🚀 Multi-Provider Enterprise RAG System

A comprehensive reference implementation for AI Engineers learning to build production-grade RAG systems with multi-provider LLM support, enterprise RBAC, and a modern React UI.

---

📦 Repository Structure

Module	Description	Readme
ai_backend/	FastAPI backend — RAG, Agents, Auth, RBAC, Conversations	ai_backend/README.md
front_end_react/	React + Vite frontend — Chat UI, Document Management, Admin	front_end_react/README.md
mcp/	Standalone MCP Tools Server — Weather, Stock, Web, Travel	mcp/README.md
lora_traning/	LoRA fine-tuning pipeline for Llama 3.2 1B	lora_traning/USAGE_GUIDE.md
vector_databases/	Standalone demos — ChromaDB, FAISS, LanceDB, SQLite	—

---

✨ Highlights

• 🤖 Multi-Provider RAG — OpenAI GPT, Google Gemini, Hugging Face, local GGUF, LlamaServer, CustomLLM
• 🔒 Enterprise RBAC — Role hierarchy (Guest → SuperAdmin), sensitivity levels, department restrictions
• 💬 Persistent Conversations — SQLite-backed history with session isolation
• 🤖 4 Agent Orchestrators — AutoGen, Custom, MCP, CrewAI via unified /api/agents/query
• 🔍 Hybrid Retrieval — BM25 + vector search with cross-encoder reranking
• 🎙️ Multimodal — Speech-to-Text, Text-to-Speech, OCR, emotion detection
• 🔌 MCP Server — Standalone tool server compatible with Claude Desktop and any MCP client

---

🚀 Getting Started

Option 1 — Docker (recommended)

Runs the backend and frontend together with a single command.

git clone https://github.com/developer-shubham101/ai_engineer.git
cd ai_engineer
docker compose up --build

Service	URL
Backend API	http://localhost:8000
Frontend UI	http://localhost:3000
Swagger Docs	http://localhost:8000/docs

To point the frontend at a custom backend URL:

VITE_API_URL=https://api.yourdomain.com docker compose up --build

---

Option 2 — Local Setup

git clone https://github.com/developer-shubham101/ai_engineer.git
cd ai_engineer

Backend

See ai_backend/README.md for full configuration, provider setup, agent workflows, and API reference.

cd ai_backend
pip install -r requirements.txt
cp .env.example .env          # add your API keys
python scripts/download_embeddings_models.py
uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

Backend available at http://localhost:8000 Interactive API docs at http://localhost:8000/docs

---

Frontend

See front_end_react/README.md for component structure and UI patterns.

cd front_end_react
npm install
npm run dev

Frontend available at http://localhost:3000

The API URL defaults to http://localhost:8000. To change it, set VITE_API_URL in a .env file inside front_end_react/.

---

MCP Tools Server

See mcp/README.md for tool list, Claude Desktop integration, and inspector usage.

cd mcp
pip install -r requirements.txt
cp .env.example .env          # optional: add OPENWEATHER_API_KEY, SERPAPI_KEY
python server.py              # stdio transport

Or run the interactive client REPL:

python client.py

---

🔧 Environment Variables

Copy .env.example to .env inside ai_backend/ and fill in the keys you need:

cp ai_backend/.env.example ai_backend/.env

Variable	Purpose
OPENAI_API_KEY	OpenAI GPT provider
GOOGLE_API_KEY	Google Gemini provider
HUGGINGFACE_API_TOKEN	Hugging Face Inference API
CUSTOMLLM_BASE_URL	Any OpenAI-compatible 3rd-party LLM
JWT_SECRET_KEY	Auth token signing — change in production
VECTOR_STORE_TYPE	faiss (default) or chroma

Full variable reference → ai_backend/README.md

---

📖 API Documentation

• Interactive (Swagger UI) → http://localhost:8000/docs
• Full reference → ai_backend/documents/API_DOCUMENTATION.md

---

🔒 RBAC & Security

Role hierarchy: Guest (0) < Employee (1) < HR (2) < Manager (3) < SuperAdmin (4)

Sensitivity levels: public_internal → department_confidential → role_confidential → highly_confidential → super_confidential

Full RBAC docs → ai_backend/README.md

---

🤝 Contributing

Contributions are welcome! Fork the repo, create a feature branch, and open a pull request.

• 🐛 Report a Bug
• 💡 Request a Feature

---

🧑💻 Author

Shubham Sharma GitHub • Repository