# HF_Space_API (FastAPI + Docker)

This Space exposes a **REST API** for document processing (summarize, embed, search, process PDFs).
It is designed to **wrap your existing project** with minimal changes.

## How to use

1. Create a new **Hugging Face Space** → **Docker** template.
2. Push this folder to that Space (or upload as a ZIP and use the Space UI).
3. Set Space **Hardware** (e.g., T4 or A10G for GPU) and add **Space secrets** if needed:
   - `HF_TOKEN` (optional, for private models)
   - `API_SECRET` (optional, simple bearer auth)

4. Wire your own modules inside `app.py` (see TODO markers).

## Endpoints

- `GET /health`
- `POST /embed` — JSON: `{ "texts": ["..."] }`
- `POST /summarize` — JSON: `{ "text": "..." }`
- `POST /search` — JSON: `{ "query": "...", "k": 5 }` (dummy store until you wire your FAISS index)
- `POST /process_pdf` — multipart/form-data: `file=@doc.pdf`

## Notes

- Default uses `faiss-cpu`. On GPU Space you may switch to `faiss-gpu` in `requirements.txt`.
- Avoid committing local cached models into git. Publish models to the Hub and **download at startup**.