🦜 LangChain Course: From Zero to Agent Hero

A collection of projects covering the LangChain ecosystem, from basic LLM integration to advanced agentic patterns using LangGraph.

📁 Repository Structure

Each directory in this repository represents a specific module or project in the LangChain learning path.

Module	Project Name	Description	Key Technologies
01	Hello World	Basic setup, instructions following, and first LLM call.	OpenAI, Few-Shot
02	Search Agent	Building agents that autonomously use web search tools.	Tavily, Tool-use
03	Agents Under the Hood	Deep dive into agent loops, ReAct prompts, and regex parsing.	Ollama, CoT, ReAct
04	RAG Gist	Foundational Retrieval Augmented Generation with vector stores.	Pinecone, LCEL
05	Doc Helper	A full-stack RAG application for intelligent documentation search.	Streamlit, Tavily
06	ReAct Agent	ReAct pattern implemented using LangGraph and ToolNode.	LangGraph, ToolNode
07	Reflection Agent	Self-correcting agents that iterate and critique their output.	Self-Correction
08	Reflexion Agent	Advanced Actor-Critic model with tool-assisted self-reflection.	Actor-Critic
09	Agentic RAG	The pinnacle: CRAG and Self-RAG for self-correcting retrieval.	CRAG, Self-RAG
50	MCP Search Agent	Decoupled search agent using Model Context Protocol (MCP).	MCP, FastMCP, Stdio

📚 Research Library

This repository links practical implementations to foundational AI research. Each module contains a docs/ folder with the relevant seminal papers:

• Prompting: Few-Shot Learning (GPT-3), Chain-of-Thought.
• Agent Architecture: ReAct, Toolformer, Reflexion, MCP Agent.
• RAG Evolution: Original RAG Paper, Corrective RAG (CRAG), Self-RAG.

🚀 Getting Started

Prerequisites

• Python 3.11+
• uv (highly recommended for lightning-fast dependency management)
• API Keys for OpenAI, Tavily, and Pinecone.

Installation & Execution

Each module is an independent project. Navigate to a directory and use uv to sync and run:

cd 09-agentic-rag
uv sync
python main.py

🧠 Learning Path

1. Foundations: Start with 01 and 04 to master basic LLM calls and the core RAG pattern.
2. Mechanics: Move to 02 and 03 to understand how agents use tools at both a high and low level.
3. The LangGraph Shift: Explore 06 and 07 to transition from linear chains to stateful, cyclic graphs.
4. Advanced Reasoning: Master 08 (Actor-Critic) and 09 (Agentic RAG) to build production-grade self-correcting systems.

🛠️ Tools Used

• Framework: LangChain 🦜🔗
• Orchestration: LangGraph 🕸️
• Models: OpenAI (GPT-4o-mini, o4-mini), Ollama
• Search: Tavily AI 🔍
• Vector DB: Pinecone 🌲
• UI: Streamlit 🖥️

---

Maintained by rasivasu

🙏 Acknowledgments & Credits

This repository is a collection of projects and tutorials from the excellent Udemy course: LangChain- Agentic AI Engineering with LangChain & LangGraph by Eden Marco.

All the implementations and architectural patterns found here are based on the course's curriculum. I highly recommend the course to anyone looking to master Agentic AI.