FastGraphAI

A Python LLM chat project comparing LangChain model calls with LangGraph workflow orchestration, exposed through FastAPI and a Streamlit interface.

What it demonstrates

• Typed request and response schemas with Pydantic.
• Separate API routes for LangChain and LangGraph.
• OpenAI-compatible model integration using Together.ai in the checked-in backend.
• A Streamlit interface with engine selection and session conversation history.
• Separation of API and interface code across branches.

The current graph is a single LLM node. This repository is a learning and portfolio project; the checked-in implementation does not establish a production-ready multi-agent system.

Find the implementation

Branch	Contents
ChatBot_V3_Backend	FastAPI service, LangChain call, LangGraph workflow and dependency list
ChatBot_V3_Frontend	Streamlit chat interface
main	Initial API implementation and this overview
LangGraph-Workflow	Separate workflow experiments

Run the backend

git clone --branch ChatBot_V3_Backend https://github.com/ShardulAswale/FastGraphAI.git fastgraphai-backend
cd fastgraphai-backend
python -m venv .venv
# Linux/macOS: source .venv/bin/activate
# Windows PowerShell: .venv\Scripts\Activate.ps1
python -m pip install -r requirements.txt

Set TOGETHER_API_KEY in your local environment or a local .env file, then run:

python -m uvicorn app:app --reload --port 8000

Explore the API at http://localhost:8000/docs.

Route	Method	Request
/chat/chain	POST	{"message": "Explain retrieval-augmented generation"}
/chat/graph	POST	Same schema

Both routes return {"response": "..."}.

Run the interface

In a separate folder and terminal:

git clone --branch ChatBot_V3_Frontend https://github.com/ShardulAswale/FastGraphAI.git fastgraphai-frontend
cd fastgraphai-frontend
python -m venv .venv
# Activate the virtual environment as above.
python -m pip install -r requirements.txt requests

Set BACKEND_URL=http://localhost:8000 in a local .env file, then run:

python -m streamlit run app.py

The frontend makes Python HTTP requests to the backend. Its default URL should be overridden for your own environment.

Project scope

This implementation demonstrates LLM API integration and workflow structure. Authentication, rate limiting, persistent chat history, automated evaluation and comprehensive deployment hardening are future work. Model calls may incur provider charges. Keep real API keys out of commits.