Repository Overview

🤖 AI Agents with LangChain & LangGraph

This repository contains multiple AI agent implementations leveraging the LangChain and LangGraph frameworks to demonstrate advanced LLM‑based workflows.

Table of Contents

• Technologies
• Project Structure
• Getting Started
• Usage
• Contributing
• License
• Contact

🌟Technologies

• LangChain: a framework for developing applications powered by LLMs, simplifying every stage of the LLM application lifecycle.
• LangGraph: a low‑level orchestration framework for building controllable, stateful agent workflows with multi‑agent support.
• FastAPI: for building high‑performance asynchronous APIs.
• Streamlit: for creating interactive web UIs for experimentation and demos.
• ChatGoogleGenerativeAI (Gemini 2.0 Flash): Google’s generative AI model integrated via LangChain.
• Ollama LLM (llama3.2:1b): local Llama 3 model powered by Ollama.
• Pinecone: vector database for retrieval‑augmented generation (RAG) pipelines.
• GroQ: an end‑to‑end AI pipeline tool integrated in LangChain workflows.
• Hugging Face: model hub integration for custom LLMs and embeddings.
• dotenv: for managing environment variables.
• LangServe: deploys LangChain pipelines as production‑ready APIs.

📁Project Structure

• chain_yt/
  Jupyter notebooks demonstrating LangChain‑based agents for YouTube video processing, covering Q&A, retrieval‑augmented generation (RAG), custom agent composition, function calling, GroQ pipeline integration, Hugging Face model usage, memory‑based retrieval, and Pinecone vector store indexing capabilities.

• graph_yt/
  LangGraph workflows for structured YouTube data processing, including single‑agent and multi‑agent graph definitions, RAG graph pipelines, exam simulation flows, and custom bot implementations under the Bots/ directory.

• project1/
  A full‑stack AI chatbot application:

  • api/

    • app.py: FastAPI app exposing /gemini, /essay, and /poem routes via add_routes, wiring LangChain prompt templates to ChatGoogleGenerativeAI and OllamaLLM backends.
    • client.py: Streamlit client for interacting with the /gemini endpoint.

  • chatbot/

    • app.py: Gemini‑powered Streamlit chatbot using LangChain prompt templates and output parsers.
    • localama.py: Local Ollama Llama 3 Streamlit chatbot showcasing an entirely on‑device pipeline.

The repository is organized to facilitate exploration and extension, with clear module boundaries and modular examples.

🚀Getting Started

1. Clone the repository

   git clone https://github.com/devagn611/AI-Agents.git
   cd AI-Agents

2. Create a virtual environment

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies

   pip install -r requirements.txt

4. Set up environment variables
   Copy .env.example to .env and fill in your API keys for Google, LangChain, LangSmith, etc.

📚Usage

Running the Jupyter Notebooks

Navigate to any folder under chain_yt/ or graph_yt/ and launch:

jupyter notebook

Launching the API

cd project1/api
uvicorn app:app --reload

• Access /gemini, /essay, and /poem endpoints at http://localhost:8000/.

Starting the Streamlit UIs

cd project1/api
streamlit run client.py

cd project1/chatbot
streamlit run app.py
# or
streamlit run localama.py

Contributing

Contributions are welcome! Please fork the repository, create a feature branch, and open a pull request. Ensure all new code is covered by appropriate documentation or tests.

📜License

This project is licensed under the MIT License. See the LICENSE file for details.

📧Contact

For questions or feedback, please open an issue or reach out via email:
devagnmaniya611@gmail.com