emarco177/langchain-course

What This Is

This repo is a course that teaches you how to build AI-powered applications using the LangChain and LangGraph libraries (Python, version 1.0+ of LangChain per the README). The course material is structured around building real projects, not toy examples — you work with live APIs and actual use cases.

What You'll Build

The README lists 10 projects (7 mentioned in one section, plus a few extra in other sections). Some live in this repo as git branches (like project/hello-world), and some live in separate external repositories. A sample:

• Hello World — your first basic AI agent.
• Modern Search Agent — an agent that uses search tools and custom tools, with structured outputs.
• RAG Gist — the basics of retrieval-augmented generation (feeding an AI relevant documents before answering).
• System One Models — fast decision-making with model routing and tool gating.
• Code Interpreter — an AI agent that can execute code.
• Reflection Agent / Reflexion Agent — agents that self-critique and self-correct.
• Agentic RAG — a more advanced retrieval-augmented system where the agent plans and has memory.

There are also references to an Ice Breaker (social media profile analyzer), a Blog Analyzer (content analysis), and a Documentation Helper.

Learning Path

The README lays out three phases:

1. Foundations — Hello World chain, then Code Interpreter.
2. Real-World Applications — Ice Breaker, Documentation Helper.
3. Advanced — Blog Analyzer, Agentic RAG.

How You Use It

• Prerequisites: Not for beginners — you should already be comfortable with git, Python, environment variables, classes, testing, and debugging. Python 3.10+ is required. You need access to an LLM (either a local one like Ollama or a cloud one like OpenAI, Anthropic, Gemini). No machine-learning experience is needed.
• Setup: Clone the repo, then check out a specific project branch and run it (e.g. git checkout project/hello-world, then uv sync and uv run python main.py). For external projects, you clone those separately.
• Learning method: Each git commit represents a lesson. The README suggests you check the commit history (git log --oneline) and step through commits to understand how the code evolves.

What You'll Learn

By the end: build agents from scratch, call external APIs, build RAG systems with vector databases, design multi-step reasoning workflows, deploy agents, handle self-correction, and optimize cost/performance.

Key Details

• Tech stack mentioned: LangChain v0.3+, LangGraph, Pinecone, FAISS, Streamlit.
• Package manager: The README says to use uv, poetry, or pipenv — explicitly NOT conda.
• License: Apache 2.0.
• Author: Eden Marco. The course also exists on Udemy (with a coupon link mentioned in the README).

What This Is NOT

The README doesn't describe this repo as a finished software product — it's a learning resource (course content, branches, external linked repos). It also doesn't claim any features beyond what's listed above.