LangChain-Mini

Welcome to Mini - a minimalistic toolkit for building sophisticated, Graph-based AI agents.

This project serves an educational purpose: to demystify the current hype around AI agents by showing how to prompt LLMs to use tools, generate structured outputs, and connect to real-world applications from scratch. Most architectural decisions for this project were inspired by LangChain and LangGraph.

You can find the source code is in the framework directory. For a complete overview of the Framework's components and a simple CoT agent (with tools), check out framework_explained.ipynb.

Key Features

• Tool Integration 🛠️: Turn any Python function into a LLM compatible tool
• Structured Outputs: Get structured outputs from LLMs for easy parsing
• Prompt Templates: Systematic prompt generation for consistent agent behavior
• Agents as State Machines: Create complex graph-based agents with ease
• Tracing and Logging: A unified module interface for easy debugging

Quick Start 🚀

1. Installation

   git clone https://github.com/TilliFe/LangChain-Mini.git
   cd LangChain-Mini
   pip install -r requirements.txt

2. Configuration

   Create .env file:

   GEMINI_API_KEY="YOUR_KEY"   # from https://aistudio.google.com/apikey
   TAVILY_API_KEY="YOUR_KEY"   # from https://app.tavily.com/home

3. Run Examples

   python run_examples.py

   You can now interact with the tool-using AI Agent directly in your console, as shown in the screenshot below. The default LLM used for this Agent is Gemini 2.0 Flash. You can easily switch the model and customize the Agent's behavior in the /examples/CoT_chat.py file.

   

Happy Building! 🔥