" +if [ -f "$STACK_DIR/docker-compose.yml" ]; then + ( cd "$STACK_DIR" && docker compose down ) +fi + +echo ">> generating keyfile $KEYFILE (root-only)" +install -d -m 0700 "$(dirname "$KEYFILE")" +if [ ! -f "$KEYFILE" ]; then + dd if=/dev/urandom of="$KEYFILE" bs=512 count=1 status=none + chmod 0400 "$KEYFILE" +fi + +echo ">> creating $SIZE sparse image at $IMG" +truncate -s "$SIZE" "$IMG" +chmod 0600 "$IMG" + +echo ">> LUKS format + open" +cryptsetup luksFormat --type luks2 --batch-mode --key-file "$KEYFILE" "$IMG" +cryptsetup luksOpen --key-file "$KEYFILE" "$IMG" "$MAPPER_NAME" + +echo ">> mkfs + mount" +mkfs.ext4 -q -L petal-data "/dev/mapper/$MAPPER_NAME" + +# Preserve whatever is already in the plaintext directory, then swap it in. +STAGING="" +if [ -d "$MOUNTPOINT" ] && [ -n "$(ls -A "$MOUNTPOINT" 2>/dev/null)" ]; then + STAGING="$(mktemp -d)" + echo ">> preserving existing plaintext data -> $STAGING" + cp -a "$MOUNTPOINT/." "$STAGING/" + + # Critical, and easy to miss: mounting over a directory HIDES its contents, + # it does not remove them. Skip this and the original plaintext petal.db sits + # on the unencrypted root filesystem forever, invisible under the mount, + # defeating the entire exercise. Clear the mountpoint before mounting. + echo ">> shredding the plaintext originals under the mountpoint" + find "$MOUNTPOINT" -mindepth 1 -type f -exec shred -uz {} + 2>/dev/null || true + find "$MOUNTPOINT" -mindepth 1 -depth -type d -exec rmdir {} + 2>/dev/null || true + [ -z "$(ls -A "$MOUNTPOINT" 2>/dev/null)" ] || { + echo "!! $MOUNTPOINT is not empty after cleanup; refusing to mount over live data" >&2 + echo " (data is preserved at $STAGING)" >&2 + exit 1 + } +fi + +mkdir -p "$MOUNTPOINT" +mount "/dev/mapper/$MAPPER_NAME" "$MOUNTPOINT" + +if [ -n "$STAGING" ]; then + echo ">> restoring data onto the encrypted volume" + cp -a "$STAGING/." "$MOUNTPOINT/" + find "$STAGING" -type f -exec shred -uz {} + 2>/dev/null || true + rm -rf "$STAGING" +fi + +# Mount-liveness sentinel: docker-compose bind-mounts this file with +# create_host_path:false, so an unmounted volume becomes a loud container start +# failure rather than Petal quietly serving an empty database. +touch "$MOUNTPOINT/.volume-ok" + +chown -R "$OWNER_UID:$OWNER_GID" "$MOUNTPOINT" + +echo ">> persisting across reboots" +# systemd-cryptsetup loop-mounts a regular file source on its own. +if ! grep -q "^$MAPPER_NAME " /etc/crypttab 2>/dev/null; then + echo "$MAPPER_NAME $IMG $KEYFILE luks,nofail" >> /etc/crypttab +fi +# nofail: a problem here must never wedge the boot of a host running half a +# dozen other services. +# x-systemd.before=docker.service is the important one: without it Docker can +# start first, find $MOUNTPOINT empty, and bring Petal up against a blank +# unencrypted directory that the real volume then hides. +if ! grep -q " $MOUNTPOINT " /etc/fstab 2>/dev/null; then + echo "/dev/mapper/$MAPPER_NAME $MOUNTPOINT ext4 defaults,nofail,x-systemd.requires=/dev/mapper/$MAPPER_NAME,x-systemd.before=docker.service 0 2" >> /etc/fstab +fi +systemctl daemon-reload + +echo ">> restarting the stack" +if [ -f "$STACK_DIR/docker-compose.yml" ]; then + ( cd "$STACK_DIR" && docker compose up -d ) +fi + +echo +status diff --git a/deploy/vllm-headscale-proxy.service b/deploy/vllm-headscale-proxy.service new file mode 100644 index 0000000..9ab0f4c --- /dev/null +++ b/deploy/vllm-headscale-proxy.service @@ -0,0 +1,40 @@ +[Unit] +Description=Expose millenia's vLLM chat server on the headscale interface only +# Why a forwarder instead of just rebinding vLLM: vllm-chat.service is shared. +# Petal, Gogobee and Open WebUI all talk to 127.0.0.1:8000, and Open WebUI keeps +# its endpoint in its own database rather than in env, so moving vLLM's bind +# address would mean editing three consumers and reloading a 35B AWQ model +# (minutes of downtime for all of them). This adds a second door instead: local +# callers keep loopback untouched, and only the headscale address gains a +# listener. Nothing about vllm-chat changes. +# +# Deliberately NOT 0.0.0.0 — this reaches a public VPS over the VPN, and the +# LAN has no business seeing an unauthenticated inference endpoint. +After=network-online.target tailscaled.service vllm-chat.service +Wants=network-online.target +BindsTo=vllm-chat.service + +[Service] +Type=simple +# fork: one child per connection, so a single client can't block the others. +# reuseaddr: survive a restart while sockets are still in TIME_WAIT. +# The bind address is millenia's headscale IP; if tailscaled hasn't brought the +# interface up yet the bind fails and Restart retries until it has. +ExecStart=/usr/bin/socat -d TCP-LISTEN:8000,bind=100.64.0.2,fork,reuseaddr TCP:127.0.0.1:8000 +Restart=always +RestartSec=5 +# Long generations hold a connection open; don't let systemd reap a healthy one. +TimeoutStopSec=10 + +# The process only shuttles bytes between two sockets — give it nothing else. +NoNewPrivileges=true +PrivateTmp=true +ProtectSystem=strict +ProtectHome=true +ProtectKernelTunables=true +ProtectControlGroups=true +RestrictAddressFamilies=AF_INET AF_INET6 +DynamicUser=true + +[Install] +WantedBy=multi-user.target diff --git a/docker-compose.yml b/docker-compose.yml new file mode 100644 index 0000000..2ee7996 --- /dev/null +++ b/docker-compose.yml @@ -0,0 +1,180 @@ +# Petal on the parodia.dev VPS. +# +# docker compose up -d --build +# +# Fronted by the host's existing Traefik (external `traefik` network, the +# `web-secure` entrypoint and the `default` cert resolver — same convention the +# other services on this box use). Petal itself never binds a host port; the +# only way in is through Traefik over HTTPS. +# +# Read-aloud runs as two sibling containers rather than host systemd services: +# each Piper HTTP server loads exactly one voice, the host has no lingering +# user session to keep systemd units alive, and keeping them on the internal +# network means the TTS ports are unreachable from anywhere but Petal. +# +# Copy deploy/petal.env.example to .env before the first `up`. + +name: petal + +services: + petal: + build: + context: . + dockerfile: Dockerfile + image: petal:local + container_name: petal + restart: unless-stopped + # ./data is a bind mount, so the image's own `petal` user (uid 10001) has no + # claim on it — the host's ownership wins and the container can't open + # petal.db. Run as whoever owns the stack directory instead. Keeping it the + # host user (rather than chowning ./data to 10001) is deliberate: the backup + # script gzips snapshots in place from the host, so the host account needs + # write access to the same directory. Still never root. + user: "${PETAL_UID:-1001}:${PETAL_GID:-1001}" + env_file: .env + environment: + # Fixed by the image layout; kept here so they're visible at a glance. + PORT: "8080" + DATABASE_PATH: /data/petal.db + IMAGE_DIR: /data/images + TTS_CACHE_DIR: /data/tts + # DreamDict's built dictionary, read-only, deployed into the data volume + # (see deploy/README.md). Absent it, word lookups fall back to the + # embedded English/Chinese datasets rather than failing. + DICT_PATH: /data/dict.db + # Piper sidecars. Each server loads one voice, so English and Chinese are + # separate containers; the handler maps language → instance from config. + TTS_ENDPOINT: http://piper-en:5000 + TTS_ENDPOINT_ZH: http://piper-zh:5000 + # A language is discovered from the TTS_ENDPOINT_/TTS_VOICE_ + # pair, so fr and es cost a service and two lines rather than a code + # change. is the base tag — an env var name can't hold pt-PT's + # hyphen, and there is one Portuguese voice loaded either way. + TTS_ENDPOINT_PT: http://piper-pt:5000 + TTS_ENDPOINT_FR: http://piper-fr:5000 + # The sidecars run piper-tts 1.6.0, which serves synthesis on + # /synthesize; millenia's older server keeps the default "/". + TTS_PATH: /synthesize + # The companion's bedtime nag and night mode read the local clock. + TZ: ${TZ:-Europe/Lisbon} + volumes: + # A bind mount, not a named volume: petal.db must be trivially reachable + # from the host for the nightly backup and for a restore. + - ./data:/data + # Mount-liveness guard. On the VPS ./data is an encrypted LUKS volume, and + # the mountpoint directory still exists when that volume is NOT mounted — + # so without this, a boot where the unlock failed would start Petal + # against an empty unencrypted directory and quietly serve a blank + # database. .volume-ok lives on the encrypted filesystem, and + # create_host_path: false turns its absence into a container start + # failure instead. Harmless elsewhere: create the file once and it is a + # no-op. See deploy/README.md §6. + - type: bind + source: ./data/.volume-ok + target: /data/.volume-ok + read_only: true + bind: + create_host_path: false + networks: + - traefik + - internal + depends_on: + - piper-en + - piper-zh + - piper-pt + labels: + traefik.enable: "true" + traefik.docker.network: traefik + traefik.http.routers.petal.rule: Host(`${PETAL_HOST:-petal.parodia.dev}`) + traefik.http.routers.petal.entrypoints: web-secure + traefik.http.routers.petal.tls: "true" + traefik.http.routers.petal.tls.certResolver: default + traefik.http.routers.petal.service: petal + # No edge gate: Petal authenticates for itself now (Authentik OIDC), so + # every /api route answers 401 without a session and the only thing served + # to an anonymous visitor is the app shell and its sign-in redirect. The + # basic-auth middleware that stood here until Phase 16 — plus the separate + # unauthenticated router /api/health needed to escape it — is gone; a + # second password in front of a real login is just one more thing to lose. + traefik.http.routers.petal.middlewares: compression@file,petal-headers + traefik.http.services.petal.loadbalancer.server.port: "8080" + # Petal is a private writing space: no framing, no sniffing, HSTS on. + traefik.http.middlewares.petal-headers.headers.customresponseheaders.Content-Security-Policy: frame-ancestors 'self' + traefik.http.middlewares.petal-headers.headers.customresponseheaders.Strict-Transport-Security: max-age=31536000; includeSubDomains + traefik.http.middlewares.petal-headers.headers.customresponseheaders.X-Content-Type-Options: nosniff + traefik.http.middlewares.petal-headers.headers.customresponseheaders.Referrer-Policy: same-origin + + piper-en: + build: + context: deploy/piper + image: petal-piper:local + container_name: petal-piper-en + restart: unless-stopped + environment: + PIPER_VOICE: ${TTS_VOICE_EN:-en_US-amy-medium} + volumes: + - piper-voices:/voices + networks: + - internal + + piper-zh: + build: + context: deploy/piper + image: petal-piper:local + container_name: petal-piper-zh + restart: unless-stopped + environment: + PIPER_VOICE: ${TTS_VOICE_ZH:-zh_CN-huayan-medium} + volumes: + - piper-voices:/voices + networks: + - internal + + # European Portuguese, for the pt-PT pair. pt_PT-tugão-medium is the *only* + # European voice in Piper's catalogue — the other five Portuguese models are + # all pt_BR — so the default anyone reaches for is the Brazilian one, exactly + # as it was with the Hunspell dictionary in Phase 21. Named here rather than + # left to the image default for that reason. + piper-pt: + build: + context: deploy/piper + image: petal-piper:local + container_name: petal-piper-pt + restart: unless-stopped + environment: + PIPER_VOICE: ${TTS_VOICE_PT:-pt_PT-tugão-medium} + volumes: + - piper-voices:/voices + networks: + - internal + + # French, for the fr pair. The opposite situation to Portuguese: every French + # voice Piper ships is fr_FR, so there is no wrong country to land on by + # default, and the name is plain ASCII so the entrypoint's percent-encoded + # fallback (added for tugão) never has to fire. siwis-medium to match the + # register of the other three. + piper-fr: + build: + context: deploy/piper + image: petal-piper:local + container_name: petal-piper-fr + restart: unless-stopped + environment: + PIPER_VOICE: ${TTS_VOICE_FR:-fr_FR-siwis-medium} + volumes: + - piper-voices:/voices + networks: + - internal + +networks: + # Created and owned by the host's Traefik stack. + traefik: + external: true + # Petal ↔ Piper only. Not reachable from the internet or the other stacks. + internal: + driver: bridge + +volumes: + # Downloaded voice models, shared read-mostly by both Piper instances so the + # same model is never fetched twice. + piper-voices: diff --git a/go.mod b/go.mod index 0935db1..e3686ef 100644 --- a/go.mod +++ b/go.mod @@ -3,7 +3,11 @@ module gitea.parodia.dev/drwily/petal go 1.25.0 require ( + github.com/coreos/go-oidc/v3 v3.20.0 github.com/go-chi/chi/v5 v5.3.0 + github.com/go-jose/go-jose/v4 v4.1.4 + github.com/prosolis/dreamdict v0.0.0-20260727163219-302a39d5c768 + golang.org/x/oauth2 v0.36.0 modernc.org/sqlite v1.53.0 ) diff --git a/go.sum b/go.sum index 860a569..95c8722 100644 --- a/go.sum +++ b/go.sum @@ -1,7 +1,11 @@ +github.com/coreos/go-oidc/v3 v3.20.0 h1:EtE0WIBHk03N+DqGkY4+UONzzZHk7amKt6IyNd7OsZE= +github.com/coreos/go-oidc/v3 v3.20.0/go.mod h1:DYCf24+ncYi+XkIH97GY1+dqoRlbaSI26KVTCI9SrY4= github.com/dustin/go-humanize v1.0.1 h1:GzkhY7T5VNhEkwH0PVJgjz+fX1rhBrR7pRT3mDkpeCY= github.com/dustin/go-humanize v1.0.1/go.mod h1:Mu1zIs6XwVuF/gI1OepvI0qD18qycQx+mFykh5fBlto= github.com/go-chi/chi/v5 v5.3.0 h1:halUjDxhshgXHMrao5bB8eNBXo/rnzwr8m5m36glehM= github.com/go-chi/chi/v5 v5.3.0/go.mod h1:R+tYY2hNuVUUjxoPtqUdgBqevM9s9njzkTLutVsOCto= +github.com/go-jose/go-jose/v4 v4.1.4 h1:moDMcTHmvE6Groj34emNPLs/qtYXRVcd6S7NHbHz3kA= +github.com/go-jose/go-jose/v4 v4.1.4/go.mod h1:x4oUasVrzR7071A4TnHLGSPpNOm2a21K9Kf04k1rs08= github.com/google/pprof v0.0.0-20250317173921-a4b03ec1a45e h1:ijClszYn+mADRFY17kjQEVQ1XRhq2/JR1M3sGqeJoxs= github.com/google/pprof v0.0.0-20250317173921-a4b03ec1a45e/go.mod h1:boTsfXsheKC2y+lKOCMpSfarhxDeIzfZG1jqGcPl3cA= github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0= @@ -12,10 +16,14 @@ github.com/mattn/go-isatty v0.0.20 h1:xfD0iDuEKnDkl03q4limB+vH+GxLEtL/jb4xVJSWWE github.com/mattn/go-isatty v0.0.20/go.mod h1:W+V8PltTTMOvKvAeJH7IuucS94S2C6jfK/D7dTCTo3Y= github.com/ncruces/go-strftime v1.0.0 h1:HMFp8mLCTPp341M/ZnA4qaf7ZlsbTc+miZjCLOFAw7w= github.com/ncruces/go-strftime v1.0.0/go.mod h1:Fwc5htZGVVkseilnfgOVb9mKy6w1naJmn9CehxcKcls= +github.com/prosolis/dreamdict v0.0.0-20260727163219-302a39d5c768 h1:+7NP78QlAVcXMviA23cMBtxwvYwDhADnABa3JBhbApI= +github.com/prosolis/dreamdict v0.0.0-20260727163219-302a39d5c768/go.mod h1:s4D+Q++6Qjq8X7/ZXEMStGXA6tScm4GadFo0+HF4BHY= github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec h1:W09IVJc94icq4NjY3clb7Lk8O1qJ8BdBEF8z0ibU0rE= github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec/go.mod h1:qqbHyh8v60DhA7CoWK5oRCqLrMHRGoxYCSS9EjAz6Eo= golang.org/x/mod v0.36.0 h1:JJjpVx6myfUsUdAzZuOSTTmRE0PfZeNWzzvKrP7amb4= golang.org/x/mod v0.36.0/go.mod h1:moc6ELqsWcOw5Ef3xVprK5ul/MvtVvkIXLziUOICjUQ= +golang.org/x/oauth2 v0.36.0 h1:peZ/1z27fi9hUOFCAZaHyrpWG5lwe0RJEEEeH0ThlIs= +golang.org/x/oauth2 v0.36.0/go.mod h1:YDBUJMTkDnJS+A4BP4eZBjCqtokkg1hODuPjwiGPO7Q= golang.org/x/sync v0.20.0 h1:e0PTpb7pjO8GAtTs2dQ6jYa5BWYlMuX047Dco/pItO4= golang.org/x/sync v0.20.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0= golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg= diff --git a/internal/auth/auth.go b/internal/auth/auth.go new file mode 100644 index 0000000..915f4e0 --- /dev/null +++ b/internal/auth/auth.go @@ -0,0 +1,77 @@ +// Package auth answers one question for every API request: who is asking? +// +// Until now Petal ran as a single hardcoded user and every query passed +// db.LocalUserID directly. That made the identity of the caller a compile-time +// constant scattered across ~35 call sites — nothing a real login could ever +// replace without touching all of them. This package moves that identity into +// the request context, resolved once by [Middleware], so handlers read the +// current user instead of naming one. +// +// The identity itself still comes from [StaticResolver] today, which returns +// the same local user for everyone. Swapping in Authentik later means writing +// one Resolver (validate the session cookie → user id) and changing the single +// line in main.go that constructs it. No handler changes. +package auth + +import ( + "context" + "net/http" + + "gitea.parodia.dev/drwily/petal/internal/httputil" +) + +// ctxKey is unexported so no other package can plant a user id in the context +// without going through [WithUser]. +type ctxKey struct{} + +// WithUser returns a copy of ctx carrying userID as the authenticated caller. +// Handlers never call this; [Middleware] does, and tests use it to build a +// request that looks authenticated. +func WithUser(ctx context.Context, userID string) context.Context { + return context.WithValue(ctx, ctxKey{}, userID) +} + +// UserID returns the authenticated user id carried by ctx, or "" if the request +// never passed through [Middleware]. +// +// Returning "" rather than panicking keeps an unauthenticated request failing +// *closed*: every query in Petal is scoped `WHERE user_id = ?`, so an empty id +// matches no rows — a missing middleware leaks nothing, it just returns empty +// results. Handlers may therefore use the value directly without checking it. +func UserID(ctx context.Context) string { + id, _ := ctx.Value(ctxKey{}).(string) + return id +} + +// Resolver maps an inbound request to the id of the user making it. Returning +// an error, or an empty id, rejects the request with a 401. +// +// This is the seam a real identity provider drops into: an Authentik resolver +// validates the session cookie and returns the user id it maps to. +type Resolver interface { + Resolve(r *http.Request) (string, error) +} + +// StaticResolver resolves every request to the same user id, ignoring the +// request entirely. It is how Petal runs today — a single-user app whose one +// user now arrives through the same path a logged-in user eventually will. +type StaticResolver string + +// Resolve implements [Resolver]. +func (s StaticResolver) Resolve(*http.Request) (string, error) { return string(s), nil } + +// Middleware resolves the caller with res and stores the result in the request +// context for [UserID]. Requests the resolver rejects — or resolves to an empty +// id — never reach the handler; they get a 401 instead. +func Middleware(res Resolver) func(http.Handler) http.Handler { + return func(next http.Handler) http.Handler { + return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) { + userID, err := res.Resolve(r) + if err != nil || userID == "" { + httputil.ErrorJSON(w, http.StatusUnauthorized, "not signed in") + return + } + next.ServeHTTP(w, r.WithContext(WithUser(r.Context(), userID))) + }) + } +} diff --git a/internal/auth/auth_test.go b/internal/auth/auth_test.go new file mode 100644 index 0000000..83abecc --- /dev/null +++ b/internal/auth/auth_test.go @@ -0,0 +1,75 @@ +package auth + +import ( + "context" + "net/http" + "net/http/httptest" + "testing" +) + +// errResolver rejects every request, standing in for a real resolver that finds +// no valid session. +type errResolver struct{ err error } + +func (e errResolver) Resolve(*http.Request) (string, error) { return "", e.err } + +func TestUserIDRoundTrip(t *testing.T) { + ctx := WithUser(context.Background(), "alice") + if got := UserID(ctx); got != "alice" { + t.Fatalf("UserID = %q, want alice", got) + } +} + +// A request that never passed through the middleware must report no user rather +// than panicking — every query is `WHERE user_id = ?`, so an empty id fails +// closed (matches nothing) instead of falling back to some default account. +func TestUserIDAbsentIsEmpty(t *testing.T) { + if got := UserID(context.Background()); got != "" { + t.Fatalf("UserID on bare context = %q, want empty", got) + } +} + +func TestMiddlewareInjectsResolvedUser(t *testing.T) { + var seen string + h := Middleware(StaticResolver("local"))(http.HandlerFunc( + func(_ http.ResponseWriter, r *http.Request) { seen = UserID(r.Context()) }, + )) + + rec := httptest.NewRecorder() + h.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/", nil)) + + if seen != "local" { + t.Fatalf("handler saw user %q, want local", seen) + } + if rec.Code != http.StatusOK { + t.Fatalf("status = %d, want 200", rec.Code) + } +} + +// Both rejection paths — an explicit error and a silent empty id — must 401 +// without ever entering the handler. The empty case matters most: a resolver +// that returns ("", nil) by mistake would otherwise hand handlers an empty user +// id, and while that fails closed at the SQL layer, it should never get there. +func TestMiddlewareRejectsUnresolved(t *testing.T) { + for name, res := range map[string]Resolver{ + "resolver error": errResolver{err: http.ErrNoCookie}, + "empty user id": StaticResolver(""), + } { + t.Run(name, func(t *testing.T) { + called := false + h := Middleware(res)(http.HandlerFunc( + func(http.ResponseWriter, *http.Request) { called = true }, + )) + + rec := httptest.NewRecorder() + h.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/", nil)) + + if called { + t.Fatal("handler ran for an unauthenticated request") + } + if rec.Code != http.StatusUnauthorized { + t.Fatalf("status = %d, want 401", rec.Code) + } + }) + } +} diff --git a/internal/auth/oidc.go b/internal/auth/oidc.go new file mode 100644 index 0000000..aa07ead --- /dev/null +++ b/internal/auth/oidc.go @@ -0,0 +1,390 @@ +package auth + +import ( + "context" + "crypto/rand" + "crypto/subtle" + "encoding/base64" + "errors" + "fmt" + "log" + "net/http" + "strings" + "sync" + + "github.com/coreos/go-oidc/v3/oidc" + "github.com/go-chi/chi/v5" + "golang.org/x/oauth2" +) + +// Temporary cookies that carry one login attempt from /auth/login to +// /auth/callback. They live for ten minutes and are cleared the moment the +// callback runs. +const ( + stateCookie = "petal_oidc_state" + nonceCookie = "petal_oidc_nonce" + pkceCookie = "petal_oidc_pkce" + + loginAttemptTTL = 600 // seconds +) + +// Options configures the OIDC client. +type Options struct { + IssuerURL string // Authentik's issuer, e.g. https://auth.example.com/application/o/petal/ + ClientID string + ClientSecret string + BaseURL string // Petal's public base URL; the redirect URI is derived from it + Allowed Allowlist +} + +// OIDC implements Petal's half of an authorization-code login against +// Authentik: /auth/login starts it, /auth/callback finishes it by provisioning +// the account and issuing a session, /auth/logout ends it. +// +// Petal is the OIDC client itself rather than trusting a proxy-injected header. +// The header approach is far less code, but it is only safe while the container +// is unreachable except through that proxy — an invariant enforced by network +// configuration, not by anything in the repository, on a public host that also +// runs half a dozen other services. Petal holds someone's private journals; it +// should be safe to expose directly. +type OIDC struct { + opts Options + sessions *SessionStore + users *UserStore + secure bool + + // The provider is discovered over the network, which means it can fail at + // startup for reasons that have nothing to do with Petal. Discovery is + // therefore lazy and retried: an Authentik outage blocks new logins but + // leaves every existing session working, since those only need the database. + mu sync.Mutex + provider *oidc.Provider + oauth *oauth2.Config + verifier *oidc.IDTokenVerifier +} + +// NewOIDC builds the login flow. It attempts discovery once so a misconfigured +// issuer shows up in the startup log rather than on the writer's first login, +// but a failure here is not fatal. +func NewOIDC(ctx context.Context, opts Options, sessions *SessionStore, users *UserStore) *OIDC { + o := &OIDC{ + opts: opts, + sessions: sessions, + users: users, + secure: strings.HasPrefix(strings.ToLower(opts.BaseURL), "https://"), + } + if err := o.discover(ctx); err != nil { + log.Printf("auth: OIDC discovery failed (%v) — login will retry on demand", err) + } + return o +} + +// RedirectURI is the callback Authentik must have registered for this client. +func (o *OIDC) RedirectURI() string { + return strings.TrimSuffix(o.opts.BaseURL, "/") + "/auth/callback" +} + +// discover resolves the provider metadata and builds the oauth2 config. +func (o *OIDC) discover(ctx context.Context) error { + o.mu.Lock() + defer o.mu.Unlock() + if o.provider != nil { + return nil + } + // The issuer is passed through exactly as configured, trailing slash and + // all: OIDC requires the discovered issuer to match the requested one + // byte-for-byte, and Authentik's ends in a slash. (go-oidc trims it itself + // when building the .well-known URL, so a slash here costs nothing.) + provider, err := oidc.NewProvider(ctx, o.opts.IssuerURL) + if err != nil { + return err + } + o.provider = provider + o.verifier = provider.Verifier(&oidc.Config{ClientID: o.opts.ClientID}) + o.oauth = &oauth2.Config{ + ClientID: o.opts.ClientID, + ClientSecret: o.opts.ClientSecret, + Endpoint: provider.Endpoint(), + RedirectURL: o.RedirectURI(), + Scopes: []string{oidc.ScopeOpenID, "profile", "email"}, + } + return nil +} + +// ready returns the discovered client, discovering it first if an earlier +// attempt failed. +func (o *OIDC) ready(ctx context.Context) (*oauth2.Config, *oidc.IDTokenVerifier, error) { + if err := o.discover(ctx); err != nil { + return nil, nil, err + } + o.mu.Lock() + defer o.mu.Unlock() + return o.oauth, o.verifier, nil +} + +// Routes mounts the login endpoints. Mount at "/auth", outside /api: these are +// browser navigations, not API calls, and they must be reachable without a +// session — that is their whole purpose. +func (o *OIDC) Routes() chi.Router { + r := chi.NewRouter() + r.Get("/login", o.login) + r.Get("/callback", o.callback) + r.Get("/logout", o.logout) + r.Post("/logout", o.logout) + return r +} + +// login starts an authorization-code flow with PKCE. +func (o *OIDC) login(w http.ResponseWriter, r *http.Request) { + // Already signed in? Don't bounce a valid session through the IdP. + if _, err := o.sessions.Resolve(r); err == nil { + http.Redirect(w, r, "/", http.StatusFound) + return + } + + conf, _, err := o.ready(r.Context()) + if err != nil { + o.page(w, http.StatusServiceUnavailable, + "登录暂时不可用", "Sign-in is unavailable right now", + "Petal 联系不上登录服务。请稍后再试。", + "Petal can't reach the sign-in service. Please try again in a moment.") + return + } + + state, err := randomToken() + if err != nil { + o.page(w, http.StatusInternalServerError, "出了点问题", "Something went wrong", "请再试一次。", "Please try again.") + return + } + nonce, err := randomToken() + if err != nil { + o.page(w, http.StatusInternalServerError, "出了点问题", "Something went wrong", "请再试一次。", "Please try again.") + return + } + pkce := oauth2.GenerateVerifier() + + // state defends the callback against CSRF (a forged callback can't know the + // cookie); nonce ties the returned ID token to this attempt; PKCE binds the + // code to this client even if it leaks in transit. + o.setTemp(w, stateCookie, state) + o.setTemp(w, nonceCookie, nonce) + o.setTemp(w, pkceCookie, pkce) + + http.Redirect(w, r, conf.AuthCodeURL(state, + oidc.Nonce(nonce), + oauth2.S256ChallengeOption(pkce), + ), http.StatusFound) +} + +// callback completes the flow: verify, allowlist, provision, issue a session. +func (o *OIDC) callback(w http.ResponseWriter, r *http.Request) { + // Expire the one-shot login cookies up front, not on the way out: every exit + // from here writes a response, and a Set-Cookie added after the header is + // written is silently dropped. They're read from the request below, so + // clearing them on the response now costs nothing. + o.clearTemp(w) + + if errParam := r.URL.Query().Get("error"); errParam != "" { + o.page(w, http.StatusForbidden, + "登录未完成", "Sign-in didn't finish", + "登录服务拒绝了这次请求。你可以再试一次。", + "The sign-in service turned that request down. You can try again.") + return + } + + state, err := r.Cookie(stateCookie) + if err != nil || state.Value == "" || + subtle.ConstantTimeCompare([]byte(state.Value), []byte(r.URL.Query().Get("state"))) != 1 { + o.page(w, http.StatusBadRequest, + "这个登录链接过期了", "That sign-in link expired", + "请回到 Petal 重新登录。", + "Head back to Petal and sign in again.") + return + } + + conf, verifier, err := o.ready(r.Context()) + if err != nil { + o.page(w, http.StatusServiceUnavailable, + "登录暂时不可用", "Sign-in is unavailable right now", + "Petal 联系不上登录服务。请稍后再试。", + "Petal can't reach the sign-in service. Please try again in a moment.") + return + } + + pkce, err := r.Cookie(pkceCookie) + if err != nil { + o.page(w, http.StatusBadRequest, "这个登录链接过期了", "That sign-in link expired", + "请回到 Petal 重新登录。", "Head back to Petal and sign in again.") + return + } + + token, err := conf.Exchange(r.Context(), r.URL.Query().Get("code"), oauth2.VerifierOption(pkce.Value)) + if err != nil { + log.Printf("auth: code exchange failed: %v", err) + o.page(w, http.StatusBadGateway, "登录没有成功", "Sign-in didn't go through", + "请再试一次。", "Please try again.") + return + } + + claims, err := o.claims(r.Context(), verifier, token) + if err != nil { + log.Printf("auth: id token rejected: %v", err) + o.page(w, http.StatusBadGateway, "登录没有成功", "Sign-in didn't go through", + "请再试一次。", "Please try again.") + return + } + + // Nonce check: this ID token must belong to the attempt that started here. + nonce, err := r.Cookie(nonceCookie) + if err != nil || subtle.ConstantTimeCompare([]byte(nonce.Value), []byte(claims.nonce)) != 1 { + o.page(w, http.StatusBadRequest, "这个登录链接过期了", "That sign-in link expired", + "请回到 Petal 重新登录。", "Head back to Petal and sign in again.") + return + } + + if !o.opts.Allowed.Permits(claims.Subject, claims.Email) { + log.Printf("auth: rejected sign-in for sub=%s email=%s (not on the allowlist)", claims.Subject, claims.Email) + o.page(w, http.StatusForbidden, + "这个 Petal 不是给你写的", "This Petal isn't yours to write in", + "你的账号是有效的,但还没有被邀请到这个 Petal。如果这是个误会,找管理员说一声就好。", + "Your account is valid, but it hasn't been invited to this Petal. If that's a mistake, a word with whoever runs it will sort it out.") + return + } + + if err := o.users.Upsert(claims.Subject, claims.Email, claims.displayName()); err != nil { + log.Printf("auth: provisioning failed: %v", err) + o.page(w, http.StatusInternalServerError, "出了点问题", "Something went wrong", + "请再试一次。", "Please try again.") + return + } + + session, err := o.sessions.Create(claims.Subject, r.UserAgent()) + if err != nil { + log.Printf("auth: session creation failed: %v", err) + o.page(w, http.StatusInternalServerError, "出了点问题", "Something went wrong", + "请再试一次。", "Please try again.") + return + } + SetSessionCookie(w, session, o.secure) + log.Printf("auth: signed in %s (%s)", claims.Email, claims.Subject) + + http.Redirect(w, r, "/", http.StatusFound) +} + +// logout revokes the session server-side and clears the cookie. Doing both +// matters: clearing only the cookie leaves a token that still works if it was +// ever captured. +func (o *OIDC) logout(w http.ResponseWriter, r *http.Request) { + if c, err := r.Cookie(SessionCookie); err == nil && c.Value != "" { + if err := o.sessions.Revoke(c.Value); err != nil { + log.Printf("auth: revoke failed: %v", err) + } + } + ClearSessionCookie(w, o.secure) + http.Redirect(w, r, "/", http.StatusFound) +} + +// idClaims is the subset of the ID token Petal cares about. +type idClaims struct { + Subject string `json:"sub"` + Email string `json:"email"` + Name string `json:"name"` + PreferredUsername string `json:"preferred_username"` + + nonce string +} + +func (c idClaims) displayName() string { + if c.Name != "" { + return c.Name + } + if c.PreferredUsername != "" { + return c.PreferredUsername + } + return c.Email +} + +// claims verifies the ID token in a token response and extracts its claims. +func (o *OIDC) claims(ctx context.Context, verifier *oidc.IDTokenVerifier, token *oauth2.Token) (idClaims, error) { + raw, ok := token.Extra("id_token").(string) + if !ok || raw == "" { + return idClaims{}, errors.New("no id_token in the token response") + } + idToken, err := verifier.Verify(ctx, raw) + if err != nil { + return idClaims{}, err + } + var claims idClaims + if err := idToken.Claims(&claims); err != nil { + return idClaims{}, err + } + if claims.Subject == "" { + claims.Subject = idToken.Subject + } + claims.nonce = idToken.Nonce + return claims, nil +} + +func (o *OIDC) setTemp(w http.ResponseWriter, name, value string) { + http.SetCookie(w, &http.Cookie{ + Name: name, + Value: value, + Path: "/auth", + HttpOnly: true, + Secure: o.secure, + SameSite: http.SameSiteLaxMode, + MaxAge: loginAttemptTTL, + }) +} + +func (o *OIDC) clearTemp(w http.ResponseWriter) { + for _, name := range []string{stateCookie, nonceCookie, pkceCookie} { + http.SetCookie(w, &http.Cookie{ + Name: name, Value: "", Path: "/auth", + HttpOnly: true, Secure: o.secure, SameSite: http.SameSiteLaxMode, MaxAge: -1, + }) + } +} + +func randomToken() (string, error) { + b := make([]byte, 24) + if _, err := rand.Read(b); err != nil { + return "", err + } + return base64.RawURLEncoding.EncodeToString(b), nil +} + +// page renders one of the flow's dead ends. Every one of them is a full stop in +// front of someone who was just trying to write, so they read as warm bilingual +// sentences rather than as a status code — the same standard as the rest of the +// app, and the reason these aren't plain http.Error calls. +func (o *OIDC) page(w http.ResponseWriter, status int, titleZH, titleEN, bodyZH, bodyEN string) { + w.Header().Set("Content-Type", "text/html; charset=utf-8") + w.WriteHeader(status) + fmt.Fprintf(w, ` + + +%s · Petal + +
+
🌸
+

%s

%s

+

%s

%s

+ 回到 Petal · Back to Petal +
+`, titleEN, titleZH, titleEN, bodyZH, bodyEN) +} diff --git a/internal/auth/oidc_test.go b/internal/auth/oidc_test.go new file mode 100644 index 0000000..5965964 --- /dev/null +++ b/internal/auth/oidc_test.go @@ -0,0 +1,402 @@ +package auth + +import ( + "context" + "crypto/rand" + "crypto/rsa" + "encoding/json" + "net/http" + "net/http/httptest" + "net/url" + "strings" + "testing" + "time" + + jose "github.com/go-jose/go-jose/v4" + + "gitea.parodia.dev/drwily/petal/internal/db" +) + +// These tests run the whole login round-trip against a stub identity provider: +// discovery, the redirect out, the callback back, and the session that comes out +// the other end. The flow is the one place in Petal where getting a detail wrong +// (an unchecked state, a nonce nobody compares) is both easy and invisible — +// everything still "works" from the browser's point of view. + +// stubIdP is a minimal OpenID provider: discovery, a JWKS, and a token endpoint +// that mints a signed ID token for whoever the test says just logged in. +type stubIdP struct { + *httptest.Server + key *rsa.PrivateKey + clientID string + // issuer as advertised by discovery and asserted in tokens. Defaults to the + // server's URL; a test can give it a trailing slash, which is what Authentik + // does and which OIDC requires to match byte-for-byte. + issuer string + + // Claims the next token exchange will assert. + sub, email, name string + // nonce echoed into the token; set from the login attempt's cookie. + nonce string + // lastForm records what Petal sent to /token, so the test can assert PKCE. + lastForm url.Values +} + +func newStubIdP(t *testing.T, clientID string) *stubIdP { + t.Helper() + key, err := rsa.GenerateKey(rand.Reader, 2048) + if err != nil { + t.Fatal(err) + } + idp := &stubIdP{key: key, clientID: clientID} + + mux := http.NewServeMux() + idp.Server = httptest.NewServer(mux) + idp.issuer = idp.URL + t.Cleanup(idp.Close) + + mux.HandleFunc("/.well-known/openid-configuration", func(w http.ResponseWriter, _ *http.Request) { + _ = json.NewEncoder(w).Encode(map[string]any{ + "issuer": idp.issuer, + "authorization_endpoint": idp.URL + "/authorize", + "token_endpoint": idp.URL + "/token", + "jwks_uri": idp.URL + "/jwks", + "id_token_signing_alg_values_supported": []string{"RS256"}, + }) + }) + + mux.HandleFunc("/jwks", func(w http.ResponseWriter, _ *http.Request) { + _ = json.NewEncoder(w).Encode(jose.JSONWebKeySet{ + Keys: []jose.JSONWebKey{{Key: key.Public(), Algorithm: "RS256", Use: "sig"}}, + }) + }) + + mux.HandleFunc("/token", func(w http.ResponseWriter, r *http.Request) { + _ = r.ParseForm() + idp.lastForm = r.PostForm + w.Header().Set("Content-Type", "application/json") + _ = json.NewEncoder(w).Encode(map[string]any{ + "access_token": "stub-access-token", + "token_type": "Bearer", + "id_token": idp.idToken(t), + }) + }) + + return idp +} + +// idToken mints a signed ID token asserting the currently configured claims. +func (idp *stubIdP) idToken(t *testing.T) string { + t.Helper() + signer, err := jose.NewSigner( + jose.SigningKey{Algorithm: jose.RS256, Key: idp.key}, + (&jose.SignerOptions{}).WithType("JWT"), + ) + if err != nil { + t.Fatal(err) + } + payload, _ := json.Marshal(map[string]any{ + "iss": idp.issuer, + "aud": idp.clientID, + "sub": idp.sub, + "email": idp.email, + "name": idp.name, + "nonce": idp.nonce, + "exp": time.Now().Add(time.Hour).Unix(), + "iat": time.Now().Unix(), + }) + signed, err := signer.Sign(payload) + if err != nil { + t.Fatal(err) + } + raw, err := signed.CompactSerialize() + if err != nil { + t.Fatal(err) + } + return raw +} + +// newFlow wires Petal's login routes to a stub provider. +func newFlow(t *testing.T, allowed Allowlist) (*stubIdP, http.Handler, *SessionStore, *UserStore) { + t.Helper() + sessions, users, _ := newStores(t) + idp := newStubIdP(t, "petal") + + o := NewOIDC(context.Background(), Options{ + IssuerURL: idp.URL, + ClientID: "petal", + ClientSecret: "shh", + BaseURL: "http://petal.test", + Allowed: allowed, + }, sessions, users) + return idp, o.Routes(), sessions, users +} + +// cookieJar collects Set-Cookie headers across the redirect chain, standing in +// for the browser that would normally carry them. +type cookieJar map[string]string + +func (j cookieJar) absorb(rec *httptest.ResponseRecorder) { + for _, c := range rec.Result().Cookies() { + if c.MaxAge < 0 || c.Value == "" { + delete(j, c.Name) + continue + } + j[c.Name] = c.Value + } +} + +func (j cookieJar) attach(r *http.Request) *http.Request { + for name, value := range j { + r.AddCookie(&http.Cookie{Name: name, Value: value}) + } + return r +} + +// start runs /auth/login and returns the redirect target plus the cookies it set. +func start(t *testing.T, flow http.Handler) (*url.URL, cookieJar) { + t.Helper() + rec := httptest.NewRecorder() + flow.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/login", nil)) + if rec.Code != http.StatusFound { + t.Fatalf("login status=%d body=%s", rec.Code, rec.Body) + } + target, err := url.Parse(rec.Header().Get("Location")) + if err != nil { + t.Fatal(err) + } + jar := cookieJar{} + jar.absorb(rec) + return target, jar +} + +func TestLoginRoundTrip(t *testing.T) { + idp, flow, sessions, users := newFlow(t, nil) + idp.sub, idp.email, idp.name = "sub-her", "her@example.com", "Her Name" + + target, jar := start(t, flow) + + // The redirect must carry everything the flow depends on later. + q := target.Query() + if q.Get("state") == "" || q.Get("nonce") == "" { + t.Fatalf("login redirect missing state/nonce: %s", target) + } + if q.Get("code_challenge") == "" || q.Get("code_challenge_method") != "S256" { + t.Fatalf("login redirect missing PKCE challenge: %s", target) + } + if q.Get("redirect_uri") != "http://petal.test/auth/callback" { + t.Fatalf("redirect_uri = %q", q.Get("redirect_uri")) + } + if jar[stateCookie] != q.Get("state") { + t.Fatal("the state cookie does not match the state sent to the provider") + } + idp.nonce = jar[nonceCookie] + + // Come back as the provider would. + rec := httptest.NewRecorder() + flow.ServeHTTP(rec, jar.attach( + httptest.NewRequest(http.MethodGet, "/callback?code=abc&state="+url.QueryEscape(q.Get("state")), nil))) + if rec.Code != http.StatusFound || rec.Header().Get("Location") != "/" { + t.Fatalf("callback status=%d location=%q body=%s", rec.Code, rec.Header().Get("Location"), rec.Body) + } + + // PKCE: the code verifier must reach the token endpoint. + if v := idp.lastForm.Get("code_verifier"); v == "" { + t.Fatal("token exchange sent no code_verifier") + } + + // The account was provisioned from the token's claims... + user, err := users.Get("sub-her") + if err != nil { + t.Fatalf("user was not provisioned: %v", err) + } + if user.Email != "her@example.com" || user.DisplayName != "Her Name" { + t.Fatalf("unexpected provisioned user %+v", user) + } + + // ...and the response carries a session that resolves to them. + jar.absorb(rec) + token := jar[SessionCookie] + if token == "" { + t.Fatal("callback issued no session cookie") + } + got, err := sessions.Resolve(withCookie(token)) + if err != nil || got != "sub-her" { + t.Fatalf("session resolved to %q (err=%v), want sub-her", got, err) + } + + // The one-shot login cookies must not linger. + for _, name := range []string{stateCookie, nonceCookie, pkceCookie} { + if jar[name] != "" { + t.Fatalf("%s survived the callback", name) + } + } + + // Signing out revokes server-side, not just in the browser. + out := httptest.NewRecorder() + flow.ServeHTTP(out, jar.attach(httptest.NewRequest(http.MethodGet, "/logout", nil))) + if out.Code != http.StatusFound { + t.Fatalf("logout status=%d", out.Code) + } + if _, err := sessions.Resolve(withCookie(token)); err == nil { + t.Fatal("the session survived signing out") + } +} + +// A callback whose state doesn't match the cookie is a forged one. +func TestCallbackRejectsBadState(t *testing.T) { + idp, flow, sessions, _ := newFlow(t, nil) + idp.sub, idp.email = "sub-her", "her@example.com" + + _, jar := start(t, flow) + idp.nonce = jar[nonceCookie] + + rec := httptest.NewRecorder() + flow.ServeHTTP(rec, jar.attach( + httptest.NewRequest(http.MethodGet, "/callback?code=abc&state=some-other-state", nil))) + if rec.Code != http.StatusBadRequest { + t.Fatalf("status=%d, want 400", rec.Code) + } + assertNoSession(t, sessions, rec) + + // And so is one with no state cookie at all. + bare := httptest.NewRecorder() + flow.ServeHTTP(bare, httptest.NewRequest(http.MethodGet, "/callback?code=abc&state=x", nil)) + if bare.Code != http.StatusBadRequest { + t.Fatalf("status=%d for a cookieless callback, want 400", bare.Code) + } +} + +// An ID token minted for a different login attempt must not be accepted, even +// though it is perfectly valid and correctly signed. +func TestCallbackRejectsReplayedNonce(t *testing.T) { + idp, flow, sessions, _ := newFlow(t, nil) + idp.sub, idp.email = "sub-her", "her@example.com" + + _, jarA := start(t, flow) + _, jarB := start(t, flow) + idp.nonce = jarB[nonceCookie] // a token belonging to the *other* attempt + + rec := httptest.NewRecorder() + flow.ServeHTTP(rec, jarA.attach( + httptest.NewRequest(http.MethodGet, "/callback?code=abc&state="+url.QueryEscape(jarA[stateCookie]), nil))) + if rec.Code != http.StatusBadRequest { + t.Fatalf("status=%d, want 400", rec.Code) + } + assertNoSession(t, sessions, rec) +} + +// Being a valid user at the identity provider is not the same as being a user +// here, and the refusal has to read like Petal rather than like a stack trace. +func TestCallbackHonoursAllowlist(t *testing.T) { + idp, flow, sessions, users := newFlow(t, ParseAllowlist("her@example.com")) + idp.sub, idp.email, idp.name = "sub-stranger", "stranger@example.com", "A Stranger" + + _, jar := start(t, flow) + idp.nonce = jar[nonceCookie] + + rec := httptest.NewRecorder() + flow.ServeHTTP(rec, jar.attach( + httptest.NewRequest(http.MethodGet, "/callback?code=abc&state="+url.QueryEscape(jar[stateCookie]), nil))) + if rec.Code != http.StatusForbidden { + t.Fatalf("status=%d, want 403", rec.Code) + } + if body := rec.Body.String(); !strings.Contains(body, "这个 Petal 不是给你写的") || + !strings.Contains(body, "isn't yours to write in") { + t.Fatalf("refusal page is not the warm bilingual one: %s", body) + } + assertNoSession(t, sessions, rec) + if _, err := users.Get("sub-stranger"); err == nil { + t.Fatal("a rejected login still provisioned an account") + } + + // The person on the list gets in through the same door. + idp.sub, idp.email, idp.name = "sub-her", "her@example.com", "Her Name" + _, jar2 := start(t, flow) + idp.nonce = jar2[nonceCookie] + ok := httptest.NewRecorder() + flow.ServeHTTP(ok, jar2.attach( + httptest.NewRequest(http.MethodGet, "/callback?code=abc&state="+url.QueryEscape(jar2[stateCookie]), nil))) + if ok.Code != http.StatusFound { + t.Fatalf("an allowed writer was turned away: status=%d body=%s", ok.Code, ok.Body) + } +} + +// The provider refusing the login (a cancelled consent, a locked account) is a +// dead end, not a session. +func TestCallbackHandlesProviderError(t *testing.T) { + _, flow, sessions, _ := newFlow(t, nil) + rec := httptest.NewRecorder() + flow.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/callback?error=access_denied", nil)) + if rec.Code != http.StatusForbidden { + t.Fatalf("status=%d, want 403", rec.Code) + } + assertNoSession(t, sessions, rec) +} + +// Authentik's issuer ends in a slash, and OIDC requires the discovered issuer to +// match the configured one byte-for-byte. Normalising it away made discovery +// fail against the real provider while every stub test still passed. +func TestDiscoveryKeepsTrailingSlashIssuer(t *testing.T) { + sessions, users, _ := newStores(t) + idp := newStubIdP(t, "petal") + idp.issuer = idp.URL + "/" + idp.sub, idp.email = "sub-her", "her@example.com" + + o := NewOIDC(context.Background(), Options{ + IssuerURL: idp.issuer, + ClientID: "petal", + ClientSecret: "shh", + BaseURL: "http://petal.test", + }, sessions, users) + flow := o.Routes() + + // A failed discovery renders the 503 "sign-in is unavailable" page instead + // of redirecting, so reaching the provider at all is the assertion. + target, jar := start(t, flow) + if !strings.HasPrefix(target.String(), idp.URL+"/authorize") { + t.Fatalf("login went to %q, want the provider's authorize endpoint", target) + } + + // And the ID token it issues, whose `iss` carries the same slash, verifies. + idp.nonce = jar[nonceCookie] + rec := httptest.NewRecorder() + flow.ServeHTTP(rec, jar.attach( + httptest.NewRequest(http.MethodGet, "/callback?code=abc&state="+url.QueryEscape(jar[stateCookie]), nil))) + if rec.Code != http.StatusFound { + t.Fatalf("callback status=%d body=%s", rec.Code, rec.Body) + } +} + +// Signing in when already signed in shouldn't bounce a good session through the +// identity provider. +func TestLoginSkipsWhenAlreadySignedIn(t *testing.T) { + _, flow, sessions, _ := newFlow(t, nil) + token, err := sessions.Create(db.LocalUserID, "") + if err != nil { + t.Fatal(err) + } + rec := httptest.NewRecorder() + flow.ServeHTTP(rec, withCookie(token)) + // withCookie builds a GET "/" request; point it at the login route. + rec = httptest.NewRecorder() + req := httptest.NewRequest(http.MethodGet, "/login", nil) + req.AddCookie(&http.Cookie{Name: SessionCookie, Value: token}) + flow.ServeHTTP(rec, req) + + if rec.Code != http.StatusFound || rec.Header().Get("Location") != "/" { + t.Fatalf("status=%d location=%q, want a redirect home", rec.Code, rec.Header().Get("Location")) + } +} + +// assertNoSession fails if a response handed out a usable session cookie. +func assertNoSession(t *testing.T, sessions *SessionStore, rec *httptest.ResponseRecorder) { + t.Helper() + for _, c := range rec.Result().Cookies() { + if c.Name == SessionCookie && c.Value != "" { + if _, err := sessions.Resolve(withCookie(c.Value)); err == nil { + t.Fatal("a rejected login was given a working session") + } + } + } +} diff --git a/internal/auth/pairlang_test.go b/internal/auth/pairlang_test.go new file mode 100644 index 0000000..c968fa1 --- /dev/null +++ b/internal/auth/pairlang_test.go @@ -0,0 +1,118 @@ +package auth + +import ( + "encoding/json" + "net/http" + "net/http/httptest" + "strings" + "testing" + + "gitea.parodia.dev/drwily/petal/internal/db" +) + +// patchMe drives UpdateMeHandler as the given user would reach it: behind the +// middleware, which is the only thing that puts an id in the context. +func patchMe(t *testing.T, users *UserStore, id, body string) *httptest.ResponseRecorder { + t.Helper() + r := httptest.NewRequest(http.MethodPatch, "/me", strings.NewReader(body)) + r = r.WithContext(WithUser(r.Context(), id)) + w := httptest.NewRecorder() + users.UpdateMeHandler()(w, r) + return w +} + +func TestSetPairLang(t *testing.T) { + _, users, _ := newStores(t) + + if err := users.SetPairLang("bob", "pt-PT"); err != nil { + t.Fatalf("set pt-PT: %v", err) + } + if u, _ := users.Get("bob"); u.PairLang != "pt-PT" { + t.Fatalf("pair_lang = %q, want pt-PT", u.PairLang) + } + + // Every pair with a langpack, not just the first one: this list and the + // frontend's PACKS are two copies of the same fact, and the day they + // disagree is the day she can pick a pair the app cannot render. + if err := users.SetPairLang("bob", "fr"); err != nil { + t.Fatalf("set fr: %v", err) + } + if u, _ := users.Get("bob"); u.PairLang != "fr" { + t.Fatalf("pair_lang = %q, want fr", u.PairLang) + } + + // And back — a writer who tries a pair and doesn't like it must be able to + // return, which is the whole reason the picker exists. + if err := users.SetPairLang("bob", "zh"); err != nil { + t.Fatalf("set zh: %v", err) + } + if u, _ := users.Get("bob"); u.PairLang != "zh" { + t.Fatalf("pair_lang = %q, want zh", u.PairLang) + } +} + +// A pair the frontend has no langpack for must not be storable. Accepting it +// would leave her looking at Chinese copy with no way back except a lucky guess. +func TestSetPairLangRejectsUnshippedPairs(t *testing.T) { + _, users, _ := newStores(t) + + // "es" is the real case here — the pair whose pack has not been written yet. + // "pt-BR" is the near-miss that matters most: a Brazilian code must not be + // quietly served European copy and a European voice. + for _, lang := range []string{"es", "pt-BR", "fr-CA", "klingon", "", " "} { + if err := users.SetPairLang("bob", lang); err == nil { + t.Fatalf("stored unshipped pair %q", lang) + } + } + if u, _ := users.Get("bob"); u.PairLang != "zh" { + t.Fatalf("a refused write still moved pair_lang to %q", u.PairLang) + } +} + +func TestSetPairLangUnknownUser(t *testing.T) { + _, users, _ := newStores(t) + if err := users.SetPairLang("nobody", "pt-PT"); err == nil { + t.Fatal("set a pair language on an account that does not exist") + } +} + +func TestUpdateMeHandler(t *testing.T) { + _, users, _ := newStores(t) + + w := patchMe(t, users, "bob", `{"pair_lang":"pt-PT"}`) + if w.Code != http.StatusOK { + t.Fatalf("status = %d, want 200 (%s)", w.Code, w.Body.String()) + } + // The whole user comes back, so the client can re-read the pair from the + // server instead of assuming its request took. + var got db.User + if err := json.Unmarshal(w.Body.Bytes(), &got); err != nil { + t.Fatalf("decode: %v", err) + } + if got.ID != "bob" || got.PairLang != "pt-PT" { + t.Fatalf("response = %+v, want bob on pt-PT", got) + } +} + +func TestUpdateMeHandlerRejects(t *testing.T) { + _, users, _ := newStores(t) + + for name, body := range map[string]string{ + "unshipped pair": `{"pair_lang":"es"}`, + "missing field": `{}`, + "not json": `pt-PT`, + } { + if w := patchMe(t, users, "bob", body); w.Code != http.StatusBadRequest { + t.Fatalf("%s: status = %d, want 400", name, w.Code) + } + } + if u, _ := users.Get("bob"); u.PairLang != "zh" { + t.Fatalf("a rejected request still moved pair_lang to %q", u.PairLang) + } + + // A caller the middleware never resolved (or whose row is gone) is a lapsed + // session, not a bad request — the client turns 401 into the sign-in overlay. + if w := patchMe(t, users, "nobody", `{"pair_lang":"pt-PT"}`); w.Code != http.StatusUnauthorized { + t.Fatalf("unknown user: status = %d, want 401", w.Code) + } +} diff --git a/internal/auth/session.go b/internal/auth/session.go new file mode 100644 index 0000000..190ea4c --- /dev/null +++ b/internal/auth/session.go @@ -0,0 +1,171 @@ +package auth + +import ( + "crypto/rand" + "crypto/sha256" + "database/sql" + "encoding/base64" + "encoding/hex" + "errors" + "net/http" + "time" +) + +// SessionCookie is the cookie carrying the opaque session token. +const SessionCookie = "petal_session" + +const ( + // sessionTTL is how long a session lives without use. Thirty days, sliding: + // every authenticated request pushes the expiry back out. An editor that + // logs you out mid-draft is hostile, and Petal auto-saves every 1.5s, so a + // surprise 401 costs real writing. + sessionTTL = 30 * 24 * time.Hour + + // sessionTTLModifier is the same span as a SQLite datetime() modifier. All + // expiry math happens inside SQLite so stored values stay canonical UTC and + // never depend on the server's local clock or on Go/SQLite parsing agreeing. + sessionTTLModifier = "+30 days" + + // sessionRenewAfter throttles the sliding extension: a session is only + // pushed forward once its expiry has drifted this far from the maximum. It + // turns "a write on every request" into "a write at most once an hour per + // session" while leaving the sliding window indistinguishable to the user. + sessionRenewAfter = "-1 hour" +) + +// ErrNoSession means the request carried no session cookie, or one that is +// unknown or expired. It is not an internal failure: the caller is simply not +// signed in. +var ErrNoSession = errors.New("no valid session") + +// SessionStore issues, validates and revokes login sessions, and is itself the +// [Resolver] the API middleware runs on. +// +// The cookie holds a random token; the table stores only its SHA-256. A dump of +// the database therefore hands an attacker no usable session — the same reason +// passwords are never stored as given. Server-side rows (rather than a signed +// stateless cookie) are what make logout and revocation actually revoke. +type SessionStore struct { + db *sql.DB +} + +// NewSessionStore returns a store backed by the given database. +func NewSessionStore(db *sql.DB) *SessionStore { return &SessionStore{db: db} } + +// Create issues a new session for userID and returns the token to put in the +// cookie. The token is never stored; only its hash is. +func (s *SessionStore) Create(userID, userAgent string) (string, error) { + raw := make([]byte, 32) + if _, err := rand.Read(raw); err != nil { + return "", err + } + token := base64.RawURLEncoding.EncodeToString(raw) + + if len(userAgent) > 256 { + userAgent = userAgent[:256] + } + _, err := s.db.Exec( + `INSERT INTO sessions (id, user_id, expires_at, user_agent) + VALUES (?, ?, datetime('now', ?), ?)`, + hashToken(token), userID, sessionTTLModifier, userAgent, + ) + if err != nil { + return "", err + } + return token, nil +} + +// Resolve implements [Resolver]: it reads the session cookie, validates it, and +// returns the user it belongs to — extending the session's life while it does. +func (s *SessionStore) Resolve(r *http.Request) (string, error) { + c, err := r.Cookie(SessionCookie) + if err != nil || c.Value == "" { + return "", ErrNoSession + } + return s.userFor(c.Value) +} + +// userFor validates a raw token and slides its expiry forward. +func (s *SessionStore) userFor(token string) (string, error) { + id := hashToken(token) + + var userID string + err := s.db.QueryRow( + `SELECT user_id FROM sessions WHERE id = ? AND expires_at > datetime('now')`, id, + ).Scan(&userID) + if errors.Is(err, sql.ErrNoRows) { + return "", ErrNoSession + } + if err != nil { + return "", err + } + + // Slide the window. Throttled, and deliberately not fatal: a failed + // extension shortens one session's life, which is no reason to reject a + // request that is otherwise perfectly authenticated. + _, _ = s.db.Exec( + `UPDATE sessions SET expires_at = datetime('now', ?) + WHERE id = ? AND expires_at < datetime('now', ?, ?)`, + sessionTTLModifier, id, sessionTTLModifier, sessionRenewAfter, + ) + return userID, nil +} + +// Revoke deletes the session behind a token. Unknown tokens are not an error — +// signing out of a session that is already gone is a success, not a failure. +func (s *SessionStore) Revoke(token string) error { + _, err := s.db.Exec(`DELETE FROM sessions WHERE id = ?`, hashToken(token)) + return err +} + +// RevokeAll deletes every session for a user, signing them out everywhere. +func (s *SessionStore) RevokeAll(userID string) error { + _, err := s.db.Exec(`DELETE FROM sessions WHERE user_id = ?`, userID) + return err +} + +// Prune removes expired rows and returns how many it deleted. Nothing depends +// on it for correctness — expired sessions are already rejected on lookup — it +// just keeps the table from accumulating dead rows forever. +func (s *SessionStore) Prune() (int64, error) { + res, err := s.db.Exec(`DELETE FROM sessions WHERE expires_at <= datetime('now')`) + if err != nil { + return 0, err + } + return res.RowsAffected() +} + +// hashToken maps a raw session token to the id stored in the table. +func hashToken(token string) string { + sum := sha256.Sum256([]byte(token)) + return hex.EncodeToString(sum[:]) +} + +// SetSessionCookie writes the session cookie. Secure is set only when Petal is +// served over https — flagging it on a plain-http dev server would make the +// browser drop the cookie and silently break local login. +func SetSessionCookie(w http.ResponseWriter, token string, secure bool) { + http.SetCookie(w, &http.Cookie{ + Name: SessionCookie, + Value: token, + Path: "/", + HttpOnly: true, + Secure: secure, + SameSite: http.SameSiteLaxMode, + MaxAge: int(sessionTTL / time.Second), + }) +} + +// ClearSessionCookie expires the session cookie in the browser. The matching +// server-side row must be revoked separately — that's the half that counts. +func ClearSessionCookie(w http.ResponseWriter, secure bool) { + http.SetCookie(w, &http.Cookie{ + Name: SessionCookie, + Value: "", + Path: "/", + HttpOnly: true, + Secure: secure, + SameSite: http.SameSiteLaxMode, + MaxAge: -1, + }) +} diff --git a/internal/auth/session_test.go b/internal/auth/session_test.go new file mode 100644 index 0000000..7860bf3 --- /dev/null +++ b/internal/auth/session_test.go @@ -0,0 +1,311 @@ +package auth + +import ( + "errors" + "net/http" + "net/http/httptest" + "path/filepath" + "strings" + "testing" + + "gitea.parodia.dev/drwily/petal/internal/db" +) + +// newStores opens a database holding two users and returns the session and user +// stores over it. +func newStores(t *testing.T) (*SessionStore, *UserStore, *db.DB) { + t.Helper() + database, err := db.Open(filepath.Join(t.TempDir(), "test.db")) + if err != nil { + t.Fatalf("open db: %v", err) + } + t.Cleanup(func() { database.Close() }) + + if _, err := database.Exec( + `INSERT INTO users (id, email, display_name) VALUES (?, ?, ?)`, + "bob", "bob@petal.local", "Bob", + ); err != nil { + t.Fatalf("seed second user: %v", err) + } + return NewSessionStore(database.DB), NewUserStore(database.DB), database +} + +// withCookie builds a request carrying a session token. +func withCookie(token string) *http.Request { + r := httptest.NewRequest(http.MethodGet, "/", nil) + r.AddCookie(&http.Cookie{Name: SessionCookie, Value: token}) + return r +} + +func TestSessionLifecycle(t *testing.T) { + sessions, _, _ := newStores(t) + + token, err := sessions.Create(db.LocalUserID, "test-agent") + if err != nil { + t.Fatalf("create: %v", err) + } + + got, err := sessions.Resolve(withCookie(token)) + if err != nil { + t.Fatalf("resolve: %v", err) + } + if got != db.LocalUserID { + t.Fatalf("resolved %q, want %q", got, db.LocalUserID) + } + + // Signing out must invalidate the token server-side, not just in the browser: + // clearing only the cookie leaves a token that still works if it ever leaked. + if err := sessions.Revoke(token); err != nil { + t.Fatalf("revoke: %v", err) + } + if _, err := sessions.Resolve(withCookie(token)); !errors.Is(err, ErrNoSession) { + t.Fatalf("revoked token still resolves (err=%v)", err) + } +} + +// A request with no cookie, or a token nobody issued, is simply not signed in. +func TestSessionRejectsUnknown(t *testing.T) { + sessions, _, _ := newStores(t) + + if _, err := sessions.Resolve(httptest.NewRequest(http.MethodGet, "/", nil)); !errors.Is(err, ErrNoSession) { + t.Fatalf("bare request err=%v, want ErrNoSession", err) + } + if _, err := sessions.Resolve(withCookie("not-a-real-token")); !errors.Is(err, ErrNoSession) { + t.Fatalf("forged token err=%v, want ErrNoSession", err) + } +} + +// The table stores a hash, so a database dump yields no usable session. +func TestSessionTokenIsNotStored(t *testing.T) { + sessions, _, database := newStores(t) + + token, err := sessions.Create(db.LocalUserID, "") + if err != nil { + t.Fatalf("create: %v", err) + } + var stored string + if err := database.QueryRow(`SELECT id FROM sessions`).Scan(&stored); err != nil { + t.Fatal(err) + } + if stored == token || strings.Contains(stored, token) { + t.Fatal("the raw session token is stored in the database") + } + if stored != hashToken(token) { + t.Fatal("stored id is not the token's hash") + } +} + +func TestSessionExpiry(t *testing.T) { + sessions, _, database := newStores(t) + + token, err := sessions.Create(db.LocalUserID, "") + if err != nil { + t.Fatalf("create: %v", err) + } + if _, err := database.Exec( + `UPDATE sessions SET expires_at = datetime('now','-1 minute')`, + ); err != nil { + t.Fatal(err) + } + + if _, err := sessions.Resolve(withCookie(token)); !errors.Is(err, ErrNoSession) { + t.Fatalf("expired session still resolves (err=%v)", err) + } + + n, err := sessions.Prune() + if err != nil { + t.Fatalf("prune: %v", err) + } + if n != 1 { + t.Fatalf("pruned %d rows, want 1", n) + } +} + +// The window slides: using a session pushes its expiry back out, so someone who +// writes in Petal every few days is never signed out mid-draft. +func TestSessionSlidesForward(t *testing.T) { + sessions, _, database := newStores(t) + + token, err := sessions.Create(db.LocalUserID, "") + if err != nil { + t.Fatalf("create: %v", err) + } + // Pretend the session has been idle for a fortnight. + if _, err := database.Exec( + `UPDATE sessions SET expires_at = datetime('now','+16 days')`, + ); err != nil { + t.Fatal(err) + } + if _, err := sessions.Resolve(withCookie(token)); err != nil { + t.Fatalf("resolve: %v", err) + } + + var extended bool + if err := database.QueryRow( + `SELECT expires_at > datetime('now','+29 days') FROM sessions`, + ).Scan(&extended); err != nil { + t.Fatal(err) + } + if !extended { + t.Fatal("using a session did not extend it") + } +} + +// Two live sessions must each resolve to their own writer — the whole point. +func TestSessionsAreNotInterchangeable(t *testing.T) { + sessions, _, _ := newStores(t) + + aliceToken, err := sessions.Create(db.LocalUserID, "") + if err != nil { + t.Fatal(err) + } + bobToken, err := sessions.Create("bob", "") + if err != nil { + t.Fatal(err) + } + + for token, want := range map[string]string{aliceToken: db.LocalUserID, bobToken: "bob"} { + got, err := sessions.Resolve(withCookie(token)) + if err != nil { + t.Fatalf("resolve: %v", err) + } + if got != want { + t.Fatalf("token resolved to %q, want %q", got, want) + } + } + + // Revoking one session leaves the other alone. + if err := sessions.Revoke(aliceToken); err != nil { + t.Fatal(err) + } + if _, err := sessions.Resolve(withCookie(bobToken)); err != nil { + t.Fatalf("bob was signed out by alice's logout: %v", err) + } + + // RevokeAll signs one writer out everywhere and nobody else. + second, _ := sessions.Create("bob", "phone") + if err := sessions.RevokeAll("bob"); err != nil { + t.Fatal(err) + } + for _, token := range []string{bobToken, second} { + if _, err := sessions.Resolve(withCookie(token)); !errors.Is(err, ErrNoSession) { + t.Fatalf("RevokeAll left a session alive (err=%v)", err) + } + } +} + +// The session store is itself the Resolver the API middleware runs on, so a +// valid cookie must carry all the way through to the handler. +func TestSessionStoreDrivesMiddleware(t *testing.T) { + sessions, _, _ := newStores(t) + token, err := sessions.Create("bob", "") + if err != nil { + t.Fatal(err) + } + + var seen string + h := Middleware(sessions)(http.HandlerFunc( + func(_ http.ResponseWriter, r *http.Request) { seen = UserID(r.Context()) }, + )) + + rec := httptest.NewRecorder() + h.ServeHTTP(rec, withCookie(token)) + if rec.Code != http.StatusOK || seen != "bob" { + t.Fatalf("status=%d user=%q, want 200/bob", rec.Code, seen) + } + + rec2 := httptest.NewRecorder() + h.ServeHTTP(rec2, httptest.NewRequest(http.MethodGet, "/", nil)) + if rec2.Code != http.StatusUnauthorized { + t.Fatalf("status=%d for a cookieless request, want 401", rec2.Code) + } +} + +func TestUserUpsert(t *testing.T) { + _, users, database := newStores(t) + + if err := users.Upsert("sub-123", "her@example.com", "Her Name"); err != nil { + t.Fatalf("upsert: %v", err) + } + user, err := users.Get("sub-123") + if err != nil { + t.Fatalf("get: %v", err) + } + if user.Email != "her@example.com" || user.DisplayName != "Her Name" { + t.Fatalf("unexpected user %+v", user) + } + if user.PairLang != "zh" { + t.Fatalf("pair_lang = %q, want the zh default", user.PairLang) + } + + // A rename upstream is reflected here; Petal's own settings are not touched. + if _, err := database.Exec(`UPDATE users SET pair_lang = 'pt-PT' WHERE id = 'sub-123'`); err != nil { + t.Fatal(err) + } + if err := users.Upsert("sub-123", "new@example.com", "New Name"); err != nil { + t.Fatalf("second upsert: %v", err) + } + user, _ = users.Get("sub-123") + if user.Email != "new@example.com" || user.DisplayName != "New Name" { + t.Fatalf("login did not refresh the profile: %+v", user) + } + if user.PairLang != "pt-PT" { + t.Fatalf("login reset pair_lang to %q", user.PairLang) + } + + // Falling back to the email keeps the sidebar from showing an empty name. + if err := users.Upsert("sub-456", "them@example.com", ""); err != nil { + t.Fatal(err) + } + if u, _ := users.Get("sub-456"); u.DisplayName != "them@example.com" { + t.Fatalf("display name = %q, want the email fallback", u.DisplayName) + } + + if err := users.Upsert("", "nobody@example.com", "Nobody"); err == nil { + t.Fatal("a login with no subject was accepted") + } +} + +// Sessions belong to their account: deleting a user takes their logins with it. +func TestSessionsCascadeWithUser(t *testing.T) { + sessions, _, database := newStores(t) + token, err := sessions.Create("bob", "") + if err != nil { + t.Fatal(err) + } + if _, err := database.Exec(`DELETE FROM users WHERE id = 'bob'`); err != nil { + t.Fatal(err) + } + if _, err := sessions.Resolve(withCookie(token)); !errors.Is(err, ErrNoSession) { + t.Fatalf("session outlived its account (err=%v)", err) + } +} + +func TestAllowlist(t *testing.T) { + // No list configured = anyone the IdP authenticates, which is the right + // default for a household instance. + if !ParseAllowlist("").Permits("anyone", "anyone@example.com") { + t.Fatal("an empty allowlist turned someone away") + } + if !ParseAllowlist(" ").Permits("anyone", "anyone@example.com") { + t.Fatal("a whitespace-only allowlist turned someone away") + } + + list := ParseAllowlist(" sub-123 , Her@Example.com ,, ") + cases := []struct { + sub, email string + want bool + }{ + {"sub-123", "someone@example.com", true}, // by subject + {"sub-999", "her@example.com", true}, // by email + {"sub-999", "HER@EXAMPLE.COM", true}, // case-insensitively + {"sub-999", "stranger@example.com", false}, // neither + {"", "", false}, // no claims at all + {"sub-1", "", false}, // a near-miss subject + } + for _, c := range cases { + if got := list.Permits(c.sub, c.email); got != c.want { + t.Fatalf("Permits(%q, %q) = %v, want %v", c.sub, c.email, got, c.want) + } + } +} diff --git a/internal/auth/users.go b/internal/auth/users.go new file mode 100644 index 0000000..bea63c0 --- /dev/null +++ b/internal/auth/users.go @@ -0,0 +1,180 @@ +package auth + +import ( + "database/sql" + "encoding/json" + "errors" + "net/http" + "strings" + + "gitea.parodia.dev/drwily/petal/internal/db" + "gitea.parodia.dev/drwily/petal/internal/httputil" +) + +// UserStore provisions and reads accounts. Petal has no signup flow: a row +// appears the first time someone Authentik vouches for signs in, and that is +// the only way one is ever created. +type UserStore struct { + db *sql.DB +} + +// NewUserStore returns a store backed by the given database. +func NewUserStore(sqlDB *sql.DB) *UserStore { return &UserStore{db: sqlDB} } + +// Upsert records the account behind an OIDC login, keyed by the issuer's +// subject id. +// +// The subject is the id — not the email, which people change and which +// Authentik does not promise is stable. Email and display name are refreshed on +// every login so a rename upstream shows up here; pair_lang is deliberately not +// touched, because it is Petal's own setting rather than the IdP's. +func (u *UserStore) Upsert(sub, email, displayName string) error { + if sub == "" { + return errors.New("oidc: empty subject") + } + if displayName == "" { + displayName = email + } + _, err := u.db.Exec( + `INSERT INTO users (id, email, display_name) VALUES (?, ?, ?) + ON CONFLICT(id) DO UPDATE SET + email = excluded.email, + display_name = excluded.display_name`, + sub, email, displayName, + ) + return err +} + +// Get loads one account. +func (u *UserStore) Get(id string) (db.User, error) { + var user db.User + err := u.db.QueryRow( + `SELECT id, email, COALESCE(display_name, ''), created_at, pair_lang + FROM users WHERE id = ?`, id, + ).Scan(&user.ID, &user.Email, &user.DisplayName, &user.CreatedAt, &user.PairLang) + return user, err +} + +// MeHandler reports who the caller is. The frontend uses it to namespace +// per-account browser state and to show the signed-in writer; it sits behind +// the auth middleware, so reaching it at all already proves a valid session. +func (u *UserStore) MeHandler() http.HandlerFunc { + return func(w http.ResponseWriter, r *http.Request) { + user, err := u.Get(UserID(r.Context())) + if err != nil { + httputil.ErrorJSON(w, http.StatusUnauthorized, "not signed in") + return + } + httputil.WriteJSON(w, http.StatusOK, user) + } +} + +// The pairs a writer may actually choose, in the order the picker offers them. +// +// This is deliberately *not* internal/llm's list of languages. That one names +// every pair the prompts know how to talk about, which is a cheap thing to add; +// this one names the pairs Petal can render itself in, which requires a langpack +// on the frontend. Accepting a code with no pack would leave her looking at +// Chinese with no way back except another guess, so the server refuses it. es +// joins this list on the day its pack lands, not before. +var shippedPairs = []string{"zh", "pt-PT", "fr"} + +func pairIsShipped(lang string) bool { + for _, p := range shippedPairs { + if p == lang { + return true + } + } + return false +} + +// SetPairLang moves an account to another (English + X) pair. +func (u *UserStore) SetPairLang(id, lang string) error { + if !pairIsShipped(lang) { + return errors.New("auth: unshipped pair language " + lang) + } + res, err := u.db.Exec(`UPDATE users SET pair_lang = ? WHERE id = ?`, lang, id) + if err != nil { + return err + } + if n, err := res.RowsAffected(); err == nil && n == 0 { + return sql.ErrNoRows + } + return nil +} + +// UpdateMeHandler changes the caller's own settings — today, the one setting +// there is: which language Petal speaks alongside her English. +// +// It answers with the whole updated user rather than an empty 204 so the client +// has one shape to trust: /api/me and this return the same thing, and the app +// re-reads the pair from the response instead of assuming its request took. +// +// The pair language reaches further than the UI copy — it picks her Hunspell +// dictionary, her read-aloud voice, which word-lookup provider answers, and the +// language the prompts ask the model to explain in. All of those read +// `users.pair_lang` at use time, so all of them follow from this one write. +func (u *UserStore) UpdateMeHandler() http.HandlerFunc { + return func(w http.ResponseWriter, r *http.Request) { + var body struct { + PairLang string `json:"pair_lang"` + } + if err := json.NewDecoder(r.Body).Decode(&body); err != nil { + httputil.BadRequest(w, "invalid request body") + return + } + lang := strings.TrimSpace(body.PairLang) + if !pairIsShipped(lang) { + // Name the ones that work. A writer who lands here has picked from a + // stale client, and "not a language" tells her nothing. + httputil.BadRequest(w, "unsupported language pair — Petal speaks "+strings.Join(shippedPairs, ", ")) + return + } + id := UserID(r.Context()) + if err := u.SetPairLang(id, lang); err != nil { + if errors.Is(err, sql.ErrNoRows) { + httputil.ErrorJSON(w, http.StatusUnauthorized, "not signed in") + return + } + httputil.ServerError(w, err) + return + } + user, err := u.Get(id) + if err != nil { + httputil.ServerError(w, err) + return + } + httputil.WriteJSON(w, http.StatusOK, user) + } +} + +// Allowlist decides which of Authentik's users may write in this Petal. +// Authentik fronts several applications; being a valid user there does not mean +// being a user here. +// +// An entry matches a subject id or an email address, case-insensitively. Both +// are accepted on purpose: a subject is an opaque uuid nobody can know before +// that person's first login, so a subject-only list means the operator must let +// someone in, read a log line, and edit config — whereas an email is knowable in +// advance. An empty list allows everyone the IdP authenticates, which is the +// right default for a single-household instance. +type Allowlist map[string]bool + +// ParseAllowlist builds an Allowlist from a comma-separated env value. +func ParseAllowlist(raw string) Allowlist { + list := Allowlist{} + for _, part := range strings.Split(raw, ",") { + if p := strings.ToLower(strings.TrimSpace(part)); p != "" { + list[p] = true + } + } + return list +} + +// Permits reports whether this login may proceed. +func (a Allowlist) Permits(sub, email string) bool { + if len(a) == 0 { + return true + } + return a[strings.ToLower(sub)] || (email != "" && a[strings.ToLower(email)]) +} diff --git a/internal/config/config.go b/internal/config/config.go index 392f482..fbc448d 100644 --- a/internal/config/config.go +++ b/internal/config/config.go @@ -2,6 +2,7 @@ package config import ( "os" + "strings" "time" ) @@ -12,6 +13,12 @@ type Config struct { BaseURL string DatabasePath string ImageDir string // on-disk store for editor image uploads + // DictPath is DreamDict's built dict.db, read-only, sitting beside + // petal.db. It is what gives Petal French, European Portuguese and Spanish + // word lookups; without it only the embedded English/Chinese datasets + // exist, which is exactly how a laptop checkout runs. A missing file is + // therefore not an error — see lexicon.OpenDreamDict. + DictPath string // LLM LLMBackend string // "vllm" | "ollama" @@ -23,19 +30,49 @@ type Config struct { // TTS (read-aloud). Off unless TTSEndpoint is set — when empty, the /api/tts // route isn't mounted and the frontend falls back to the browser's Web Speech // API. Endpoint points at a local Piper HTTP server. - TTSEndpoint string // Piper instance serving the English voice - TTSEndpointZH string // Piper instance serving the Chinese voice; empty = zh falls back to Web Speech - TTSVoiceEN string // Piper voice id for English (e.g. en_US-amy-medium) - TTSVoiceZH string // Piper voice id for Chinese (e.g. zh_CN-huayan-medium) + TTSEndpoint string // Piper instance serving the English voice; also the on/off switch + // TTSVoices is every language Petal can read aloud, keyed by base language + // tag ("en", "zh", "pt", …). Each Piper server loads exactly one model, so + // a language *is* an instance — and the instances are discovered from the + // environment rather than named in this struct: one + // TTS_ENDPOINT_/TTS_VOICE_ pair per language, so the fr and es + // pairs cost a compose service and two lines of .env rather than a code + // change. English keeps the unsuffixed TTS_ENDPOINT/TTS_VOICE_EN it has + // always had. + TTSVoices map[string]TTSVoice + // TTSPath is the path Piper serves synthesis on. Piper moved it from "/" to + // "/synthesize" in 1.6.0 with an unchanged request body, so this is a + // version knob, not a feature: millenia's older server keeps the default, + // the containerised sidecars set "/synthesize". + TTSPath string TTSCacheDir string // on-disk store for synthesized clips (content-addressed) TTSTimeout time.Duration TTSFormat string // mp3 | opus | wav — mp3/opus transcode Piper's WAV via ffmpeg - // Auth (deferred — not wired in the local-dev build, kept for later) - AuthentikURL string + // Auth. OIDC against Authentik. Login is enabled only when the issuer, the + // client id and the secret are all present; with any of them missing Petal + // falls back to the single hardcoded local user, which is what local + // development wants and what every deployment did before Phase 16. + AuthentikURL string // issuer URL of the Petal provider in Authentik AuthentikClientID string AuthentikClientSecret string - SessionSecret string + // AllowedSubs gates who may sign in, as a comma-separated list of OIDC + // subject ids and/or email addresses. Empty means everyone Authentik + // authenticates — right for a single-household instance, wrong the moment + // the IdP serves an audience wider than Petal's. + AllowedSubs string +} + +// TTSVoice is one Piper instance and the single voice it has loaded. +type TTSVoice struct { + Endpoint string + Voice string +} + +// AuthEnabled reports whether real logins are configured. When false, Petal +// resolves every request to the local user. +func (c *Config) AuthEnabled() bool { + return c.AuthentikURL != "" && c.AuthentikClientID != "" && c.AuthentikClientSecret != "" } // Load reads configuration from the environment, applying sane local-dev defaults. @@ -45,6 +82,7 @@ func Load() *Config { BaseURL: env("BASE_URL", "http://localhost:8080"), DatabasePath: env("DATABASE_PATH", "./data/petal.db"), ImageDir: env("IMAGE_DIR", "./data/images"), + DictPath: env("DICT_PATH", "./data/dict.db"), LLMBackend: env("LLM_BACKEND", "vllm"), LLMEndpoint: env("LLM_ENDPOINT", "http://localhost:8000"), @@ -52,10 +90,9 @@ func Load() *Config { LLMChatModel: env("LLM_CHAT_MODEL", ""), LLMTimeout: envDuration("LLM_TIMEOUT", 30*time.Second), - TTSEndpoint: env("TTS_ENDPOINT", ""), - TTSEndpointZH: env("TTS_ENDPOINT_ZH", ""), - TTSVoiceEN: env("TTS_VOICE_EN", "en_US-amy-medium"), - TTSVoiceZH: env("TTS_VOICE_ZH", "zh_CN-huayan-medium"), + TTSEndpoint: env("TTS_ENDPOINT", ""), + TTSVoices: ttsVoices(os.Environ()), + TTSPath: env("TTS_PATH", "/"), TTSCacheDir: env("TTS_CACHE_DIR", "./data/tts"), TTSTimeout: envDuration("TTS_TIMEOUT", 15*time.Second), TTSFormat: env("TTS_AUDIO_FORMAT", "mp3"), @@ -63,10 +100,68 @@ func Load() *Config { AuthentikURL: env("AUTHENTIK_URL", ""), AuthentikClientID: env("AUTHENTIK_CLIENT_ID", ""), AuthentikClientSecret: env("AUTHENTIK_CLIENT_SECRET", ""), - SessionSecret: env("SESSION_SECRET", "dev-insecure-secret-change-me"), + AllowedSubs: env("PETAL_ALLOWED_SUBS", ""), } } +// ttsVoices reads the Piper instances out of an environment slice (as returned +// by os.Environ) into a map keyed by base language tag. +// +// English is the unsuffixed pair, TTS_ENDPOINT + TTS_VOICE_EN, because that is +// what every deployment already sets and read-aloud has always been English +// first. Every other language is a TTS_ENDPOINT_/TTS_VOICE_ pair, +// discovered rather than enumerated — TTS_ENDPOINT_ZH is what millenia and the +// VPS already use, and TTS_ENDPOINT_PT is all the Portuguese pair needs. +// +// is the *base* tag: an environment variable name cannot hold the hyphen +// in "pt-PT", and the handler routes on the base tag anyway (a request for +// pt-PT, pt-BR or bare pt reaches the same instance, because there is only one +// Portuguese voice loaded). A pair is ignored unless both halves are set: half +// a configuration should read as "no voice for this language" and fall back to +// the browser, not as an instance that answers every request with an error. +func ttsVoices(environ []string) map[string]TTSVoice { + vals := make(map[string]string, len(environ)) + for _, kv := range environ { + if k, v, ok := strings.Cut(kv, "="); ok { + vals[k] = v + } + } + + voices := map[string]TTSVoice{} + add := func(lang, endpoint, voice string) { + endpoint = strings.TrimRight(strings.TrimSpace(endpoint), "/") + voice = strings.TrimSpace(voice) + if endpoint == "" || voice == "" { + return + } + voices[lang] = TTSVoice{Endpoint: endpoint, Voice: voice} + } + + // The two languages that shipped before this was a map keep their voice + // defaults, so an existing deployment that names only the endpoints (as + // millenia's unit does) sounds exactly as it did. + voiceOr := func(key, fallback string) string { + if v := strings.TrimSpace(vals[key]); v != "" { + return v + } + return fallback + } + + add("en", vals["TTS_ENDPOINT"], voiceOr("TTS_VOICE_EN", "en_US-amy-medium")) + for k, endpoint := range vals { + suffix, ok := strings.CutPrefix(k, "TTS_ENDPOINT_") + if !ok || suffix == "" { + continue + } + voice := vals["TTS_VOICE_"+suffix] + if suffix == "ZH" { + voice = voiceOr("TTS_VOICE_ZH", "zh_CN-huayan-medium") + } + add(strings.ToLower(suffix), endpoint, voice) + } + return voices +} + func env(key, fallback string) string { if v := os.Getenv(key); v != "" { return v diff --git a/internal/config/config_test.go b/internal/config/config_test.go new file mode 100644 index 0000000..c75d616 --- /dev/null +++ b/internal/config/config_test.go @@ -0,0 +1,88 @@ +package config + +import "testing" + +// The Piper instances are discovered from the environment rather than named in +// code, so that a new pair costs a compose service and two .env lines. These +// assert the discovery rule, including the two shapes that already exist in the +// wild (millenia's systemd unit and the VPS compose file). +func TestTTSVoicesDiscovery(t *testing.T) { + voices := ttsVoices([]string{ + "TTS_ENDPOINT=http://piper-en:5000", + "TTS_VOICE_EN=en_US-amy-medium", + "TTS_ENDPOINT_ZH=http://piper-zh:5000/", + "TTS_VOICE_ZH=zh_CN-huayan-medium", + "TTS_ENDPOINT_PT=http://piper-pt:5000", + "TTS_VOICE_PT=pt_PT-tugão-medium", + "TTS_ENDPOINT_FR=http://piper-fr:5000", + "TTS_VOICE_FR=fr_FR-siwis-medium", + // Noise that must not become a language. + "TTS_PATH=/synthesize", + "PATH=/usr/bin", + }) + + want := map[string]TTSVoice{ + "en": {"http://piper-en:5000", "en_US-amy-medium"}, + // The trailing slash is trimmed here so the synthesis path concatenates + // cleanly rather than producing a double slash at every call site. + "zh": {"http://piper-zh:5000", "zh_CN-huayan-medium"}, + "pt": {"http://piper-pt:5000", "pt_PT-tugão-medium"}, + // Phase 24's whole TTS change: a fourth language costs two lines here + // and a compose service, and no Go at all. + "fr": {"http://piper-fr:5000", "fr_FR-siwis-medium"}, + } + if len(voices) != len(want) { + t.Fatalf("discovered %v, want %v", voices, want) + } + for lang, w := range want { + if voices[lang] != w { + t.Errorf("%s = %+v, want %+v", lang, voices[lang], w) + } + } +} + +// Half a configuration is not a language. An endpoint with no voice (or the +// reverse) must read as "no voice for this language" — a 404 the client answers +// by falling back to Web Speech — rather than as an instance that exists and +// errors on every request. +func TestTTSVoicesIgnoresHalfConfiguredLanguages(t *testing.T) { + voices := ttsVoices([]string{ + "TTS_ENDPOINT=http://piper-en:5000", + "TTS_VOICE_EN=en_US-amy-medium", + "TTS_ENDPOINT_FR=http://piper-fr:5000", // no TTS_VOICE_FR + "TTS_VOICE_ES=es_ES-davefx-medium", // no TTS_ENDPOINT_ES + }) + if _, ok := voices["fr"]; ok { + t.Errorf("fr routed with no voice configured") + } + if _, ok := voices["es"]; ok { + t.Errorf("es routed with no endpoint configured") + } + if len(voices) != 1 { + t.Errorf("discovered %v, want English only", voices) + } +} + +// A deployment that predates the map names only the endpoints and relies on the +// voice defaults; it must sound exactly as it did. +func TestTTSVoicesKeepsTheOriginalDefaults(t *testing.T) { + voices := ttsVoices([]string{ + "TTS_ENDPOINT=http://127.0.0.1:5005", + "TTS_ENDPOINT_ZH=http://127.0.0.1:5006", + }) + if got := voices["en"].Voice; got != "en_US-amy-medium" { + t.Errorf("en voice = %q, want the default", got) + } + if got := voices["zh"].Voice; got != "zh_CN-huayan-medium" { + t.Errorf("zh voice = %q, want the default", got) + } +} + +// Read-aloud is off when no English instance is configured; nothing else may +// switch it on. (tts.New gates on TTSEndpoint, so a stray TTS_ENDPOINT_PT with +// no English sibling must not produce a routable map that outlives that gate.) +func TestTTSVoicesEmptyWithoutEndpoints(t *testing.T) { + if voices := ttsVoices([]string{"TTS_VOICE_EN=en_US-amy-medium"}); len(voices) != 0 { + t.Errorf("discovered %v, want none", voices) + } +} diff --git a/internal/db/backup.go b/internal/db/backup.go new file mode 100644 index 0000000..bd77d66 --- /dev/null +++ b/internal/db/backup.go @@ -0,0 +1,79 @@ +package db + +import ( + "database/sql" + "fmt" + "os" + "path/filepath" +) + +// Backup writes a consistent copy of the database at srcPath to destPath using +// SQLite's `VACUUM INTO`. +// +// Why not copy the file: Petal runs in WAL mode, so at any instant the newest +// committed pages may live in petal.db-wal rather than petal.db. Copying the +// three files separately can capture them mid-checkpoint and produce a backup +// that is subtly torn. `VACUUM INTO` runs inside a read transaction, so it sees +// one coherent snapshot including the WAL, and emits a single defragmented file +// with no -wal/-shm companions — exactly what you want to ship off-box. +// +// It takes no write lock, so this is safe to run against the live database +// while someone is writing. +// +// destPath must not already exist: SQLite refuses to overwrite, which keeps a +// failed run from destroying the previous good backup. +func Backup(srcPath, destPath string) error { + if _, err := os.Stat(srcPath); err != nil { + return fmt.Errorf("source database: %w", err) + } + if _, err := os.Stat(destPath); err == nil { + return fmt.Errorf("destination %s already exists", destPath) + } + if dir := filepath.Dir(destPath); dir != "" && dir != "." { + if err := os.MkdirAll(dir, 0o755); err != nil { + return fmt.Errorf("create backup dir: %w", err) + } + } + + // Opened directly rather than through Open: a backup must never migrate or + // seed the database it is copying. + conn, err := sql.Open("sqlite", dsn(srcPath)) + if err != nil { + return fmt.Errorf("open source: %w", err) + } + defer conn.Close() + conn.SetMaxOpenConns(1) + + if err := conn.Ping(); err != nil { + return fmt.Errorf("ping source: %w", err) + } + + // The path is interpolated because VACUUM INTO takes a literal, not a bound + // parameter. Quotes are doubled so a path containing one can't break out. + quoted := "'" + escapeSQLiteString(destPath) + "'" + if _, err := conn.Exec("VACUUM INTO " + quoted); err != nil { + return fmt.Errorf("vacuum into %s: %w", destPath, err) + } + + // A zero-byte result would mean the vacuum silently produced nothing; catch + // it here rather than discovering it during a restore. + info, err := os.Stat(destPath) + if err != nil { + return fmt.Errorf("stat backup: %w", err) + } + if info.Size() == 0 { + return fmt.Errorf("backup %s is empty", destPath) + } + return nil +} + +func escapeSQLiteString(s string) string { + out := make([]byte, 0, len(s)) + for i := 0; i < len(s); i++ { + if s[i] == '\'' { + out = append(out, '\'') + } + out = append(out, s[i]) + } + return string(out) +} diff --git a/internal/db/backup_test.go b/internal/db/backup_test.go new file mode 100644 index 0000000..894e204 --- /dev/null +++ b/internal/db/backup_test.go @@ -0,0 +1,108 @@ +package db + +import ( + "database/sql" + "os" + "path/filepath" + "testing" +) + +// The point of VACUUM INTO over a file copy is that it captures rows still +// sitting in the WAL. This writes with the source connection open (so the WAL +// is hot and unlikely to have been checkpointed) and asserts the backup has +// them. +func TestBackupCapturesLiveWrites(t *testing.T) { + dir := t.TempDir() + src := filepath.Join(dir, "petal.db") + dest := filepath.Join(dir, "backups", "petal-backup.db") + + d, err := Open(src) + if err != nil { + t.Fatalf("open source: %v", err) + } + defer d.Close() + + if _, err := d.Exec( + `INSERT INTO documents (id, user_id, title, content_text) VALUES ('d1', ?, '春天', 'hello 春天')`, + LocalUserID, + ); err != nil { + t.Fatalf("insert: %v", err) + } + + if err := Backup(src, dest); err != nil { + t.Fatalf("backup: %v", err) + } + + // VACUUM INTO emits a single self-contained file — no -wal/-shm to ship + // alongside it. + for _, suffix := range []string{"-wal", "-shm"} { + if _, err := os.Stat(dest + suffix); err == nil { + t.Errorf("backup left a %s companion file behind", suffix) + } + } + + copyConn, err := sql.Open("sqlite", dsn(dest)) + if err != nil { + t.Fatalf("open backup: %v", err) + } + defer copyConn.Close() + + var title string + if err := copyConn.QueryRow(`SELECT title FROM documents WHERE id = 'd1'`).Scan(&title); err != nil { + t.Fatalf("row missing from backup: %v", err) + } + if title != "春天" { + t.Errorf("title = %q, want 春天", title) + } + + // The seeded user has to come across too, or a restore would orphan every + // document's foreign key. + var users int + if err := copyConn.QueryRow(`SELECT COUNT(*) FROM users WHERE id = ?`, LocalUserID).Scan(&users); err != nil { + t.Fatalf("count users: %v", err) + } + if users != 1 { + t.Errorf("users in backup = %d, want 1", users) + } +} + +// A second run to the same path must fail loudly rather than clobber or +// half-write the previous good backup. +func TestBackupRefusesExistingDestination(t *testing.T) { + dir := t.TempDir() + src := filepath.Join(dir, "petal.db") + dest := filepath.Join(dir, "petal-backup.db") + + d, err := Open(src) + if err != nil { + t.Fatalf("open source: %v", err) + } + defer d.Close() + + if err := Backup(src, dest); err != nil { + t.Fatalf("first backup: %v", err) + } + before, err := os.ReadFile(dest) + if err != nil { + t.Fatalf("read backup: %v", err) + } + + if err := Backup(src, dest); err == nil { + t.Fatal("second backup to the same path succeeded; want an error") + } + + after, err := os.ReadFile(dest) + if err != nil { + t.Fatalf("re-read backup: %v", err) + } + if len(before) != len(after) { + t.Errorf("existing backup was modified: %d bytes → %d", len(before), len(after)) + } +} + +func TestBackupMissingSource(t *testing.T) { + dir := t.TempDir() + if err := Backup(filepath.Join(dir, "nope.db"), filepath.Join(dir, "out.db")); err == nil { + t.Fatal("backup of a nonexistent database succeeded; want an error") + } +} diff --git a/internal/db/db.go b/internal/db/db.go index d52dbee..8e91c31 100644 --- a/internal/db/db.go +++ b/internal/db/db.go @@ -365,6 +365,138 @@ SELECT id, doc_id, from_pos, to_pos, original, replacement, explanation, type, s DROP TABLE suggestions; ALTER TABLE suggestions_new RENAME TO suggestions; CREATE INDEX idx_suggestions_doc_id ON suggestions(doc_id); +`, + }, + { + // Writing passport: evidence that a document was written, not pasted. + // + // `preserve_history` opts a document out of auto-snapshot pruning. The + // 40-snapshot cap is right for recovery (you want recent states) but + // wrong for provenance (you want the *whole* span, oldest included), so + // a writer who may need to defend authorship flags the doc and keeps + // every snapshot. + // + // `content_hash`/`prev_hash` chain each snapshot to the one before it: + // hash = sha256(prev_hash | doc_id | created_at | word_count | text). + // This proves the local history is internally consistent — no snapshot + // was edited, reordered, or removed after the fact without breaking + // every link downstream. It is NOT third-party attestation: anyone with + // the DB and the algorithm could forge a fresh chain. It raises the cost + // of a doctored history from "edit one row" to "rebuild all of them". + // Pre-existing snapshots keep empty hashes and are reported as + // unverifiable rather than as failures. + name: "0009_writing_passport", + stmt: ` +ALTER TABLE documents ADD COLUMN preserve_history INTEGER NOT NULL DEFAULT 0; +ALTER TABLE document_versions ADD COLUMN content_hash TEXT NOT NULL DEFAULT ''; +ALTER TABLE document_versions ADD COLUMN prev_hash TEXT NOT NULL DEFAULT ''; +`, + }, + { + // Real accounts. Three separate things land together because they are + // one change: Petal can now tell users apart. + // + // `sessions` backs server-side login state. The cookie carries an opaque + // random token and this table stores only its SHA-256 — a leaked database + // copy therefore yields no usable session, the same reason passwords are + // hashed. Server-side rows (rather than a signed stateless cookie) are + // what make logout and revocation actually revoke. + // + // `images` gives the content-addressed image store an owner. Until now it + // was a flat directory with no database row at all: any caller holding a + // hash could fetch anyone's image, which is capability-URL security, not + // access control. The primary key is (name, user_id), so the same picture + // uploaded by two people is still stored once on disk and simply has two + // rows — deduplication survives; the file is deleted only with its last + // row. Rows for images already on disk are backfilled at startup by the + // images package, which is the only code that knows the storage path. + // + // `users.pair_lang` is the writer's language pair (English + X). It is + // unused until the langpack work, but it belongs to provisioning and + // costs nothing to add while the users table is already being touched. + name: "0010_sessions_images_and_pair_lang", + stmt: ` +CREATE TABLE sessions ( + id TEXT PRIMARY KEY, + user_id TEXT NOT NULL REFERENCES users(id) ON DELETE CASCADE, + expires_at DATETIME NOT NULL, + created_at DATETIME DEFAULT CURRENT_TIMESTAMP, + user_agent TEXT NOT NULL DEFAULT '' +); +CREATE INDEX idx_sessions_user_id ON sessions(user_id); + +CREATE TABLE images ( + name TEXT NOT NULL, + user_id TEXT NOT NULL REFERENCES users(id) ON DELETE CASCADE, + content_type TEXT NOT NULL DEFAULT '', + size INTEGER NOT NULL DEFAULT 0, + created_at DATETIME DEFAULT CURRENT_TIMESTAMP, + PRIMARY KEY (name, user_id) +); +CREATE INDEX idx_images_user_id ON images(user_id); + +ALTER TABLE users ADD COLUMN pair_lang TEXT NOT NULL DEFAULT 'zh'; +`, + }, + { + // The personal spelling dictionary moves off the browser. It used to be a + // single `petal.spell.personal` key in localStorage, which meant two + // people sharing a device shared a word list built from one person's + // private writing — and one person writing on two devices had two + // unrelated lists. + // + // `lang` is the *dictionary's* language, not the writer's: a word is only + // ever added while a particular Hunspell dictionary flagged it, and an + // en-US personal word must not silence a pt-PT flag (or vice versa) once + // the second pair ships. `word` is stored as typed; matching is exact, + // because case carries meaning to a speller ("polish" vs "Polish"). + name: "0011_personal_dictionary", + stmt: ` +CREATE TABLE personal_words ( + user_id TEXT NOT NULL REFERENCES users(id) ON DELETE CASCADE, + lang TEXT NOT NULL, + word TEXT NOT NULL, + created_at DATETIME DEFAULT CURRENT_TIMESTAMP, + PRIMARY KEY (user_id, lang, word) +); +`, + }, + { + // The growth journal reads the suggestions table as a record of what the + // writer has been learning, and that reading only works if a row is dated + // by *her decision* rather than by the model's proposal. `created_at` is + // when a checkpoint offered the edit; a suggestion offered in April and + // accepted in June is June's growth, not April's. + // + // Existing rows are backfilled to created_at — which is exactly the + // approximation the journal would have had to make anyway, and is very + // nearly right in practice since edits are settled minutes after a + // checkpoint. Only pending rows keep a NULL: nothing has been decided. + name: "0012_suggestion_resolved_at", + stmt: ` +ALTER TABLE suggestions ADD COLUMN resolved_at DATETIME; +UPDATE suggestions SET resolved_at = created_at WHERE status != 'pending'; +CREATE INDEX idx_suggestions_resolved ON suggestions(status, resolved_at); +`, + }, + { + // Which engine proposed a row. Until now `type` doubled as that answer — + // 'mechanics' meant "the offline rule pack found this" and everything else + // meant "the model did". That breaks the moment an offline rule proposes a + // *collocation*: the miscollocation list (SUGGESTIONS §6) is the same + // family, the same rail and the same warm phrasing as the LLM coach, and it + // must stay type='collocation' so an accepted chunk still plants in the + // garden and still counts in the journal. With type no longer naming the + // engine, the two passes could not scope their own DELETEs — the coach + // would wipe the offline flags, and the offline pass would leave the + // coach's behind to accumulate. + // + // Existing mechanics rows are local by definition; everything else came + // from a model. + name: "0013_suggestion_source", + stmt: ` +ALTER TABLE suggestions ADD COLUMN source TEXT NOT NULL DEFAULT 'llm'; +UPDATE suggestions SET source = 'local' WHERE type = 'mechanics'; `, }, } diff --git a/internal/db/db_test.go b/internal/db/db_test.go index 1eb12b8..e705c0e 100644 --- a/internal/db/db_test.go +++ b/internal/db/db_test.go @@ -83,3 +83,147 @@ func TestOpenMigratesAndSeeds(t *testing.T) { t.Errorf("expected exactly 1 local user after reopen, got %d", users) } } + +// TestResolvedAtBackfill runs migration 0012 against a database that predates +// it, which is the only shape that matters: on the live box the suggestions +// table is years of settled edits with no resolved_at to their name. Backfilling +// to created_at is exactly the approximation the growth journal would otherwise +// have had to make, and a pending row must stay NULL — nothing has been decided. +func TestResolvedAtBackfill(t *testing.T) { + path := filepath.Join(t.TempDir(), "old.db") + d, err := Open(path) + if err != nil { + t.Fatalf("open: %v", err) + } + + // Rewind to the state before 0012: drop the column and forget the migration. + if _, err := d.Exec(`DROP INDEX idx_suggestions_resolved`); err != nil { + t.Fatalf("rewind index: %v", err) + } + if _, err := d.Exec(`ALTER TABLE suggestions DROP COLUMN resolved_at`); err != nil { + t.Fatalf("rewind schema: %v", err) + } + if _, err := d.Exec(`DELETE FROM schema_migrations WHERE name = '0012_suggestion_resolved_at'`); err != nil { + t.Fatalf("rewind migration record: %v", err) + } + if _, err := d.Exec(`INSERT INTO documents (id, user_id) VALUES ('d1', ?)`, LocalUserID); err != nil { + t.Fatalf("insert document: %v", err) + } + for _, s := range []struct{ id, status string }{ + {"s-old", "accepted"}, + {"s-open", "pending"}, + } { + if _, err := d.Exec( + `INSERT INTO suggestions (id, doc_id, from_pos, to_pos, original, replacement, explanation, type, status, created_at) + VALUES (?, 'd1', 0, 3, 'teh', 'the', 'x', 'grammar', ?, '2026-01-02 03:04:05')`, + s.id, s.status, + ); err != nil { + t.Fatalf("seed %s: %v", s.id, err) + } + } + d.Close() + + d2, err := Open(path) + if err != nil { + t.Fatalf("reopen (migrate): %v", err) + } + defer d2.Close() + + // Compared against created_at read back the same way: the driver renders a + // DATETIME column itself, so the assertion is "the same instant", not a + // particular text format. + var settled, created *string + if err := d2.QueryRow( + `SELECT resolved_at, created_at FROM suggestions WHERE id = 's-old'`, + ).Scan(&settled, &created); err != nil { + t.Fatalf("read settled row: %v", err) + } + if settled == nil || created == nil || *settled != *created { + t.Errorf("resolved_at = %v, want it backfilled from created_at (%v)", settled, created) + } + + var pending *string + if err := d2.QueryRow(`SELECT resolved_at FROM suggestions WHERE id = 's-open'`).Scan(&pending); err != nil { + t.Fatalf("read pending row: %v", err) + } + if pending != nil { + t.Errorf("pending row got resolved_at = %v, want NULL — nothing was decided", *pending) + } +} + +// TestSuggestionSourceBackfill runs migration 0013 against a database that +// predates it — the shape the live box is actually in. `source` is the column +// that lets the offline rule pack and the model share the collocation family +// without deleting each other's rows, and it can only do that if the existing +// rows are labelled correctly on the way in: everything the old deterministic +// pass wrote is local, and everything else came from a model. +func TestSuggestionSourceBackfill(t *testing.T) { + path := filepath.Join(t.TempDir(), "old.db") + d, err := Open(path) + if err != nil { + t.Fatalf("open: %v", err) + } + + // Rewind to the state before 0013. + if _, err := d.Exec(`ALTER TABLE suggestions DROP COLUMN source`); err != nil { + t.Fatalf("rewind schema: %v", err) + } + if _, err := d.Exec(`DELETE FROM schema_migrations WHERE name = '0013_suggestion_source'`); err != nil { + t.Fatalf("rewind migration record: %v", err) + } + if _, err := d.Exec(`INSERT INTO documents (id, user_id) VALUES ('d1', ?)`, LocalUserID); err != nil { + t.Fatalf("insert document: %v", err) + } + for _, s := range []struct{ id, typ string }{ + {"s-mech", SuggestionTypeMechanics}, + {"s-gram", SuggestionTypeGrammar}, + {"s-coll", SuggestionTypeCollocation}, + } { + if _, err := d.Exec( + `INSERT INTO suggestions (id, doc_id, from_pos, to_pos, original, replacement, explanation, type) + VALUES (?, 'd1', 0, 3, 'teh', 'the', 'x', ?)`, + s.id, s.typ, + ); err != nil { + t.Fatalf("seed %s: %v", s.id, err) + } + } + d.Close() + + d2, err := Open(path) + if err != nil { + t.Fatalf("reopen (migrate): %v", err) + } + defer d2.Close() + + // A pre-0013 collocation row can only have come from the coach — the offline + // miscollocation list did not exist yet — so it must NOT be claimed as local. + for id, want := range map[string]string{ + "s-mech": SuggestionSourceLocal, + "s-gram": SuggestionSourceLLM, + "s-coll": SuggestionSourceLLM, + } { + var got string + if err := d2.QueryRow(`SELECT source FROM suggestions WHERE id = ?`, id).Scan(&got); err != nil { + t.Fatalf("read %s: %v", id, err) + } + if got != want { + t.Errorf("%s: source = %q, want %q", id, got, want) + } + } + + // And a row written after the migration defaults to the model, so a code path + // that forgets to name a source can never silently claim to be offline. + if _, err := d2.Exec( + `INSERT INTO suggestions (id, doc_id, from_pos, to_pos, original, replacement, explanation, type) + VALUES ('s-new', 'd1', 0, 3, 'teh', 'the', 'x', 'grammar')`, + ); err != nil { + t.Fatalf("insert new row: %v", err) + } + var fresh string + if err := d2.QueryRow(`SELECT source FROM suggestions WHERE id = 's-new'`).Scan(&fresh); err != nil { + t.Fatalf("read new row: %v", err) + } + if fresh != SuggestionSourceLLM { + t.Errorf("default source = %q, want %q", fresh, SuggestionSourceLLM) + } +} diff --git a/internal/db/models.go b/internal/db/models.go index 635d7cb..1c9e073 100644 --- a/internal/db/models.go +++ b/internal/db/models.go @@ -2,14 +2,19 @@ package db import "time" -// User is an account. With auth deferred, the app runs as a single hardcoded -// `local` user (see LocalUserID); the user_id columns and this type exist so -// real auth can drop in later without a schema migration. +// User is an account. Its ID is the OIDC subject for anyone who signed in, or +// LocalUserID for the pre-auth single user (and for local development, where +// StaticResolver still hands out that id). type User struct { ID string `json:"id"` Email string `json:"email"` DisplayName string `json:"display_name"` CreatedAt time.Time `json:"created_at"` + + // PairLang is the X in this writer's (English + X) language pair — "zh" + // today, "pt-PT"/"fr"/"es" once the langpacks land. It selects the UI copy + // and dictionary set, not the language they may type in. + PairLang string `json:"pair_lang"` } // Document is a single piece of writing. `Content` is the Tiptap JSON document @@ -25,6 +30,10 @@ type Document struct { WordCount int `json:"word_count"` CreatedAt time.Time `json:"created_at"` UpdatedAt time.Time `json:"updated_at"` + + // PreserveHistory opts this document out of auto-snapshot pruning so its + // full writing trail survives as authorship evidence (see the passport). + PreserveHistory bool `json:"preserve_history"` } // DocumentVersion is a point-in-time snapshot of a document's body, captured so @@ -42,6 +51,13 @@ type DocumentVersion struct { WordCount int `json:"word_count"` Kind string `json:"kind"` // auto | manual | pre_restore CreatedAt time.Time `json:"created_at"` + + // ContentHash chains this snapshot to the previous one (PrevHash), so a + // history that was edited or thinned after the fact fails verification. + // Both are empty for snapshots taken before the chain existed. Omitted from + // list responses; the passport loads them explicitly. + ContentHash string `json:"content_hash,omitempty"` + PrevHash string `json:"prev_hash,omitempty"` } // Document version kinds, mirrored from the schema CHECK constraint. @@ -90,7 +106,11 @@ type Suggestion struct { Explanation string `json:"explanation"` Type string `json:"type"` // grammar | phrasing | idiom | clarity | voice | collocation Status string `json:"status"` // pending | accepted | rejected - CreatedAt time.Time `json:"created_at"` + // Source names the engine that proposed the edit, not its family: an offline + // rule and the model can both propose a collocation, and the writer is never + // told which one spoke. It exists so each pass can replace its own rows. + Source string `json:"source"` // llm | local + CreatedAt time.Time `json:"created_at"` } // Suggestion type and status values, mirrored from the schema CHECK constraints. @@ -103,6 +123,12 @@ const ( SuggestionTypeCollocation = "collocation" SuggestionTypeMechanics = "mechanics" // deterministic rule-based pass (no LLM) + // Who proposed it. The offline rule pack ('local') runs on every edit inside + // the browser and survives a VPN-down box; the model ('llm') adds the long + // tail when it is reachable. + SuggestionSourceLLM = "llm" + SuggestionSourceLocal = "local" + SuggestionStatusPending = "pending" SuggestionStatusAccepted = "accepted" SuggestionStatusRejected = "rejected" diff --git a/internal/docs/export.go b/internal/docs/export.go index 8eafdd9..f2c49e7 100644 --- a/internal/docs/export.go +++ b/internal/docs/export.go @@ -13,6 +13,7 @@ import ( "github.com/go-chi/chi/v5" + "gitea.parodia.dev/drwily/petal/internal/auth" "gitea.parodia.dev/drwily/petal/internal/db" "gitea.parodia.dev/drwily/petal/internal/httputil" ) @@ -46,7 +47,7 @@ func (h *Handler) exportAll(w http.ResponseWriter, r *http.Request) { FROM documents WHERE user_id = ? ORDER BY updated_at DESC`, - db.LocalUserID, + auth.UserID(r.Context()), ) if err != nil { httputil.ServerError(w, err) @@ -143,7 +144,7 @@ func (h *Handler) export(w http.ResponseWriter, r *http.Request) { return } - doc, err := h.fetch(chi.URLParam(r, "id")) + doc, err := h.fetch(auth.UserID(r.Context()), chi.URLParam(r, "id")) if errors.Is(err, sql.ErrNoRows) { notFound(w) return diff --git a/internal/docs/handlers.go b/internal/docs/handlers.go index a946460..42f5a28 100644 --- a/internal/docs/handlers.go +++ b/internal/docs/handlers.go @@ -1,6 +1,7 @@ // Package docs implements the document CRUD HTTP handlers — the create / list / // read / update / delete surface that backs the editor and its 1.5s auto-save. -// All access is scoped to the single hardcoded local user while auth is deferred. +// Every query is scoped to the caller resolved by the auth middleware, so a +// document is only ever reachable by the user who owns it. package docs import ( @@ -12,6 +13,7 @@ import ( "github.com/go-chi/chi/v5" + "gitea.parodia.dev/drwily/petal/internal/auth" "gitea.parodia.dev/drwily/petal/internal/db" "gitea.parodia.dev/drwily/petal/internal/httputil" ) @@ -52,15 +54,16 @@ type docSummary struct { Tags []db.Tag `json:"tags"` } -// list returns the local user's documents, most-recently-updated first, each +// list returns the caller's documents, most-recently-updated first, each // decorated with its tags. func (h *Handler) list(w http.ResponseWriter, r *http.Request) { + userID := auth.UserID(r.Context()) rows, err := h.DB.Query( `SELECT id, title, word_count, updated_at FROM documents WHERE user_id = ? ORDER BY updated_at DESC`, - db.LocalUserID, + userID, ) if err != nil { httputil.ServerError(w, err) @@ -84,7 +87,7 @@ func (h *Handler) list(w http.ResponseWriter, r *http.Request) { return } - byDoc, err := h.tagsByDoc(ids) + byDoc, err := h.tagsByDoc(userID, ids) if err != nil { httputil.ServerError(w, err) return @@ -104,7 +107,7 @@ func (h *Handler) create(w http.ResponseWriter, r *http.Request) { err := h.DB.QueryRow( `INSERT INTO documents (user_id) VALUES (?) RETURNING id, user_id, title, content, content_text, tone, word_count, created_at, updated_at`, - db.LocalUserID, + auth.UserID(r.Context()), ).Scan( &doc.ID, &doc.UserID, &doc.Title, &doc.Content, &doc.ContentText, &doc.Tone, &doc.WordCount, &doc.CreatedAt, &doc.UpdatedAt, @@ -118,7 +121,7 @@ func (h *Handler) create(w http.ResponseWriter, r *http.Request) { // get returns a single full document by id. func (h *Handler) get(w http.ResponseWriter, r *http.Request) { - doc, err := h.fetch(chi.URLParam(r, "id")) + doc, err := h.fetch(auth.UserID(r.Context()), chi.URLParam(r, "id")) if errors.Is(err, sql.ErrNoRows) { notFound(w) return @@ -139,12 +142,17 @@ type updateRequest struct { ContentText *string `json:"content_text"` Tone *string `json:"tone"` WordCount *int `json:"word_count"` + + // PreserveHistory toggles the passport's keep-everything mode. Sent alone + // by the History panel's toggle, never by the auto-save path. + PreserveHistory *bool `json:"preserve_history"` } // update applies the provided fields to a document and returns the saved row. // content and content_text are kept in sync by the client and written together. func (h *Handler) update(w http.ResponseWriter, r *http.Request) { id := chi.URLParam(r, "id") + userID := auth.UserID(r.Context()) var req updateRequest if err := json.NewDecoder(r.Body).Decode(&req); err != nil { @@ -159,9 +167,11 @@ func (h *Handler) update(w http.ResponseWriter, r *http.Request) { content_text = COALESCE(?, content_text), tone = COALESCE(?, tone), word_count = COALESCE(?, word_count), + preserve_history = COALESCE(?, preserve_history), updated_at = CURRENT_TIMESTAMP WHERE id = ? AND user_id = ?`, - req.Title, req.Content, req.ContentText, req.Tone, req.WordCount, id, db.LocalUserID, + req.Title, req.Content, req.ContentText, req.Tone, req.WordCount, + req.PreserveHistory, id, userID, ) if err != nil { httputil.ServerError(w, err) @@ -172,7 +182,7 @@ func (h *Handler) update(w http.ResponseWriter, r *http.Request) { return } - doc, err := h.fetch(id) + doc, err := h.fetch(userID, id) if err != nil { httputil.ServerError(w, err) return @@ -194,7 +204,7 @@ func (h *Handler) update(w http.ResponseWriter, r *http.Request) { func (h *Handler) delete(w http.ResponseWriter, r *http.Request) { res, err := h.DB.Exec( `DELETE FROM documents WHERE id = ? AND user_id = ?`, - chi.URLParam(r, "id"), db.LocalUserID, + chi.URLParam(r, "id"), auth.UserID(r.Context()), ) if err != nil { httputil.ServerError(w, err) @@ -207,17 +217,21 @@ func (h *Handler) delete(w http.ResponseWriter, r *http.Request) { w.WriteHeader(http.StatusNoContent) } -// fetch loads one full document scoped to the local user. -func (h *Handler) fetch(id string) (db.Document, error) { +// fetch loads one full document, scoped to its owner. Callers pass the id from +// [auth.UserID]; a document belonging to anyone else comes back as +// sql.ErrNoRows, which handlers surface as a 404 rather than a 403 (a stranger's +// document should be indistinguishable from one that doesn't exist). +func (h *Handler) fetch(userID, id string) (db.Document, error) { var doc db.Document err := h.DB.QueryRow( - `SELECT id, user_id, title, content, content_text, tone, word_count, created_at, updated_at + `SELECT id, user_id, title, content, content_text, tone, word_count, + created_at, updated_at, preserve_history FROM documents WHERE id = ? AND user_id = ?`, - id, db.LocalUserID, + id, userID, ).Scan( &doc.ID, &doc.UserID, &doc.Title, &doc.Content, &doc.ContentText, - &doc.Tone, &doc.WordCount, &doc.CreatedAt, &doc.UpdatedAt, + &doc.Tone, &doc.WordCount, &doc.CreatedAt, &doc.UpdatedAt, &doc.PreserveHistory, ) return doc, err } diff --git a/internal/docs/handlers_test.go b/internal/docs/handlers_test.go index 1d0e35e..d80c9e7 100644 --- a/internal/docs/handlers_test.go +++ b/internal/docs/handlers_test.go @@ -8,10 +8,14 @@ import ( "path/filepath" "testing" + "gitea.parodia.dev/drwily/petal/internal/auth" "gitea.parodia.dev/drwily/petal/internal/db" ) -// newTestServer spins up an isolated on-disk database and the docs router. +// newTestServer spins up an isolated on-disk database and the docs router, +// behind the same auth middleware main.go installs. Tests must go through it: +// handlers read the caller from the request context, so a router mounted bare +// would see an empty user id and match no rows. func newTestServer(t *testing.T) http.Handler { t.Helper() database, err := db.Open(filepath.Join(t.TempDir(), "test.db")) @@ -19,7 +23,13 @@ func newTestServer(t *testing.T) http.Handler { t.Fatalf("open db: %v", err) } t.Cleanup(func() { database.Close() }) - return New(database).Routes() + return withAuth(New(database).Routes()) +} + +// withAuth wraps a router so every test request arrives authenticated as the +// seeded local user — the stand-in for a real session until Authentik lands. +func withAuth(h http.Handler) http.Handler { + return auth.Middleware(auth.StaticResolver(db.LocalUserID))(h) } func do(t *testing.T, srv http.Handler, method, path, body string) *httptest.ResponseRecorder { diff --git a/internal/docs/isolation_test.go b/internal/docs/isolation_test.go new file mode 100644 index 0000000..8622cda --- /dev/null +++ b/internal/docs/isolation_test.go @@ -0,0 +1,222 @@ +package docs + +import ( + "encoding/json" + "net/http" + "path/filepath" + "testing" + + "github.com/go-chi/chi/v5" + + "gitea.parodia.dev/drwily/petal/internal/auth" + "gitea.parodia.dev/drwily/petal/internal/db" +) + +// This file is the point of the auth plumbing: it proves that swapping the +// hardcoded user for a request-scoped one actually isolates accounts. Every +// handler resolves its user from the request, so mounting the same routers twice +// behind two different resolvers gives us two "logged-in" users over one +// database — which is exactly the situation a real login will create. + +// newTwoUserServer opens one database holding two users and returns a router for +// each, identical but for who the auth middleware says is calling. +func newTwoUserServer(t *testing.T) (alice, bob http.Handler) { + t.Helper() + database, err := db.Open(filepath.Join(t.TempDir(), "test.db")) + if err != nil { + t.Fatalf("open db: %v", err) + } + t.Cleanup(func() { database.Close() }) + + // db.Open seeds the local user; add a second so both sides have a valid FK. + if _, err := database.Exec( + `INSERT INTO users (id, email, display_name) VALUES (?, ?, ?)`, + "bob", "bob@petal.local", "Bob", + ); err != nil { + t.Fatalf("seed second user: %v", err) + } + + mount := func(userID string) http.Handler { + h := New(database) + r := chi.NewRouter() + r.Mount("/docs", h.Routes()) + r.Mount("/tags", h.TagRoutes()) + r.Mount("/search", h.SearchRoutes()) + return auth.Middleware(auth.StaticResolver(userID))(r) + } + return mount(db.LocalUserID), mount("bob") +} + +// TestDocumentIsolation walks every read and write path that takes a document id +// and asserts Bob cannot reach Alice's document through any of them. A stranger's +// document must be indistinguishable from a nonexistent one — 404, never 403. +func TestDocumentIsolation(t *testing.T) { + alice, bob := newTwoUserServer(t) + + docID := createDoc(t, alice, "Alice's diary", "a private sentence about my day") + + t.Run("not in list", func(t *testing.T) { + rec := do(t, bob, http.MethodGet, "/docs", "") + var out []docSummary + if err := json.Unmarshal(rec.Body.Bytes(), &out); err != nil { + t.Fatalf("decode list: %v", err) + } + if len(out) != 0 { + t.Fatalf("bob sees %d of alice's documents, want 0", len(out)) + } + }) + + t.Run("not in search", func(t *testing.T) { + rec := do(t, bob, http.MethodGet, "/search?q=private", "") + var out []searchResult + if err := json.Unmarshal(rec.Body.Bytes(), &out); err != nil { + t.Fatalf("decode search: %v", err) + } + if len(out) != 0 { + t.Fatalf("search leaked %d of alice's documents", len(out)) + } + }) + + // The FTS index is a separate table joined back to documents; a missing + // user_id filter there would leak content even though the list query is + // scoped, so assert the owner still finds her own document. + t.Run("owner still finds it", func(t *testing.T) { + rec := do(t, alice, http.MethodGet, "/search?q=private", "") + var out []searchResult + if err := json.Unmarshal(rec.Body.Bytes(), &out); err != nil { + t.Fatalf("decode search: %v", err) + } + if len(out) != 1 { + t.Fatalf("alice found %d results for her own document, want 1", len(out)) + } + }) + + for _, tc := range []struct { + name, method, path, body string + }{ + {"get", http.MethodGet, "/docs/" + docID, ""}, + {"update", http.MethodPut, "/docs/" + docID, `{"title":"defaced"}`}, + {"delete", http.MethodDelete, "/docs/" + docID, ""}, + {"export", http.MethodGet, "/docs/" + docID + "/export?format=md", ""}, + {"passport", http.MethodGet, "/docs/" + docID + "/passport", ""}, + {"snapshot", http.MethodPost, "/docs/" + docID + "/versions", ""}, + } { + t.Run(tc.name, func(t *testing.T) { + rec := do(t, bob, tc.method, tc.path, tc.body) + if rec.Code != http.StatusNotFound { + t.Fatalf("%s %s as bob = %d, want 404 (body: %s)", + tc.method, tc.path, rec.Code, rec.Body) + } + }) + } + + // The document must have survived every attempt above unchanged. + rec := do(t, alice, http.MethodGet, "/docs/"+docID, "") + if rec.Code != http.StatusOK { + t.Fatalf("alice lost access to her own document: %d %s", rec.Code, rec.Body) + } + var doc db.Document + if err := json.Unmarshal(rec.Body.Bytes(), &doc); err != nil { + t.Fatalf("decode doc: %v", err) + } + if doc.Title != "Alice's diary" { + t.Fatalf("title = %q, want %q — bob's update went through", doc.Title, "Alice's diary") + } +} + +// TestVersionIsolation covers the history endpoints, which scope through a join +// to documents rather than a direct user_id column — an easy place to forget the +// filter, and one where the leak would be the full text of every draft. +func TestVersionIsolation(t *testing.T) { + alice, bob := newTwoUserServer(t) + + docID := createDoc(t, alice, "Draft", "the first version of my essay") + rec := do(t, alice, http.MethodPost, "/docs/"+docID+"/versions", "") + if rec.Code != http.StatusCreated { + t.Fatalf("snapshot: %d %s", rec.Code, rec.Body) + } + var v db.DocumentVersion + if err := json.Unmarshal(rec.Body.Bytes(), &v); err != nil { + t.Fatalf("decode version: %v", err) + } + + t.Run("list is empty for stranger", func(t *testing.T) { + rec := do(t, bob, http.MethodGet, "/docs/"+docID+"/versions", "") + var out []db.DocumentVersion + if err := json.Unmarshal(rec.Body.Bytes(), &out); err != nil { + t.Fatalf("decode: %v", err) + } + if len(out) != 0 { + t.Fatalf("bob sees %d of alice's snapshots, want 0", len(out)) + } + }) + + for _, tc := range []struct{ name, method, path string }{ + {"preview", http.MethodGet, "/docs/" + docID + "/versions/" + v.ID}, + {"restore", http.MethodPost, "/docs/" + docID + "/versions/" + v.ID + "/restore"}, + } { + t.Run(tc.name, func(t *testing.T) { + rec := do(t, bob, tc.method, tc.path, "") + if rec.Code != http.StatusNotFound { + t.Fatalf("%s as bob = %d, want 404 (body: %s)", tc.name, rec.Code, rec.Body) + } + }) + } +} + +// TestTagIsolation checks the tag roster and, more importantly, that a document +// and a tag can't be cross-linked across accounts — the assignment endpoint takes +// two ids from different tables and must own-check both. +func TestTagIsolation(t *testing.T) { + alice, bob := newTwoUserServer(t) + + docID := createDoc(t, alice, "Essay", "some words") + + rec := do(t, alice, http.MethodPost, "/tags", `{"name":"school","color":"mint"}`) + if rec.Code != http.StatusCreated { + t.Fatalf("create tag: %d %s", rec.Code, rec.Body) + } + var aliceTag db.Tag + if err := json.Unmarshal(rec.Body.Bytes(), &aliceTag); err != nil { + t.Fatalf("decode tag: %v", err) + } + + rec = do(t, bob, http.MethodPost, "/tags", `{"name":"bobs","color":"sky"}`) + var bobTag db.Tag + if err := json.Unmarshal(rec.Body.Bytes(), &bobTag); err != nil { + t.Fatalf("decode bob tag: %v", err) + } + + t.Run("roster is per user", func(t *testing.T) { + rec := do(t, bob, http.MethodGet, "/tags", "") + var out []db.Tag + if err := json.Unmarshal(rec.Body.Bytes(), &out); err != nil { + t.Fatalf("decode: %v", err) + } + if len(out) != 1 || out[0].Name != "bobs" { + t.Fatalf("bob's roster = %+v, want just his own tag", out) + } + }) + + t.Run("cannot tag a stranger's document", func(t *testing.T) { + body, _ := json.Marshal(map[string]string{"tag_id": bobTag.ID}) + rec := do(t, bob, http.MethodPost, "/docs/"+docID+"/tags", string(body)) + if rec.Code != http.StatusNotFound { + t.Fatalf("bob tagging alice's doc = %d, want 404", rec.Code) + } + }) + + t.Run("cannot rename a stranger's tag", func(t *testing.T) { + rec := do(t, bob, http.MethodPatch, "/tags/"+aliceTag.ID, `{"name":"stolen"}`) + if rec.Code != http.StatusNotFound { + t.Fatalf("bob renaming alice's tag = %d, want 404", rec.Code) + } + }) + + t.Run("cannot delete a stranger's tag", func(t *testing.T) { + rec := do(t, bob, http.MethodDelete, "/tags/"+aliceTag.ID, "") + if rec.Code != http.StatusNotFound { + t.Fatalf("bob deleting alice's tag = %d, want 404", rec.Code) + } + }) +} diff --git a/internal/docs/passport.go b/internal/docs/passport.go new file mode 100644 index 0000000..d5fe298 --- /dev/null +++ b/internal/docs/passport.go @@ -0,0 +1,283 @@ +package docs + +// Writing passport: a standalone, printable report showing *how* a document was +// written — when each snapshot landed, how the word count grew, how the work +// broke into sessions. It exists because automated "AI detector" verdicts are +// unreliable and skew against non-native English writers, so the useful thing to +// hand someone who doubts your authorship is not a score but a record. +// +// The report is deliberately modest about what it proves (see passportLimits): +// it evidences a plausible writing process, it does not certify one. + +import ( + "crypto/sha256" + "database/sql" + "encoding/hex" + "errors" + "fmt" + "net/http" + "time" + + "github.com/go-chi/chi/v5" + + "gitea.parodia.dev/drwily/petal/internal/auth" + "gitea.parodia.dev/drwily/petal/internal/db" + "gitea.parodia.dev/drwily/petal/internal/httputil" +) + +// Passport tuning. +const ( + // sessionGap is the idle time that separates one writing session from the + // next. Auto-snapshots fire at most every 3 minutes while typing, so any + // gap far above that means the writer stepped away. 45 minutes keeps a + // coffee break inside one session but splits morning from evening work. + sessionGap = 45 * time.Minute + + // jumpNoteThreshold is the share of the final word count a single + // snapshot-to-snapshot increase must exceed before the report calls it out. + // A large jump is the first thing a skeptical reader will ask about, so the + // report raises it rather than leaving it to be discovered. + jumpNoteThreshold = 0.25 +) + +// chainHash links a snapshot to its predecessor. Covering prev_hash makes each +// hash depend on the entire history before it, so altering any earlier snapshot +// invalidates every later one; covering created_at means a row cannot be +// silently backdated. +// +// This detects tampering with the local database. It is not third-party +// attestation — someone with the database and this function could regenerate a +// consistent chain from scratch. +func chainHash(prevHash, docID string, createdAt time.Time, wordCount int, text string) string { + h := sha256.New() + fmt.Fprintf(h, "%s\x00%s\x00%d\x00%d\x00%s", + prevHash, docID, createdAt.UTC().UnixNano(), wordCount, text) + return hex.EncodeToString(h.Sum(nil)) +} + +// --- report model ----------------------------------------------------------- + +// passportSession is one continuous stretch of work — snapshots with no +// sessionGap-sized pause between them. +type passportSession struct { + Start, End time.Time + Snapshots int + WordsAdded int // net change across the session; negative when trimming +} + +// Duration is the observed length of the session: first snapshot to last. A +// single-snapshot session reports zero, which is why total active time is +// described as a lower bound. +func (s passportSession) Duration() time.Duration { return s.End.Sub(s.Start) } + +// chain verification outcomes, in the order the report prefers to report them. +const ( + chainVerified = "verified" // every hash recomputes and every link holds + chainGaps = "gaps" // hashes valid, links broken — consistent with pruning + chainPartial = "partial" // some snapshots predate the hash chain + chainUnverifiable = "unverifiable" // no snapshot carries a hash + chainBroken = "broken" // a hash does not match its own contents +) + +// passportData is everything the template renders. +type passportData struct { + Doc db.Document + Versions []db.DocumentVersion // ascending by time + + Sessions []passportSession + FirstAt time.Time + LastAt time.Time + Span time.Duration // wall-clock first snapshot → last + ActiveTime time.Duration // summed session durations; a lower bound + + LargestJump int // biggest single snapshot-to-snapshot word increase + LargestJumpAt time.Time + LargestJumpIdx int // index into Versions, so the chart can mark it + NoteJump bool // jump is large enough to be worth pre-empting + + ChainStatus string + UnhashedCount int + GeneratedAt time.Time +} + +// buildPassport derives the report from a document and its snapshots, which must +// be ordered oldest-first. It assumes nothing about snapshot spacing. +func buildPassport(doc db.Document, versions []db.DocumentVersion) passportData { + d := passportData{ + Doc: doc, + Versions: versions, + GeneratedAt: time.Now(), + } + if len(versions) == 0 { + d.ChainStatus = chainUnverifiable + return d + } + + d.FirstAt = versions[0].CreatedAt + d.LastAt = versions[len(versions)-1].CreatedAt + d.Span = d.LastAt.Sub(d.FirstAt) + + cur := passportSession{Start: versions[0].CreatedAt, End: versions[0].CreatedAt, Snapshots: 1} + + // Baseline for the running session's net-words figure. Later sessions + // measure from the *previous* session's final count, not from their own + // first snapshot, because that first snapshot already contains the few + // minutes of typing that preceded it — measuring from it would drop that + // work. The first session is the exception: it measures from its own first + // snapshot rather than from zero, so a history whose early snapshots were + // pruned understates session one instead of reporting the words it never + // saw as a sudden addition. + startWords := versions[0].WordCount + + for i := 1; i < len(versions); i++ { + v, prev := versions[i], versions[i-1] + + if delta := v.WordCount - prev.WordCount; delta > d.LargestJump { + d.LargestJump, d.LargestJumpAt, d.LargestJumpIdx = delta, v.CreatedAt, i + } + + if v.CreatedAt.Sub(prev.CreatedAt) > sessionGap { + cur.WordsAdded = prev.WordCount - startWords + d.Sessions = append(d.Sessions, cur) + cur = passportSession{Start: v.CreatedAt, End: v.CreatedAt, Snapshots: 1} + startWords = prev.WordCount + continue + } + cur.End = v.CreatedAt + cur.Snapshots++ + } + cur.WordsAdded = versions[len(versions)-1].WordCount - startWords + d.Sessions = append(d.Sessions, cur) + + for _, s := range d.Sessions { + d.ActiveTime += s.Duration() + } + + final := versions[len(versions)-1].WordCount + d.NoteJump = final > 0 && float64(d.LargestJump)/float64(final) > jumpNoteThreshold + + d.ChainStatus, d.UnhashedCount = verifyChain(doc, versions) + return d +} + +// verifyChain recomputes every snapshot's hash and checks that each links to the +// one before it. Returns the outcome and how many snapshots predate the chain. +// +// Broken *links* are not evidence of tampering on their own: auto-snapshot +// pruning legitimately removes rows from the middle of the history, which severs +// the links across the hole. So a link break is reported as a gap unless the +// document is in preserve-history mode, where nothing should ever be removed. A +// hash that fails to match its *own* contents is unambiguous, and always broken. +func verifyChain(doc db.Document, versions []db.DocumentVersion) (status string, unhashed int) { + var ( + hashed int + linkBreak bool + prevHash string + havePrev bool + ) + + for _, v := range versions { + if v.ContentHash == "" { + unhashed++ + havePrev = false // can't vouch for what follows an unhashed row + continue + } + hashed++ + + want := chainHash(v.PrevHash, v.DocID, v.CreatedAt, v.WordCount, v.ContentText) + if want != v.ContentHash { + return chainBroken, unhashed + } + if havePrev && v.PrevHash != prevHash { + linkBreak = true + } + prevHash, havePrev = v.ContentHash, true + } + + switch { + case hashed == 0: + return chainUnverifiable, unhashed + case linkBreak && doc.PreserveHistory: + // Nothing should have been removed from a preserved history. + return chainBroken, unhashed + case linkBreak: + return chainGaps, unhashed + case unhashed > 0: + return chainPartial, unhashed + default: + return chainVerified, unhashed + } +} + +// --- HTTP ------------------------------------------------------------------- + +// passport renders the report for one document as a standalone HTML download. +// HTML rather than PDF for the same reason as the other exports: a CJK-safe PDF +// needs an embedded Unicode font or a headless browser. The page is styled for +// printing, so "Save as PDF" in the browser produces the handoff artifact. +func (h *Handler) passport(w http.ResponseWriter, r *http.Request) { + docID := chi.URLParam(r, "id") + userID := auth.UserID(r.Context()) + + doc, err := h.fetch(userID, docID) + if errors.Is(err, sql.ErrNoRows) { + notFound(w) + return + } + if err != nil { + httputil.ServerError(w, err) + return + } + + versions, err := h.passportVersions(userID, docID) + if err != nil { + httputil.ServerError(w, err) + return + } + + body := renderPassport(buildPassport(doc, versions)) + + filename := sanitizeFilename(doc.Title) + if filename == "" { + filename = "untitled" + } + filename += " - writing passport.html" + + w.Header().Set("Content-Type", "text/html; charset=utf-8") + w.Header().Set("Content-Disposition", + fmt.Sprintf("attachment; filename*=UTF-8''%s", urlEscapeFilename(filename))) + w.Header().Set("Content-Length", fmt.Sprintf("%d", len(body))) + _, _ = w.Write(body) +} + +// passportVersions loads every snapshot oldest-first with the fields the report +// and the chain check need — including content_text, which the list endpoint +// omits as too heavy but verification cannot do without. +func (h *Handler) passportVersions(userID, docID string) ([]db.DocumentVersion, error) { + rows, err := h.DB.Query( + `SELECT v.id, v.doc_id, v.title, v.content_text, v.word_count, v.kind, + v.created_at, v.content_hash, v.prev_hash + FROM document_versions v + JOIN documents d ON d.id = v.doc_id + WHERE v.doc_id = ? AND d.user_id = ? + ORDER BY v.created_at ASC, v.rowid ASC`, + docID, userID, + ) + if err != nil { + return nil, err + } + defer rows.Close() + + var out []db.DocumentVersion + for rows.Next() { + var v db.DocumentVersion + if err := rows.Scan( + &v.ID, &v.DocID, &v.Title, &v.ContentText, &v.WordCount, &v.Kind, + &v.CreatedAt, &v.ContentHash, &v.PrevHash, + ); err != nil { + return nil, err + } + out = append(out, v) + } + return out, rows.Err() +} diff --git a/internal/docs/passport_render.go b/internal/docs/passport_render.go new file mode 100644 index 0000000..501d0b6 --- /dev/null +++ b/internal/docs/passport_render.go @@ -0,0 +1,387 @@ +package docs + +// HTML rendering for the writing passport. Self-contained (no external assets) +// and styled for print, so the browser's "Save as PDF" turns it into the file a +// writer actually hands over. + +import ( + "fmt" + "math" + "strings" + "time" +) + +// Chart geometry. The plot is wide and short on purpose: the report's question +// is "what shape did this document grow in", and a wide aspect makes a steady +// climb read as steady rather than dramatic. +const ( + chartW, chartH = 760, 260 + padL, padR, padT, padB = 52, 20, 18, 34 + plotW, plotH = chartW - padL - padR, chartH - padT - padB + minBandW = 2.0 // so a single-snapshot session still shows + + // gutterSlots is the space between sessions, in snapshot-slot widths. Wide + // enough to read as a break and to seat its duration label. + gutterSlots = 2.5 +) + +// Palette — the export stylesheet's tokens, reused so a passport looks like it +// came from the same application as the document it describes. +const ( + rose = "#b04a6a" + roseLight = "#f6d6e0" + roseWash = "#fdeef3" + surface = "#fffafb" +) + +func renderPassport(d passportData) []byte { + var b strings.Builder + + fmt.Fprintf(&b, passportHead, htmlEscape(d.Doc.Title)) + + fmt.Fprintf(&b, `
+

Writing passport

+

%s

+

Generated %s

+
+`, htmlEscape(d.Doc.Title), htmlEscape(formatWhen(d.GeneratedAt))) + + if len(d.Versions) == 0 { + b.WriteString(`

This document has no saved history yet, so there is +nothing to report. History builds up automatically as you write.

+`) + return []byte(b.String()) + } + + b.WriteString(renderStats(d)) + b.WriteString(renderChart(d)) + b.WriteString(renderSessions(d)) + b.WriteString(renderIntegrity(d)) + b.WriteString(passportLimits) + b.WriteString("\n") + + return []byte(b.String()) +} + +// renderStats is the headline row — the numbers a reader wants before deciding +// whether to study the chart. +func renderStats(d passportData) string { + final := d.Versions[len(d.Versions)-1].WordCount + + tiles := []struct{ value, label string }{ + {fmt.Sprintf("%d", len(d.Versions)), "snapshots saved"}, + {humanDuration(d.Span), "from first to last edit"}, + {fmt.Sprintf("%d", len(d.Sessions)), pluralize(len(d.Sessions), "writing session", "writing sessions")}, + {humanDuration(d.ActiveTime), "spent actively editing"}, + {fmt.Sprintf("%d", final), "words in the final draft"}, + } + + var b strings.Builder + b.WriteString(`
`) + for _, t := range tiles { + fmt.Fprintf(&b, `
%s%s
`, + htmlEscape(t.value), htmlEscape(t.label)) + } + b.WriteString("
\n") + return b.String() +} + +// renderChart draws word count as a step line — the count is only known at +// snapshot moments, and a step says that honestly where a smooth curve would +// invent values in between. Shaded bands mark writing sessions. +// +// The x axis is snapshot order, not wall-clock time, and this is deliberate. On +// a linear time axis an essay written in three half-hour sittings across three +// days renders as three vertical cliffs separated by empty space: all the actual +// writing is crushed into one percent of the width, and the result looks exactly +// like text pasted in three chunks — the opposite of what happened. Because +// auto-snapshots are throttled to roughly one per few minutes of *active* +// editing, snapshot order is already close to proportional to time spent +// writing. So the plot gives its width to the writing and compresses the breaks, +// which are drawn as explicit labelled gaps rather than silently removed. +// +// One series, so no legend: the heading names it. +func renderChart(d passportData) string { + maxW := 0 + for _, v := range d.Versions { + if v.WordCount > maxW { + maxW = v.WordCount + } + } + yTop := niceCeil(maxW) + + y := func(words int) float64 { + if yTop <= 0 { + return padT + plotH + } + return padT + plotH - float64(words)/float64(yTop)*plotH + } + + // Lay snapshots out in slots: one per snapshot, plus a gutter between + // sessions for the break marker. + slots := float64(len(d.Versions)) + gutterSlots*float64(len(d.Sessions)-1) + sw := plotW / slots + + xs := make([]float64, len(d.Versions)) + bandStart := make([]float64, len(d.Sessions)) + bandEnd := make([]float64, len(d.Sessions)) + + cursor, vi := 0.0, 0 + for si, s := range d.Sessions { + if si > 0 { + cursor += gutterSlots + } + bandStart[si] = padL + cursor*sw + for k := 0; k < s.Snapshots; k++ { + xs[vi] = padL + (cursor+0.5)*sw + cursor++ + vi++ + } + bandEnd[si] = padL + cursor*sw + } + + var b strings.Builder + fmt.Fprintf(&b, `
+

How the draft grew

+ +`, chartW, chartH) + + // Session bands sit behind everything; each carries a native tooltip. + for i, s := range d.Sessions { + w := bandEnd[i] - bandStart[i] + if w < minBandW { + w = minBandW + } + fmt.Fprintf(&b, `Session %d: %s, %s, %d snapshots +`, bandStart[i], padT, w, plotH, roseWash, i+1, + htmlEscape(formatWhen(s.Start)), htmlEscape(humanDuration(s.Duration())), s.Snapshots) + } + + // Recessive gridlines with y labels at 0 / half / top. + for _, gv := range []int{0, yTop / 2, yTop} { + gy := y(gv) + fmt.Fprintf(&b, ` +%d +`, padL, gy, chartW-padR, gy, roseLight, padL-8, gy+4, gv) + } + + // Break markers in the gutters, so compressed time is stated, not hidden. + for i := 1; i < len(d.Sessions); i++ { + mid := (bandEnd[i-1] + bandStart[i]) / 2 + gap := d.Sessions[i].Start.Sub(d.Sessions[i-1].End) + fmt.Fprintf(&b, ` +%s +`, mid, padT, mid, padT+plotH, roseLight, mid, padT+plotH+13, htmlEscape(humanDuration(gap)+" away")) + } + + // Step path: hold the previous value until the next snapshot lands. + var path strings.Builder + fmt.Fprintf(&path, "M %.1f %.1f", xs[0], y(d.Versions[0].WordCount)) + for i := 1; i < len(d.Versions); i++ { + fmt.Fprintf(&path, " L %.1f %.1f L %.1f %.1f", + xs[i], y(d.Versions[i-1].WordCount), xs[i], y(d.Versions[i].WordCount)) + } + fmt.Fprintf(&b, ` +`, path.String(), rose) + + // Pre-empt the obvious question: label the largest single jump when it is a + // big share of the finished draft, rather than letting a reader find it. + if d.NoteJump && d.LargestJumpIdx < len(xs) { + jx, jy := xs[d.LargestJumpIdx], y(d.Versions[d.LargestJumpIdx].WordCount) + + // Flip the label inboard near the right edge so it can't overflow, and + // push it below the point when the point sits near the top. + anchor, dx := "start", 9.0 + if jx > float64(chartW)*0.6 { + anchor, dx = "end", -9.0 + } + ly := jy - 10 + if ly < padT+12 { + ly = jy + 18 + } + + fmt.Fprintf(&b, ` +largest single addition: +%d words +`, jx, jy, rose, surface, jx+dx, ly, anchor, d.LargestJump) + } + + fmt.Fprintf(&b, ` +

Each shaded band is one writing session, %s to %s. Width follows +snapshots saved, so time spent writing gets the space and breaks are compressed to +the labelled gaps.

+
+`, htmlEscape(formatDay(d.FirstAt)), htmlEscape(formatDay(d.LastAt))) + + return b.String() +} + +func renderSessions(d passportData) string { + var b strings.Builder + b.WriteString(`
+

Writing sessions

+ + + +`) + for i, s := range d.Sessions { + fmt.Fprintf(&b, ` +`, i+1, htmlEscape(formatWhen(s.Start)), htmlEscape(humanDuration(s.Duration())), + s.Snapshots, s.WordsAdded) + } + b.WriteString("
#StartedLengthSnapshotsNet words
%d%s%s%d%+d
\n
\n") + return b.String() +} + +// renderIntegrity explains the hash chain in plain language, including when it +// cannot vouch for something. +func renderIntegrity(d passportData) string { + var headline, detail string + + switch d.ChainStatus { + case chainVerified: + headline = "History intact" + detail = "Every snapshot matches its own contents and links correctly to the one before it. Nothing in this history has been altered or removed since it was recorded." + case chainGaps: + headline = "History intact, with gaps" + detail = "Every snapshot matches its own contents, but some older automatic snapshots have been cleared to save space, so the record is not continuous. Turn on “keep full history” for this document to stop that happening." + case chainPartial: + headline = "Partly verifiable" + detail = fmt.Sprintf("%d snapshot(s) were recorded before this document started tracking integrity, so they cannot be checked. Everything recorded since then matches.", d.UnhashedCount) + case chainUnverifiable: + headline = "Not verifiable" + detail = "These snapshots were recorded before integrity tracking existed. The timeline above is still the record that was saved as you wrote; it simply cannot be checked for later alteration." + case chainBroken: + headline = "Integrity check failed" + detail = "At least one snapshot does not match what was recorded for it. This can mean the history was edited after the fact, or that the database was restored from a backup or copied between machines." + } + + return fmt.Sprintf(`
+

%s

+

%s

+
+`, htmlEscape(d.ChainStatus), htmlEscape(headline), htmlEscape(detail)) +} + +// passportLimits states plainly what the report does and does not establish. +// Overclaiming would be worse than useless: a reader who catches the report +// overstating its case discounts the whole thing. +const passportLimits = `
+

How to read this

+

A document written over time leaves a trail: many snapshots, uneven growth, +words added and cut and added again across separate sittings. A document that was +pasted in from elsewhere tends to arrive nearly whole, in one or two snapshots, +with little revision after.

+

What this report shows is the record Petal saved automatically while the +document was open, roughly every few minutes of active editing.

+

What it does not show. It cannot prove who was at the +keyboard, and it cannot tell whether text typed into the editor was composed +there or copied from another window. It is evidence of a writing process, not a +certificate of authorship. It is most useful read alongside the drafts +themselves.

+
+` + +// --- formatting helpers ----------------------------------------------------- + +func formatWhen(t time.Time) string { return t.Local().Format("2 Jan 2006, 3:04 PM") } +func formatDay(t time.Time) string { return t.Local().Format("2 Jan 2006") } + +// humanDuration renders a span at the coarsest useful precision — a reader cares +// that a session ran "2h 40m", never that it ran 2h40m12s. +func humanDuration(d time.Duration) string { + if d < time.Minute { + return "under a minute" + } + days := int(d.Hours()) / 24 + hours := int(d.Hours()) % 24 + mins := int(d.Minutes()) % 60 + + switch { + case days > 0 && hours > 0: + return fmt.Sprintf("%dd %dh", days, hours) + case days > 0: + return fmt.Sprintf("%dd", days) + case hours > 0 && mins > 0: + return fmt.Sprintf("%dh %dm", hours, mins) + case hours > 0: + return fmt.Sprintf("%dh", hours) + default: + return fmt.Sprintf("%dm", mins) + } +} + +func pluralize(n int, one, many string) string { + if n == 1 { + return one + } + return many +} + +// niceCeil rounds a maximum up to a round number so gridlines land on values a +// reader can hold in their head. +func niceCeil(n int) int { + if n <= 0 { + return 0 + } + mag := math.Pow(10, math.Floor(math.Log10(float64(n)))) + return int(math.Ceil(float64(n)/(mag/2)) * (mag / 2)) +} + +// passportHead is the page shell: one %s for the title. Print rules keep the +// chart and the caveats on the page rather than letting them break across sheets. +const passportHead = ` + + + + +Writing passport — %s + + + +` diff --git a/internal/docs/passport_test.go b/internal/docs/passport_test.go new file mode 100644 index 0000000..edada25 --- /dev/null +++ b/internal/docs/passport_test.go @@ -0,0 +1,376 @@ +package docs + +import ( + "encoding/json" + "fmt" + "net/http" + "net/http/httptest" + "regexp" + "strconv" + "strings" + "testing" + "time" + + "gitea.parodia.dev/drwily/petal/internal/db" +) + +// chained builds a valid hash-chained snapshot run from (minutes-offset, words) +// pairs, so tests can describe a writing history in terms a reader recognises +// and get correct hashes for free. +func chained(docID string, base time.Time, points ...[2]int) []db.DocumentVersion { + var ( + out []db.DocumentVersion + prev string + ) + for i, p := range points { + at := base.Add(time.Duration(p[0]) * time.Minute) + text := strings.Repeat("word ", p[1]) + v := db.DocumentVersion{ + ID: fmt.Sprintf("v%d", i), + DocID: docID, + ContentText: text, + WordCount: p[1], + Kind: db.VersionKindAuto, + CreatedAt: at, + PrevHash: prev, + } + v.ContentHash = chainHash(prev, docID, at, p[1], text) + prev = v.ContentHash + out = append(out, v) + } + return out +} + +func TestBuildPassportSessions(t *testing.T) { + base := time.Date(2026, 3, 2, 9, 0, 0, 0, time.UTC) + doc := db.Document{ID: "d1", Title: "Essay"} + + // Two sittings: 09:00–09:30, then a three-hour break, then 12:30–13:00. + vs := chained("d1", base, + [2]int{0, 40}, [2]int{15, 120}, [2]int{30, 210}, + [2]int{210, 260}, [2]int{240, 330}, + ) + + d := buildPassport(doc, vs) + + if len(d.Sessions) != 2 { + t.Fatalf("sessions = %d, want 2", len(d.Sessions)) + } + if got := d.Sessions[0].Duration(); got != 30*time.Minute { + t.Errorf("session 1 duration = %v, want 30m", got) + } + if got := d.Sessions[1].Snapshots; got != 2 { + t.Errorf("session 2 snapshots = %d, want 2", got) + } + if got := d.Span; got != 4*time.Hour { + t.Errorf("span = %v, want 4h", got) + } + // Active time counts only time inside sessions, never the break. + if got := d.ActiveTime; got != 60*time.Minute { + t.Errorf("active time = %v, want 60m", got) + } + // Session 2 measures from session 1's final count (210 → 330). + if got := d.Sessions[1].WordsAdded; got != 120 { + t.Errorf("session 2 words = %d, want 120", got) + } + if d.ChainStatus != chainVerified { + t.Errorf("chain = %q, want %q", d.ChainStatus, chainVerified) + } +} + +func TestBuildPassportFlagsLargeJump(t *testing.T) { + base := time.Date(2026, 3, 2, 9, 0, 0, 0, time.UTC) + doc := db.Document{ID: "d1"} + + t.Run("steady growth is not flagged", func(t *testing.T) { + vs := chained("d1", base, [2]int{0, 100}, [2]int{5, 200}, [2]int{10, 300}, [2]int{15, 400}) + if d := buildPassport(doc, vs); d.NoteJump { + t.Errorf("even growth flagged a jump of %d", d.LargestJump) + } + }) + + t.Run("a paste-shaped jump is flagged", func(t *testing.T) { + vs := chained("d1", base, [2]int{0, 20}, [2]int{5, 40}, [2]int{10, 900}) + d := buildPassport(doc, vs) + if !d.NoteJump { + t.Fatal("large jump not flagged") + } + if d.LargestJump != 860 { + t.Errorf("largest jump = %d, want 860", d.LargestJump) + } + if !d.LargestJumpAt.Equal(base.Add(10 * time.Minute)) { + t.Errorf("jump at %v, want +10m", d.LargestJumpAt) + } + }) +} + +func TestVerifyChain(t *testing.T) { + base := time.Date(2026, 3, 2, 9, 0, 0, 0, time.UTC) + good := func() []db.DocumentVersion { + return chained("d1", base, [2]int{0, 50}, [2]int{5, 90}, [2]int{10, 160}) + } + + t.Run("intact chain verifies", func(t *testing.T) { + got, _ := verifyChain(db.Document{ID: "d1"}, good()) + if got != chainVerified { + t.Errorf("got %q, want %q", got, chainVerified) + } + }) + + t.Run("edited content breaks it", func(t *testing.T) { + vs := good() + vs[1].ContentText = "something else entirely" + if got, _ := verifyChain(db.Document{ID: "d1"}, vs); got != chainBroken { + t.Errorf("got %q, want %q", got, chainBroken) + } + }) + + t.Run("backdating breaks it", func(t *testing.T) { + vs := good() + vs[2].CreatedAt = base.Add(-time.Hour) + if got, _ := verifyChain(db.Document{ID: "d1"}, vs); got != chainBroken { + t.Errorf("got %q, want %q", got, chainBroken) + } + }) + + t.Run("a removed snapshot reads as a gap when pruning is allowed", func(t *testing.T) { + vs := good() + pruned := []db.DocumentVersion{vs[0], vs[2]} // middle snapshot gone + got, _ := verifyChain(db.Document{ID: "d1"}, pruned) + if got != chainGaps { + t.Errorf("got %q, want %q", got, chainGaps) + } + }) + + t.Run("a removed snapshot is tampering when history is preserved", func(t *testing.T) { + vs := good() + pruned := []db.DocumentVersion{vs[0], vs[2]} + got, _ := verifyChain(db.Document{ID: "d1", PreserveHistory: true}, pruned) + if got != chainBroken { + t.Errorf("got %q, want %q", got, chainBroken) + } + }) + + t.Run("pre-chain snapshots are partial, not failures", func(t *testing.T) { + vs := good() + vs[0].ContentHash, vs[0].PrevHash = "", "" + got, unhashed := verifyChain(db.Document{ID: "d1"}, vs) + if got != chainPartial { + t.Errorf("got %q, want %q", got, chainPartial) + } + if unhashed != 1 { + t.Errorf("unhashed = %d, want 1", unhashed) + } + }) + + t.Run("no hashes at all is unverifiable", func(t *testing.T) { + vs := good() + for i := range vs { + vs[i].ContentHash, vs[i].PrevHash = "", "" + } + if got, _ := verifyChain(db.Document{ID: "d1"}, vs); got != chainUnverifiable { + t.Errorf("got %q, want %q", got, chainUnverifiable) + } + }) +} + +// The report must survive the degenerate histories — one snapshot, an empty +// document — rather than dividing by a zero span or a zero maximum. +func TestRenderPassportEdgeCases(t *testing.T) { + base := time.Date(2026, 3, 2, 9, 0, 0, 0, time.UTC) + + t.Run("no history", func(t *testing.T) { + out := string(renderPassport(buildPassport(db.Document{Title: "Empty"}, nil))) + if !strings.Contains(out, "no saved history") { + t.Errorf("missing empty-state copy:\n%s", out) + } + }) + + t.Run("single snapshot", func(t *testing.T) { + vs := chained("d1", base, [2]int{0, 12}) + out := string(renderPassport(buildPassport(db.Document{ID: "d1", Title: "One"}, vs))) + if strings.Contains(out, "NaN") || strings.Contains(out, "+Inf") { + t.Errorf("degenerate geometry leaked into output:\n%s", out) + } + }) + + t.Run("empty document", func(t *testing.T) { + vs := chained("d1", base, [2]int{0, 0}, [2]int{5, 0}) + out := string(renderPassport(buildPassport(db.Document{ID: "d1"}, vs))) + if strings.Contains(out, "NaN") { + t.Errorf("zero word count produced NaN:\n%s", out) + } + }) + + t.Run("title is escaped", func(t *testing.T) { + doc := db.Document{ID: "d1", Title: ``} + out := string(renderPassport(buildPassport(doc, chained("d1", base, [2]int{0, 5})))) + if strings.Contains(out, "