feat: add Dockerfile and Hugging Face Spaces config

.dockerignore

@@ -0,0 +1,9 @@
+.venv/
+.env
+__pycache__/
+*.pyc
+.git/
+.DS_Store
+data/
+docs/
+*.wav

Dockerfile

@@ -0,0 +1,28 @@
+FROM python:3.11-slim
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+      espeak-ng \
+      curl \
+      ca-certificates \
+    && rm -rf /var/lib/apt/lists/*
+
+RUN useradd -m -u 1000 hal
+WORKDIR /app
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY download_model.py .
+RUN python download_model.py
+
+COPY hal_prompt.py main.py ./
+COPY static/ ./static/
+COPY profile.md ./
+
+RUN mkdir -p /app/data && chown -R hal:hal /app
+USER hal
+
+ENV HAL_DATA_DIR=/app/data
+EXPOSE 7860
+
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -1,8 +1,19 @@
+---
+title: HAL Voice MVP
+emoji: 🔴
+colorFrom: red
+colorTo: gray
+sdk: docker
+app_port: 7860
+pinned: false
+short_description: Push-to-talk conversation with HAL 9000's voice.
+---
+
 # HAL Voice MVP
 
-Push-to-talk web app that lets you talk to HAL 9000. FastAPI backend wires Groq Whisper (STT) → Claude (LLM) → Piper TTS (HAL voice). In-memory sessions, no auth, no database.
+Push-to-talk web app that lets you talk to HAL 9000. FastAPI backend wires Groq Whisper (STT) → Claude (LLM) → Piper TTS (HAL voice). Sessions persist to disk as JSON.
 
-## Setup
+## Local setup
 
 1. Install espeak-ng (macOS):
    ```
@@ -18,13 +29,14 @@ Push-to-talk web app that lets you talk to HAL 9000. FastAPI backend wires Groq
    ```
    python download_model.py
    ```
-4. Create `.env` (copy `.env.example`):
+4. Create `.env`:
    ```
    GROQ_API_KEY=gsk_...
    ANTHROPIC_API_KEY=sk-ant-...
    ```
+5. (Optional) Edit `profile.md` with facts you want HAL to know about you.
 
-## Run
+## Run locally
 
 ```
 .venv/bin/uvicorn main:app --reload --port 8000
@@ -32,9 +44,27 @@ Push-to-talk web app that lets you talk to HAL 9000. FastAPI backend wires Groq
 
 Open http://localhost:8000, hold the red eye to talk, release to send.
 
+## Deploy to Hugging Face Spaces
+
+1. Create a new Space: https://huggingface.co/new-space
+   - SDK: **Docker**
+   - Hardware: **CPU Basic** (free)
+   - Visibility: Public (free) or Private ($9/mo)
+2. Add repository secrets (Settings → Variables and secrets):
+   - `GROQ_API_KEY`
+   - `ANTHROPIC_API_KEY`
+3. Push this repo to the Space:
+   ```
+   git remote add space https://huggingface.co/spaces/<your-username>/<space-name>
+   git push space main
+   ```
+4. The Space builds from the `Dockerfile`. First build takes ~5 min (downloads the HAL model). Subsequent builds are cached.
+5. Open the Space URL, hold the eye, talk.
+
 ## Notes
 
 - English only. The Piper HAL model is English-only by design.
-- Sessions live in process memory; restarting the server wipes history.
+- Sessions live in `data/sessions/` as JSON. On HF Spaces free tier, the filesystem is ephemeral — sessions reset on restart/rebuild. For persistence across restarts, upgrade to a Space with persistent storage ($5/mo) and set `HAL_DATA_DIR=/data`.
 - Piper outputs 22050 Hz mono WAV, which browsers play natively.
-- The HAL voice is loaded once at startup — first request has no extra latency.
+- The HAL voice loads once at startup.
+- Edit `profile.md` to give HAL static context about you; it gets prepended to the system prompt on startup.