CompassIA / README.md
dembasowmr's picture
Reorganized the project: -Documents hosted on Firestore db, -conversations saved
15d9931
---
title: CompassIA
emoji: πŸ’¬πŸ“š # You can choose other emojis here
colorFrom: green
colorTo: blue
sdk: docker # This must be 'docker' as you selected
app_port: 7860 # This is the port your uvicorn app is listening on
# Optional: You can specify a base image for the Docker SDK if not in Dockerfile directly.
# base_image: python:3.10-slim-buster
# pinned: false # Usually not needed unless you want to pin a specific commit
---
# CompassIA
This Space hosts the backend API for CompassIA, a PDF Question Answering system. CompassIA is a smart agent for MaarifCompass that allows users to ask questions about PDF documents, leveraging advanced AI models for natural language understanding and information retrieval.
It uses:
- **Python FastAPI** for the web API.
- **DeepSeek Chat** (via OpenRouter) as the Large Language Model.
- **BGE-M3** from FlagEmbedding for local, multilingual text embeddings.
- **ChromaDB** for persistent vector storage.
- **pdfminer.six** and **PyTesseract (with Poppler)** for robust PDF text extraction and OCR.
## How to Use:
This Space exposes a `/compassia/` API endpoint. You can interact with it using `curl`, Postman, Insomnia, or by integrating it with your Next.js frontend.
### API Endpoint: `/compassia/` (POST request)
**Request Body (JSON):**
```json
{
"question": "Your question about the PDF documents"
}