README.md

metadata

title: RAG SaaS Platform
emoji: 🧠
colorFrom: blue
colorTo: indigo
sdk: docker
python_version: '3.11'
pinned: false

RAG Systems as a Service (SAAS)

Enterprise-grade, multi-tenant RAG platform designed for Hugging Face Spaces. This project consists of two main components:

1. rag_saas Backend: A FastAPI-based REST API for managing multi-tenant retrieval and generation.
2. webapp Playground: A Streamlit interface for interactive testing, document ingestion, and "How it works" diagrams.

Quick Start on HF Spaces

Since this is a decoupled system, we recommend deploying as two separate Spaces for optimal performance.

1. Backend (API)

• Space Type: Docker Space.
• Root Files: Use the root Dockerfile and app/ folder.
• Secrets: Add OPENAI_API_KEY or HUGGINGFACE_API_TOKEN.

2. Frontend (Playground)

• Space Type: Docker Space (using webapp/Dockerfile).
• Files: Only upload the contents of the webapp/ folder.
• Link: Paste your Backend Space URL into the "API Connection" section of the playground sidebar.

Key Features

• Multi-Tenant Isolation: Secure data segregation using X-Tenant-ID headers.
• Playground UI: Modern chat interface, ingestion dashboard, and educational onboarding.
• Enterprise-ready: DB-backed tenant management and storage (SQLite/Postgres).
• HF Optimized: Docker images configured with user 1000 and port 7860.

API Documentation

Once the backend is deployed, visit /docs for the OpenAPI Swagger UI.

Core Architecture

graph LR
    User --> Streamlit[Web App Playground]
    Streamlit --> API[FastAPI Backend]
    API --> DB[(SQLite/Postgres)]
    API --> VectorDB[(Vector Store)]
    API --> LLM[OpenAI/HuggingFace]

Local Development

# 1. Start Backend
uvicorn app.main:app --reload --port 7860

# 2. Start Web App
cd webapp
streamlit run app.py