langgraph-ollama-local 0.1.0

Local agent building at scale using Ollama - comprehensive LangGraph patterns for local LLMs

LangGraph Ollama Local

Learn LangGraph by building agents that run entirely on your hardware.

---

What This Is

A hands-on tutorial series for building LangGraph agents with local LLMs via Ollama. Each notebook teaches a concept from scratch - no cloud APIs required.

You'll learn:

• LangGraph fundamentals: StateGraph, nodes, edges, reducers
• Tool calling and the ReAct pattern
• Memory and conversation persistence
• Human-in-the-loop workflows
• Self-reflection and critique patterns
• RAG patterns: Basic RAG, Self-RAG, CRAG, Adaptive RAG, Agentic RAG
• Build a Perplexity-style research assistant

Quick Start

Prerequisites

1. Python 3.12+
2. Ollama running locally or on your LAN:

   # Install Ollama
   curl -fsSL https://ollama.ai/install.sh | sh
   
   # Pull a model
   ollama pull llama3.2:3b
   

Installation

git clone https://github.com/AbhinaavRamesh/langgraph-ollama-tutorial.git
cd langgraph-ollama-tutorial

# Install with RAG dependencies
pip install -e ".[all]"

# Verify connection
langgraph-local check

Configuration

cp .env.example .env
# Edit .env with your settings

Variable	Default	Description
OLLAMA_HOST	127.0.0.1	Ollama server address
OLLAMA_PORT	11434	Ollama server port
OLLAMA_MODEL	llama3.2:3b	Default model
TAVILY_API_KEY	(optional)	For web search in CRAG tutorials

Web Search Setup (Optional)

For CRAG and Perplexity-style tutorials, get a free Tavily API key:

1. Sign up at https://tavily.com
2. Get your API key from the dashboard
3. Add to .env:

   TAVILY_API_KEY=tvly-your-key-here
   

LAN Server with Monitoring

To host Ollama on a GPU machine accessible across your network, use ollama-local-serve:

pip install ollama-local-serve[all]
make init && make up
# Dashboard at http://your-server:3000

Tutorials

Core Patterns (Phase 2)

#	Notebook	Documentation	What You'll Learn
01	Chatbot Basics	docs	StateGraph, nodes, edges, message handling
02	Tool Calling	docs	Tools, ReAct loop from scratch
03	Memory & Persistence	docs	Checkpointers, threads, conversation history
04	Human-in-the-Loop	docs	Interrupts, approvals, resume
05	Reflection	docs	Generate, Critique, Revise loops
06	Plan & Execute	docs	Structured outputs, multi-step planning
07	Research Assistant	docs	Capstone: all patterns combined

RAG Patterns (Phase 3)

#	Notebook	Documentation	What You'll Learn
08	Basic RAG	docs	Document loading, chunking, embeddings, ChromaDB
09	Self-RAG	docs	Document grading, hallucination detection, retry loops
10	CRAG	docs	Web search fallback, corrective retrieval
11	Adaptive RAG	docs	Query classification, strategy routing
12	Agentic RAG	docs	Agent-controlled retrieval, multi-step
13	Perplexity Clone	docs	Citations, source metadata, follow-ups

Run any notebook:

jupyter lab examples/

RAG Quick Start

Index your documents and start querying:

from langgraph_ollama_local.rag import DocumentIndexer, LocalRetriever

# Index documents
indexer = DocumentIndexer()
indexer.index_directory("sources/")

# Query
retriever = LocalRetriever()
docs = retriever.retrieve_documents("What is Self-RAG?", k=4)

Project Structure

langgraph-ollama-tutorial/
├── examples/
│   ├── core_patterns/          # Tutorials 01-07
│   │   ├── 01_chatbot_basics.ipynb
│   │   └── ...
│   └── rag_patterns/           # Tutorials 08-13
│       ├── 08_basic_rag.ipynb
│       └── ...
├── docs/
│   ├── core_patterns/          # Core pattern documentation
│   └── rag_patterns/           # RAG pattern documentation
├── sources/                    # PDF sources for RAG indexing
├── langgraph_ollama_local/
│   ├── config.py               # Configuration
│   ├── cli.py                  # CLI tools
│   └── rag/                    # RAG infrastructure
│       ├── embeddings.py       # Local embeddings
│       ├── indexer.py          # Document indexing
│       ├── retriever.py        # Document retrieval
│       └── graders.py          # Quality graders
├── tests/                      # Test suite
└── pyproject.toml

CLI

langgraph-local check    # Verify Ollama connection
langgraph-local config   # Show current configuration
langgraph-local list     # List available examples

Development

make test          # Run tests
make test-int      # Integration tests (requires Ollama)
make lint          # Linting
make format        # Format code

Adapted From

These tutorials are adapted from the official LangGraph documentation (MIT License), optimized for local Ollama deployment.

License

MIT License