A card vanishing and coming back seconds later, with different words, was
never about latency: every pass deleted its whole family and re-inserted
it, so each round minted new row ids. The rail keys on suggestion.id, so a
full remount was guaranteed — new id, new created_at (hence the re-fired
chime), and a fresh explanation from a model that re-reasons every time it
is asked. One unchanged mistake carried three different explanations in a
single sitting.
Passes now reconcile instead of replace. A re-proposed edit keeps its row:
its id, its created_at, and the wording she has already read. And the
grammar checkpoint stops asking about sentences nobody touched — the
document is split into hashed sentences, checked_chunks records which ones
a family has read, and only the difference is sent. When nothing changed
it doesn't call the model at all, and doesn't spend its rate-limit slot on
having done nothing.
The tone is part of a sentence's identity: cached advice was written for
the old register, so switching doc type re-reads every line.
replaceMechanics reconciles too, which mattered more than expected — the
rule pack fires 250 ms after a keystroke, so it was re-minting every local
card's id several times a sentence.
Only the grammar checkpoint is chunked. Voice is a property of the whole
document, and the collocation coach is a button she pressed asking for a
fresh read.
No client change was needed; stable ids were the whole of it.
Claude-Session: https://claude.ai/code/session_016y6gyuHkQXPiEuW8RGQyua
A security review of the whole repo. The queries were already scoped, the
OIDC flow already did state and nonce and PKCE, the session tokens were
already stored as hashes. What it found was mostly the seam between the
code and the deployment — and one place where the deployment quietly
undid the code.
The one that matters: with any AUTHENTIK_* variable missing, Petal fell
back to resolving every request to the single `local` user. That is right
on a laptop and a catastrophe on a public host, and Phase 16 removed the
Traefik basic-auth gate that used to stand behind the mistake. A typo in
the client secret would have served her journals to the open internet and
said so only in a log line nobody reads. It now refuses to start, guarded
by default for any BASE_URL that isn't loopback.
Then the one that would have been fixed and wasn't: stored images now
serve under `default-src 'none'; sandbox`, so an SVG pasted into a
document can't run as a page on Petal's own origin. Traefik's
customresponseheaders *overwrites*, so the CSP declared in the compose
labels would have silently replaced that per-route policy in production.
The whole header block moved into the binary, where a route can tighten
its own and a test can prove it; only HSTS stays at the edge, where TLS
actually terminates.
The rest, smaller:
- PETAL_ALLOWED_SUBS empty means everyone authentik authenticates, and
authentik here fronts half a dozen applications. Still legal, now
said out loud every boot, and set in both env examples.
- LLM failures relayed err.Error() to the browser, which carries the
address of the inference box on the far side of the VPN. Logged
instead; the client only ever rendered "the helper is resting".
- Exports scheme-check their links. Escaping makes a URL safe to sit
in an attribute and says nothing about following it, and an export
is the one artifact here meant to leave. Writing the test found the
markdown image src, which I'd missed reading it.
- The draft rescue is namespaced per account and cleared on sign-out.
Everything else in localStorage is a preference; this is her unsaved
writing, sitting in a profile two people share.
- /auth/logout is POST-only. With SameSite=Lax a GET route lets any
page on the internet sign her out mid-draft.
- Image uploads get a per-account allowance and the TTS cache a size
cap. Both share the encrypted volume the database is on, and a full
disk is SQLite failing to write, not a feature degrading.
- The session cookie takes the __Host- prefix over https, so nothing
else under parodia.dev can plant one. Old cookies still resolve;
nobody is signed out to get there.
- npm audit: linkify-it and postcss.
Verified: go build, go vet, the full Go suite, tsc, 195 frontend tests,
npm audit clean. The startup guard and both CSPs checked against a
running server rather than only asserted.
Claude-Session: https://claude.ai/code/session_016y6gyuHkQXPiEuW8RGQyua
Grammar lite, the false-friend list, the daily invitation and the offline
miscollocations — the four remaining §5–§6 items, all client-side and all
alive on a box that cannot reach the model.
The offline collocations forced a schema change. `type` had been doubling
as the answer to "which engine found this" — `mechanics` meant offline —
and that stops being true the moment an offline rule proposes a
collocation. Migration 0013 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. Existing rows backfill by type, so a pre-0013
collocation row is claimed as the coach's, which it was: the offline list
did not exist yet.
The rule pack is hand-curated rather than mined, and the entries left out
are the point — `married with` is wrong until "married with children",
`arrive to` wants at or in depending on the noun. A pack running on every
keystroke must not correct correct writing.
Claude-Session: https://claude.ai/code/session_016y6gyuHkQXPiEuW8RGQyua
Two halves of the same idea, both read out of work Petal already
records.
Planting: an accepted collocation is a learnable chunk, so it becomes a
phrase card. The scheduler didn't need to know — a three-word chunk
climbs the ladder exactly like a looked-up word. What needed care was
deciding what *isn't* a chunk (single words are word choice; a
six-word-plus "collocation" is a rewritten sentence, and sentences make
miserable flashcards), and that the example must be the *corrected*
sentence — the stored draft still holds the phrasing she just left
behind. Re-accepting the same chunk leaves the existing card alone
rather than resetting a schedule it has been climbing. The whole thing
is best-effort: accepting an edit must never fail because a flashcard
couldn't be made.
The growth journal: kept this month beside kept the month before, the
phrasing that stuck, the patterns that faded. The queries were the easy
part; the honesty is the feature. "Stuck" needs the phrase in a *second*
document, because one document is just the edit where she left it.
"Faded" says nothing at all unless she has been writing lately —
otherwise a month away from Petal comes back to her as progress, which
is the one way this could lie. And a suggestion had to start recording
when she *decided* it, not when the model proposed it, so 0012 adds
resolved_at and backfills the old rows to their created_at.
It lives as a second tab in the garden, and it feeds the kitten: after
an accept she now sometimes hears something true of her alone, once per
line, half the time, never waited for.
Claude-Session: https://claude.ai/code/session_016y6gyuHkQXPiEuW8RGQyua
Every `中文 · English` string moves out of ~29 components into
web/src/i18n: one Pack type, a verbatim zh pack, and two ways to read
it — usePack() for components, pack() for the modules that build a line
when something happens rather than when something renders.
Anything with a value in it is a function on the pack rather than a
template at the call site, English pluralisation included: word order
isn't universal, and a pack author has to be able to move the number.
The roster constants (tones, rewrite styles, export formats, companions)
keep only value + emoji, so a label can't drift from its key.
On the server, internal/llm/lang.go replaces "Simplified Chinese" in the
three prompts that actually name her language. pt-PT is spelled
"European Portuguese (pt-PT, never Brazilian Portuguese)" in the prompt
itself, and each Lang carries her word for "why" so the tutor prompt
still recognises the question when she asks it her way.
pair_lang reaches the model through the row-scoped query each handler
already ran — the one that proves she owns the document — rather than a
second lookup that could disagree with it.
Also records Phase 18's deploy: migration 0011 rehearsed against a copy
of the live VPS database, then applied for real.
Petal ran as a single hardcoded user, with db.LocalUserID named directly
at ~35 query sites. That made the caller's identity a compile-time
constant scattered across every package — nothing a real login could
replace without touching all of them.
New internal/auth moves it into the request context:
- Middleware(Resolver) resolves the caller once per API request
- handlers read auth.UserID(r.Context()) instead of naming a user
- Resolver is the seam an Authentik session check drops into
- StaticResolver(db.LocalUserID) keeps Petal single-user today
Behavior is unchanged. UserID returns "" rather than panicking when the
middleware is absent, so a mis-wired route fails closed: every query is
WHERE user_id = ?, which then matches nothing.
main.go splits /api into a public group (/health, /version) and an
authenticated group for everything else — a monitoring probe must not
need a session.
Two pre-existing access-control gaps fixed while threading, both
harmless with one user and not with two:
- setStatus (accept/dismiss) updated a suggestion by bare id with no
ownership check at all
- listForDoc/fetchPending read a document's suggestions by doc_id
alone; a suggestion quotes the sentence it corrects, so that leaked
the source prose
Both now scope through documents.user_id.
Tests: internal/auth covers the context round-trip, the absent-context
case, and both 401 paths. Two-user isolation suites in docs and
suggestions mount the same routers twice behind two resolvers over one
database and assert a stranger gets 404 on every id-taking path, sees
nothing in list/search, and leaves the owner's data untouched.
Those suites earned their keep immediately: docs.fetch gained a userID
parameter but kept binding db.LocalUserID in the query. Unused
parameters are legal Go, so it compiled clean, vet was silent, and every
existing test passed while the lookup stayed unscoped.
Still global, out of scope and flagged in BUILD_PLAN.md: the image store
has no per-user association, and frontend localStorage keys are
per-browser rather than per-account.
Suppression keyed on the exact original->replacement pair, which the
model routinely sidestepped: it reverses an accepted edit (reverse
pair), re-polishes its own accepted output (new original == accepted
replacement), and the editor's smart-quote churn ("..." -> '...')
defeated even a byte-exact match. Result: a few sentences got nudged
back and forth pass after pass.
Replace the exact-pair actionedKeys with a suppressor that compares
under a normalization folding all quote variants and collapsing
whitespace, and drops a fresh suggestion when it re-touches an
already-settled span: same edit re-proposed, an original the user
already accepted/dismissed, the model re-touching its own accepted
output, or a multi-word sub-clause contained in an accepted span.
Tradeoff: once a sentence is accepted/dismissed it won't be re-flagged
until its text changes — stability over marginal improvement, the right
call for the calm ESL persona.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
Reuse the companion's prose.ts rules engine as the single source of
deterministic detection instead of duplicating it. Applyable rules now
also emit exact-span fixes (original -> replacement) that surface as
suggestion cards; awareness-only rules (run-ons, splices, ...) stay
companion bubbles. The companion hides fix-bearing hints so a span is
never both a bubble and a card.
Spans are widened to a distinctive phrase ("a old" -> "an old",
"She have" -> "She has") so they re-anchor by string in the editor; a
lone lowercase "i" stays awareness-only since a single char can't anchor.
Backend: detection lives client-side, so the new persist-only
POST /docs/{id}/mechanics endpoint receives findings and stores them as
the 'mechanics' family with their exact offsets. It honours
actioned-suppression, leaves the LLM families untouched, and a checkpoint
no longer wipes it. fetchPending dedupes spans with mechanics winning any
collision against an LLM card (its span is exact). Migration 0008 adds the
'mechanics' suggestion type.
Client renders the mechanics fixes immediately (no LLM wait) and the cards
use a calm sage "Tidy-up" accent.
Verified end-to-end in a real browser on millenia: detect -> persist ->
render -> accept applies the fix.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
Backend:
- Extract shared internal/httputil (WriteJSON/ErrorJSON/BadRequest/
ServerError); drop the triple-duplicated helpers in docs, suggestions,
vocab. ServerError now logs the real error and returns a generic 500 so
raw DB/internal errors never reach the client.
- vocab capture: validate doc_id ownership (blank -> none, unknown -> 400
instead of a leaked FK 500); rune-safe clamp word/gloss/definition/
phonetic/example.
- vocab review(): wrap the read-modify-write in a transaction (TOCTOU).
- /api request-size cap via MaxBytesReader middleware (2 MiB), exempting
/api/images (own 10 MiB limit).
Frontend:
- StatusBar: drive the checking/voicing/collocating indicators from one
array; llmDown uses !anyBusy.
- Slide-overs: new useFocusTrap hook (focus-in, Tab trap, focus-restore)
on GardenPanel + HistoryPanel, both role=dialog/aria-modal/aria-label.
- speech.ts: export stopSpeech(); GardenPanel cancels audio on unmount.
Tests: add doc_id-validation and field-clamp coverage; full suite green.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
Phase 12 — collocation coach: a third suggestion family for gentle
"natives usually say…" hints on non-native word pairings, reusing the
existing runPass/pendingScope/rail machinery.
- llm/collocation.go (RunCollocation, 25s floor, reuses ParseCheckpoint)
+ collocationSystemPrompt/CollocationMessages (warm, Mandarin gloss,
defers grammar/spelling to the grammar family)
- migration 0005 rebuilds the suggestions table to extend the type CHECK
(SQLite can't ALTER a CHECK)
- collocationScope + CollocationLimit + POST /{id}/collocation
- fix: grammarScope was `type != 'voice'` and would wipe the new
collocation flags; now `type NOT IN ('voice','collocation')`
- frontend: --color-blossom, "Make it sound natural 🌸" pill,
collocating/runCollocation in useCheckpoint, StatusBar dot
Phase 13 — vocabulary garden: capture looked-up words and surface them
for gentle spaced repetition.
- new internal/vocab package: migration 0006 (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 then geometric; gentle
"again", no streak-shaming), handlers (capture-upsert/list/due/
review/delete, owner-scoped, SQLite-side datetime math)
- auto-capture on word lookup (dictionary-known words only, captures
the surrounding sentence + doc_id) + 🤍/💚 toggle on WordCard
- GardenPanel: blossom grid (bloom by reps), flashcard review (sentence
blanked, flip, again/good/easy, recognition↔production), sleepy-kitten
footer; opened from a global 🌷 header button
Tests: TestCollocationPassCoexists, vocab scheduler + handlers, db CHECK
extended. go build/vet/test + tsc + vite + vitest (51/51) clean;
migration verified against a copy of the live DB; live backend smoke
walked the full vocab lifecycle + the warm-502 collocation path.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
A paste fires exactly one grammar checkpoint, and a failed one never retried
until the next keystroke — stranding the writer on "Petal's helper is resting"
after a paste. Long docs make it worse: their 15-25s checks have a wide window
to catch a transient 502 from the shared Ollama (co-tenant apps load other
models and evict the 9B). A failed pass also burned the per-document rate-limit
slot, so a retry within 30s hit the throttle path and got an empty set back.
- llm.RateLimiter.Release rolls back a slot when its pass fails; Allow now
returns the recorded timestamp so Release only frees its own slot.
- suggestions.runPass releases the slot on LLM failure before returning 502.
- useCheckpoint auto-retries a failed checkpoint with backoff (3/12/35s),
keeping the breathing dot up and only flagging "resting" once retries exhaust.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
Surface every outstanding suggestion as a card in the right-hand
whitespace, vertically aligned to the text it flags — so the writer sees
the whole queue at once instead of hovering each highlight. Cards stack
with collision avoidance, link both ways with their highlight (hover/click
↔ soft text wash, driven through the decoration plugin so it survives
edit repaints), and carry the same Accept / Dismiss / Ask Petal actions.
The rail is a progressive enhancement: it mounts only when there's room
beside the editor, otherwise the existing inline hover card is unchanged.
Stacked cards that reach the bottom-right corner tuck behind the
companion mascot (z-order).
When a card is expanded, the Ask Petal bubble now opens with the
Simplified-Chinese translation of the explanation (the English stays in
the card body) instead of repeating the same text twice — a new
POST /api/suggestions/{id}/translate one-shot LLM endpoint, loaded
lazily on open with an English fallback.
Verified live against the local LLM via the uitest harness: rail
stacking, hover↔text wash, expand/Ask Petal, accept-from-rail, narrow
fallback, and the Mandarin bubble.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
Two fixes for the "accept Petal's change, then it nags about the same
sentence moments later" report:
- replacePending now suppresses any freshly-generated suggestion whose
original->replacement matches one the user already accepted or
dismissed for that doc. The model has no memory between passes, so
without this it re-proposes the identical edit on the next checkpoint.
- findRange anchored suggestions by exact string match, which missed
whenever the model echoed an `original` with plain ASCII (straight
quotes, --, ...) while the document held the Typography-converted
glyphs (curly quotes, em-dash, single-char ellipsis). The miss meant
no highlight AND a silent no-op on accept, which then fed the re-nag
above. foldTypography canonicalizes those variants (with a source
index map for length changes) so matching survives the mismatch.
Covered by a server-side regression test (accept+dismiss then re-check
returns nothing) and frontend unit tests for the fold and anchoring.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
Inline Chinese gloss (offline) and a "say it more naturally" / tone-rewrite,
the two ESL features for the Mandarin-speaking writer.
Gloss: embedded English→Chinese dictionary (gloss.json.gz, 57k common words
built from ECDICT via scripts/build_gloss.py). lexicon gains Gloss()/Result.Gloss
and a lightweight GET /api/gloss/{word}; the right-click WordCard leads with the
中文; GlossTip shows it on a 350ms hover (reuses wordAt, so CJK is never glossed).
Offline + instant, works with the LLM down.
Rewrite: selecting text pops a SelectionBubble (✨更自然 + the tone vocabulary);
picking a style calls POST /api/docs/:id/rewrite (llm.RunRewrite, stateless,
owner-scoped) and shows a RewritePreview (original→rewrite, accept/cancel/retry).
Accept applies it in-editor.
Tests added in lexicon and suggestions. go build/vet/test, tsc, vite all clean;
live smoke vs a fake vLLM verified gloss + rewrite + 400/404/502 paths.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
Four enhancements to make the editor fit real school usage:
- Per-document tone (academic/professional/casual/humorous/creative/
persuasive/general): new documents.tone column (migration 0002), threaded
through the docs API, a bilingual ToneSelect dropdown on the title row, and
injected into the grammar-checkpoint LLM prompt so advice fits the register.
The voice pass stays tone-agnostic.
- Right-click word lookup: a new offline `lexicon` package serves definitions
(Wordset, modern ESL-friendly glosses) and synonyms (WordNet synsets first,
then frequency+stopword-ranked Moby for breadth) from gzipped embedded data,
behind /api/word/{word} with light morphology. The WordCard popover shows the
definition and tappable synonym pills that swap the word in place.
- Expanded writing stats: clicking the word count opens a StatsPanel with page
count, sentences, paragraphs, reading time, average word length, word variety,
and Flesch-Kincaid reading level — all computed client-side.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
Tier-1 voice-consistency pass: whole-document LLM review surfacing passages
that read tonally out of place (formal/over-polished/paraphrased-too-closely),
as honey-decorated `voice` flags with no correction (awareness-only).
- internal/llm/voice.go: RunVoice sends the whole document (no TruncateDoc),
MaxTokens 2048, 20s per-doc floor (VoiceInterval). Standalone voice prompt
in prompts.go (not bundled with the grammar checkpoint, per spec).
- internal/suggestions: POST /api/docs/:id/voice. replacePending is now
family-scoped (pendingScope) so grammar and voice never clobber each other's
pending flags; both passes return the unified pending set. check/voice share
one runPass helper. TestVoicePassCoexists covers both directions.
- Frontend: api.voiceDoc, useCheckpoint voicing/runVoice, honey "Check my
voice" toolbar pill, breathing honey dot in StatusBar.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
Conversational follow-up on a suggestion, streamed token-by-token.
Backend (interface-only; handlers never touch a concrete LLM client):
- internal/llm/chat.go: StreamAskPetal with conversational sampling
(max_tokens 512, temp 0.7, rep 1.15, top_p 0.92, stop "\n\n\n"),
reusing AskPetalSystemPrompt + TrimHistory.
- internal/suggestions/chat.go: POST /api/suggestions/:id/chat. One
user-scoped join loads the suggestion + parent content_text;
surroundingParagraph extracts the \n\n-bounded paragraph at from_pos
(whole-doc fallback when unlocated) and injects it server-side.
Streams event: token / event: done SSE frames with JSON-encoded data
so token newlines can't break framing; real http.Flusher per chunk.
LLM-unreachable -> 502 before SSE headers; unknown suggestion -> 404.
Frontend:
- streamSuggestionChat: fetch + ReadableStream SSE parser (not
EventSource, needs POST), abortable.
- AskPetal.tsx: whole conversation in component state (no persistence,
cleared on close), Petal's first bubble pre-seeded with the
explanation, rose/lavender bubbles, CJK font stack on the bubbles
only (Note #17), streaming caret.
- SuggestionCard "Ask Petal" pill pins the card open while chatting
(hover-close suppressed, click-away closes) and widens it to 340px.
Tests: chat_test.go covers streamed-text concat + done event,
server-side context injection on the system message, sampling params,
404, and surroundingParagraph. go build/vet/test clean, tsc clean,
vite build OK. Live SSE smoke-tested against a fake streaming vLLM:
tokens flushed individually through the chi middleware stack, done
terminator, 502 on LLM-down, 404 on unknown suggestion.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
Backend (internal/llm): backend-agnostic LLMClient interface + factory
with vLLM (OpenAI-compat) and Ollama (native) clients, each Complete +
Stream. prompts.go holds the checkpoint and Ask Petal templates;
checkpoint.go salvages JSON from model output (brace-matched), enforces a
per-doc 30s RateLimiter, and truncates the doc to a latency cap.
internal/suggestions: POST /api/docs/:id/check runs a checkpoint and
replaces the doc's pending suggestions in one tx (accepted/rejected kept
as history); GET /api/docs/:id/suggestions lists pending;
POST /api/suggestions/:id/{accept,dismiss} resolves one. Throttled checks
return the current set rather than erroring.
Frontend: useCheckpoint (4s debounce, loads existing on open, stale-guard
tokens); SuggestionHighlight renders ProseMirror decorations re-anchored
by the `original` string on every doc change (not stored marks), with
precise textblock-offset→PM-position mapping; SuggestionCard shows the
type tag + diff + explanation and applies the replacement in-editor on
accept; breathing rose checkpoint dot in the StatusBar; fade-float +
breathe animations.
Tests: llm parse/rate-limit/truncate; suggestions full flow + rate-limit
over httptest with a stub client. Smoke-tested end-to-end against a fake
vLLM endpoint (anchoring verified) and the LLM-unreachable 502 path.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
prosolis