README.md

metadata

title: Internal Portal API
emoji: 🏢
colorFrom: indigo
colorTo: blue
sdk: docker
app_port: 7860
pinned: false

Internal Portal API (backend)

FastAPI backend for the internal portal chat app. Provides domain/project management, a chat endpoint backed by DeepSeek, and an HR assistant grounded in uploaded documents.

Deploy on Hugging Face Spaces

1. Create a new Space → Docker SDK.
2. Push the contents of this backend/ folder to the Space repo (this README's YAML header configures the Space; app_port: 7860 matches the Dockerfile).
3. In Settings → Variables and secrets, add:
   • DEEPSEEK_API_KEY (secret) — your DeepSeek key
   • DEEPSEEK_MODEL — your exact model id (e.g. the "v4 flash" model string)
   • DEEPSEEK_BASE_URL — optional, defaults to https://api.deepseek.com
   • ALLOWED_ORIGINS — your Vercel URL, e.g. https://your-app.vercel.app (defaults to * for easy first testing)
4. The Space builds and serves at https://<user>-<space>.hf.space.

Check it's up: GET https://<space-url>/api/health → {"ok": true, "llm_configured": true, "model": "..."}.

Run locally

cd webapp/backend
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
export DEEPSEEK_API_KEY=sk-...           # your key
export DEEPSEEK_MODEL=deepseek-chat       # or your "v4 flash" model id
uvicorn app.main:app --reload --port 7860

Then open http://localhost:7860 — the backend also serves the static frontend from ../frontend for convenience, so you can test the whole app from one process locally.

Storage note

Data is persisted to a JSON file under DATA_DIR (/data in Docker). A free HF Space's disk is ephemeral — it resets on rebuild/restart. That's fine for development and demos. When durable storage is needed, the store.py layer is swappable for a real database without changing the rest of the app.