🎓 Simple RAG: FastAPI + LangGraph + LangChain

This repository serves as a didactic reference for implementing a production-ready RAG (Retrieval-Augmented Generation) agent. It demonstrates how to orchestrate complex LLM logic using LangGraph and serve it as a high-performance API via FastAPI.

---

🏛️ Architecture & Workflow

The project is designed to show students the "Life of a Query":

1. API Layer (FastAPI): Receives the user request and manages the asynchronous execution.
2. Orchestration (LangGraph): Manages the state and logic flow (e.g., should I retrieve documents?, is the context sufficient?).
3. RAG Logic (LangChain): Handles document loading, splitting, embedding, and vector search.

Key Educational Concepts Included:

• State Management: Using LangGraph to maintain context across the agentic workflow.
• Vector Embeddings: How documents are transformed and stored for semantic search.
• RESTful AI: Serving model outputs through standard HTTP endpoints.

---

🛠️ Tech Stack

Component	Technology	Role
Framework	LangChain	RAG primitives and LLM integration.
Orchestration	LangGraph	Cyclic/Acyclic graph-based agent logic.
API	FastAPI	Asynchronous web server and OpenAPI documentation.
Vector Store	Chroma	Local vector database for document retrieval.

---

🚀 Getting Started

1. Environment Setup

Install UV

curl -LsSf https://astral.sh/uv/install.sh | sh

Clone the repo and install dependencies:

git clone https://github.com/rikDei/simple_rag.git
cd simple_rag
uv sync

2. Configuration

Create a .env file with your credentials:

NVIDIA_API_KEY=****
TAVILY_API_KEY=****

3. Running the API

Loading the document into the VectorDB

python -m embed_documents.py

4. Running the API

Start the FastAPI server:

fastapi run app:app

Once running, you can access the interactive Swagger UI at: http://127.0.0.1:8000/docs.

---

📂 Project Structure

├── app/
│   ├── __init__.py          # FastAPI entry point
│   ├── router.py            # Router to handle HTTP Requests
│   └── validators.py        # Pydantic BaseModel to validate user input
├── agent/                   # All the agent logic
│   ├── __init__.py          # Initialize model instance
│   └── agent.py             # Langgraph agent
├── embeddings/              # Embedding model
│   └── __init__.py          # Initialize instance
├── tools/                   # Custom tools
│   ├── __init__.py          # Lists tools
│   ├── vector_store.py      # Tools to interact with the vector store
│   └── web_search.py        # Tavily built in tool
├── embed_documents.py       # Python script to load the documents inside the vector DB
└── config.py                # Gets all the environment variables

---

🎓 For Students: Things to Explore

• Modify the Prompt: Look at prompts/chat_bot.md and see how changing the System Message affects the agent's behavior.
• Add a Node: Try adding a "Tools" node in tools module to allow the agent more capabilities.

---

Created for educational purposes by rikDei