🧠 RAG-based Chatbot with User & API Key Management

This is a FastAPI-based backend that implements a Retrieval-Augmented Generation (RAG) chatbot using Azure AI Search as the vector database. It includes user authentication, API key management, chat history tracking, and MSSQL visualization through SQLPAD.

---

🚀 Features

🔍 RAG Chatbot (Azure AI Search)

• Semantic search using Azure AI Search
• Maintains chat history per user and session (Chat ID)
• Like / Dislike feedback for chat responses
• Favourite chat
• Rename chat titles
• Disable chats
• Retrieve:
  • All user chats
  • Chats by session ID

👤 User Management

• Sign up / Register
• Login to get JWT access and refresh tokens

🔑 API Key Management

• Create API key per registered user
• Rate limit enforced per API key associated to type of tier
• Delete API key
• Validate API key on API requests

---

🐳 Docker Setup

🧱 Build and Run Containers

# Build images and start containers in background
sudo docker compose up --build -d

# Stop and remove containers
sudo docker compose down

URL

• FastAPI http://localhost:8000/docs
• SQLPAD UI http://localhost:3000

Adminer Database Configuration

To log into SQLPAD at http://localhost:3000, first use admin as user and password for SQLPAD dashboard then connect to your mssql-server container to fill in the connection form as follows:

Field	Value
System	Microsoft SQL Server
Server	mssql-server (container name)
Username	sa
Password	yourStrong)_1_(Password
Database	master

Enviroment Variable (.env)

Make sure to provide a .env file in your roject root.

Redis_URL="redis://redis-server:6379/0"
DATABASE_URL="mssql+pymssql://sa:yourStrong%29_1_%28Password@mssql:1433/master"
OPENAI_API_KEY=your_openai_api_key
LANGSMITH_KEY=your_langsmith_key
OPENAI_API_VERSION="2023-05-15"
EMBEDD_MODEL=text-embedding-ada-002
GPT_MODEL="gpt-4o-mini-2024-07-18"
VECTOR_STORE_ADDRESS=Azure Ai search index url
VECTOR_STORE_KEY=your_azure_ai_search_key
JWT_SECRET_KEY=super-secret-add any thing 
ACCESS_TOKEN_EXPIRE_MINUTES=5
REFRESH_TOKEN_EXPIRE_DAYS=7

Project Structure

.
├── ai
│   ├── gen_models.py
│   ├── __init__.py
│   └── retriever.py
├── alembic
│   ├── env.py
│   ├── README
│   ├── script.py.mako
│   └── versions
│       ├── 60547dc1cfdb_fix_relationship_mismatch.py
│       └── 913ec573b6a4_initial_migration.py
├── alembic.ini
├── api
│   ├── api_key.py
│   ├── chat.py
│   ├── __init__.py
│   └── user.py
├── core
│   ├── api_key_auth.py
│   ├── config.py
│   ├── __init__.py
│   └── jwt_utils.py
├── db
│   ├── chat_memory.py
│   └── db_connection.py
├── db_models
│   ├── __init__.py
│   └── models.py
├── docker-compose.yml
├── Dockerfile
├── main.py
├── mssql_data  [error opening dir]
├── pyproject.toml
├── README.md
├── schemas
│   ├── chat_schemas.py
│   ├── __init__.py
│   └── user_schema.py
├── services
│   ├── chat_service.py
│   └── __init__.py
├── tests
│   ├── db test.py
│   └── __init__.py
├── uv.lock
└── wait-for-it.sh

🛠️ Tech Stack

• Python – Core programming language
• FastAPI – Backend web framework
• Uvicorn – ASGI server for running FastAPI
• SQLAlchemy – ORM for database models and interactions
• Alembic – Database migrations for SQLAlchemy models
• LangChain – Framework for building LLM-based applications
• LangChain OpenAI – Integration with OpenAI APIs for embeddings and completions
• LangChain Unstructured – For parsing and chunking unstructured documents (e.g. PDFs, Word)
• LangChain Expression Language (LCEL) – Used to build modular RAG pipelines
• LangSmith – Tracing, observability, and debugging for LangChain apps
• SQLChatMessageHistory – Used to persist chat history in MSSQL
• Azure AI Search – Vector database for semantic document search
• AzureSearch (LangChain) – LangChain wrapper for index creation and search in Azure AI Search
• Redis (Docker) – In-memory store for caching and temporary storage
• MSSQL (Docker) – SQL Server for chat/message storage
• Docker – Containerization of all services
• Docker Compose – Orchestration of multi-container application
• OpenAI API Key – Access to OpenAI’s GPT and embedding models
• python-jose – JWT authentication and token verification
• uv – Lightweight Python dependency and virtual environment manager