feat: deploy LifeOS on hf-pdocker-02 (:8095)

Self-hosted LifeOS (github.com/lifeos-app/lifeos, pinned 3009bf6) in local
mode: no Supabase. Multi-stage build - Node builds the Vite bundle, gunicorn
serves it plus the bundled Flask/SQLite API from one process.

LAN-only on purpose: the backend is single-user with no auth, so anyone who
reaches the port has full read/write. Needs Traefik basicAuth before it can
ever be exposed publicly.

Two upstream bugs patched via sed (build fails loudly if they are renamed):
the chat model was hardcoded to the author's private 'sentient-teddy' bridge,
and a hardcoded 30s timeout failed every first message because Ollama needs
~46s to cold-load the model (0.7s warm). Both are env vars now.

SQLite stays on local disk - NFSv3 breaks WAL locking.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

compose/hf-pdocker-02/lifeos/Dockerfile

@@ -0,0 +1,51 @@
+# LifeOS — https://github.com/lifeos-app/lifeos
+#
+# Two stages: Node builds the Vite/React bundle, Python serves that bundle plus
+# the Flask API from a single process. Build context is the upstream source tree
+# (cloned to /opt/lifeos on the node); this file is copied in alongside it.
+
+# ─── Stage 1: build the frontend ──────────────────────────────────────
+FROM node:22-bookworm-slim AS frontend
+WORKDIR /build
+
+COPY package.json package-lock.json ./
+# --ignore-scripts skips the electron-builder postinstall, which pulls a desktop
+# toolchain we never use and fails without an X stack.
+RUN npm ci --ignore-scripts
+
+COPY . .
+
+# Routes the data layer (src/lib/data-access.ts) to the local Flask API instead
+# of Supabase Cloud. VITE_API_BASE_URL is deliberately left unset: local-api.ts
+# then falls back to a relative /api, so the bundle works on any host and port.
+ENV VITE_USE_LOCAL_API=true
+ENV NODE_OPTIONS=--max-old-space-size=3072
+RUN npm run build
+
+# ─── Stage 2: Flask API + built frontend ──────────────────────────────
+FROM python:3.12-slim
+WORKDIR /app
+
+COPY backend/requirements.txt ./
+RUN pip install --no-cache-dir -r requirements.txt gunicorn
+
+COPY backend/ /app/
+COPY serve_container.py /app/
+COPY --from=frontend /build/dist /app/dist
+
+# Upstream hardcodes the author's own model name in the chat proxy, and a 30s
+# timeout that a cold Ollama model load (~46s for a 9.6GB model) blows through.
+# Make both env vars. Fails the build if upstream renames these out from under us.
+RUN sed -i 's/"model": "sentient-teddy"/"model": os.environ.get("AI_MODEL", "sentient-teddy")/' app.py \
+    && sed -i 's/timeout=30$/timeout=int(os.environ.get("AI_TIMEOUT", "120"))/' app.py \
+    && grep -q 'os.environ.get("AI_MODEL"' app.py \
+    && grep -q 'os.environ.get("AI_TIMEOUT"' app.py
+
+# app.py hardcodes its SQLite path to ~/.lifeos/data.db, so HOME is what pins
+# the database into the mounted volume.
+ENV HOME=/data
+RUN mkdir -p /data/.lifeos
+
+EXPOSE 8080
+CMD ["gunicorn", "--bind", "0.0.0.0:8080", "--workers", "2", "--threads", "4", \
+     "--timeout", "60", "--access-logfile", "-", "serve_container:app"]

compose/hf-pdocker-02/lifeos/docker-compose.yml

@@ -0,0 +1,32 @@
+services:
+  lifeos:
+    build:
+      context: /opt/lifeos
+      dockerfile: Dockerfile
+    image: lifeos:local
+    container_name: lifeos
+    restart: unless-stopped
+    ports:
+      # 8080 on this node is taken by unifi-os-server.
+      - "8095:8080"
+    environment:
+      - TZ=America/Chicago
+      # AI companion -> Ollama on the GPU machine (OpenAI-compatible endpoint).
+      - AI_BRIDGE_URL=http://192.168.86.150:11434
+      - AI_MODEL=gemma4:latest
+      # Ollama needs ~46s to cold-load gemma4 (0.7s once warm); upstream's
+      # hardcoded 30s timeout would fail every first message.
+      - AI_TIMEOUT=120
+    volumes:
+      # SQLite (WAL) stays on local disk: NFS v3 locking breaks it, same reason
+      # guacamole-db lives on /var/lib.
+      - /var/lib/lifeos:/data/.lifeos
+    labels:
+      - "com.centurylinklabs.watchtower.enable=false"
+    mem_limit: 1g
+    healthcheck:
+      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8080/api/health-check').read()"]
+      interval: 60s
+      timeout: 10s
+      retries: 3
+      start_period: 30s

compose/hf-pdocker-02/lifeos/serve_container.py

@@ -0,0 +1,28 @@
+"""Container entrypoint: serve the Flask API and the built SPA from one process.
+
+Upstream's backend/serve.py does this too, but it also shells out to `npm run
+build` and tries to launch a browser, neither of which works in a container.
+"""
+
+from pathlib import Path
+
+from flask import send_file, send_from_directory
+
+from app import app  # importing app runs init_db(), creating the SQLite schema
+
+DIST = Path(__file__).parent / "dist"
+
+
+@app.route("/")
+def _index():
+    return send_file(DIST / "index.html")
+
+
+@app.route("/<path:path>")
+def _static(path):
+    asset = DIST / path
+    if asset.is_file():
+        return send_from_directory(DIST, path)
+    if path.startswith("api/"):
+        return {"error": "Not found"}, 404
+    return send_file(DIST / "index.html")  # SPA client-side routing