LangGraph Mastery: Stateful LLM Workflows

A comprehensive laboratory for building resilient, stateful, and cyclic LLM applications using LangGraph. Master the transition from linear chains to complex agentic graphs.

---

Project Overview

This repository is a structured learning journey into LangGraph, the framework for building stateful, multi-actor applications with LLMs. It covers the complete spectrum of graph-based orchestration, starting from basic mathematical logic to advanced persistent chatbots.

Strategic Learning Path

1. Foundations & Control Flow

• 0_test_installation: Verification of the graph execution environment.
• 1_bmi_workflow: Introduction to StateGraph, TypedDict, and basic node-to-node transitions.
• 2_simple_llm_workflow: Integrating LLMs as functional nodes within a graph.
• 6_quadratic_equation: Handling complex mathematical reasoning and multi-step logic.

2. Chain of Thought & Pattern Design

• 3_prompt_chaining: Implementing sequential prompt logic for refined outputs.
• 4_batsman_workflow: Dynamic routing and domain-specific logic implementation.
• 7_review_reply_workflow: Automated customer sentiment analysis and response generation.

3. Iterative Generation & Advanced State

• 5_CSS_essay_workflow: Multi-stage content generation with refinement loops.
• 8_X_post_generation: Automating social media content creation using agentic workflows.
• 10_persistence: Mastering Checkpointers for state recovery, human-in-the-loop, and time-travel debugging.

4. Conversational Agents

• 9_basic_chatbot: Building a standard conversational interface using stateful memory.
• langgraph-chatbot/: A full-scale implementation of a persistent, tool-enabled chatbot.

---

Key Concepts Demonstrated

• State Management: Using TypedDict to maintain and update application state across nodes.
• Nodes & Edges: Defining functional units and the logic that connects them (including START and END).
• Conditional Routing: Implementing branches based on the graph's current state.
• Persistence: Using MemorySaver to persist graph state across sessions.
• Visualization: Generating Mermaid diagrams to visualize complex logic flows.

---

Setup & Execution

1. Environment Setup

# Create and activate virtual environment
python -m venv myenv
source myenv/Scripts/activate  # Windows

2. Install Dependencies

pip install langgraph langchain_openai langchain_groq ipykernel

3. Run Notebooks

Launch Jupyter and explore the modules in order:

jupyter notebook

---

Why LangGraph?

While standard LangChain is excellent for linear DAGs, LangGraph enables:

1. Cycles: Iterative refinement where an LLM can check its own work.
2. Persistence: Seamlessly saving and resuming long-running agent workflows.
3. Fine-Grained Control: Explicitly defining exactly how data flows between actors.

---