---
title: Smart Document QA
emoji: 📄
colorFrom: blue
colorTo: indigo
sdk: docker
app_port: 7860
pinned: false
---

# Smart Document QA

FastAPI-based document Q&A with asynchronous processing, grounded answers, and a simple demo UI.

Live endpoints:

- Demo: [https://agni512-hirehub-smart-document-qa.hf.space/demo](https://agni512-hirehub-smart-document-qa.hf.space/demo)
- API docs: [https://agni512-hirehub-smart-document-qa.hf.space/docs](https://agni512-hirehub-smart-document-qa.hf.space/docs)
- Health: [https://agni512-hirehub-smart-document-qa.hf.space/health](https://agni512-hirehub-smart-document-qa.hf.space/health)

This Space runs the existing app in a single Docker container and starts:

- PostgreSQL
- Redis
- Celery worker
- FastAPI app

The public entrypoints are:

- `/demo`
- `/docs`
- `/health`

Required Space secrets:

- `HF_TOKEN`
- `OPENAI_API_KEY`

Recommended Space variables:

- `OPENAI_BASE_URL=https://router.huggingface.co/v1`
- `OPENAI_MODEL=openai/gpt-oss-20b`
- `HF_INFERENCE_PROVIDER=auto`
- `HF_EMBEDDING_MODEL=Qwen/Qwen3-Embedding-8B`
- `EMBEDDING_MODEL=infgrad/Jasper-Token-Compression-600M`
- `EMBEDDING_TRUST_REMOTE_CODE=true`

`HF_EMBEDDING_MODEL` is the primary API-backed embedder. `EMBEDDING_MODEL` remains the local fallback model if the HF embedding API is unavailable.

The generated Space bundle is created under [`build/hf-space`](./build/hf-space) by [`scripts/build_hf_space_bundle.py`](./scripts/build_hf_space_bundle.py).