Phase 15: containerize Petal for the parodia.dev VPS

Deploy plumbing so Petal can run on the public VPS behind the Traefik
already on that box, with vLLM reached over headscale.

- Dockerfile: node build -> go build -> alpine runtime. CGO stays off
  (modernc SQLite is pure Go), so the runtime layer exists only for
  ffmpeg (read-aloud transcodes Piper's WAV) and tzdata (the companion's
  bedtime nag and night mode read the local clock). Runs as uid 10001
  with /data as the single writable mount.

- docker-compose.yml: Traefik labels following this host's convention
  (external `traefik` network, `web-secure` entrypoint, `default` cert
  resolver). Petal publishes no host port. ./data is a bind mount, not a
  named volume, so the nightly backup and a restore are reachable from
  the host.

- Piper runs as two sibling containers rather than host systemd units.
  The plan assumed Piper was already installed on the VPS; it is not,
  the host has no lingering user session to keep user units alive, and
  containers keep the TTS ports on an internal network unreachable from
  anywhere but Petal. One image, voice chosen per service, model cached
  in a shared volume -- so the pt-PT voice is a new service, not a new
  image.

- db.Backup + a `-backup` flag: VACUUM INTO, not a file copy. Petal runs
  in WAL mode, so the newest committed pages may live in petal.db-wal;
  copying the three files separately can capture a torn mid-checkpoint
  state. VACUUM INTO reads one coherent snapshot without taking a write
  lock, and emits a single file with no -wal/-shm companions. Refuses an
  existing destination so a failed run can't destroy the last good
  backup.

- deploy/backup-petal.sh: nightly snapshot, compress, push to millenia
  over headscale with a post-transfer size check, prune both sides.

- deploy/petal.env.example: LLM_TIMEOUT raised 30s -> 90s for the
  WAN+VPN round trip, since the voice and collocation passes send a
  whole document and the timeout is a hard deadline on Complete.

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']+'/', \
+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"]

deploy/piper/entrypoint.sh

@@ -0,0 +1,22 @@
+#!/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 ${data_dir}"
+  python -m piper.download_voices "${voice}" --data-dir "${data_dir}"
+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}"