contextinator 2.0.2

Filesystem tools for AI agents with optional RAG capabilities

Turn any codebase into semantically-aware, searchable knowledge for AI-powered workflows.

Key Features

• AST-Powered Chunking - Extract functions, classes, and methods from 23+ programming languages
• Parent-Child Relationships - Maintain hierarchical chunk-context for complete understanding
• Semantic Search - Find relevant code using natural language queries
• Multiple Search Modes - Semantic, symbol-based, pattern matching, and hybrid search
• Smart Deduplication - Hash-based detection of duplicate code
• TOON Format Export - Token-efficient output format for LLM prompts (40-60% token savings)
• Full Pipeline Automation - One command to chunk, embed, and store
• Docker-Ready - ChromaDB server included

Use Cases

Agentic AI Systems	RAG Applications	Code Intelligence
Dynamic code retrieval for autonomous coding agents	High-precision code retrieval for question answering	Cross-repository code search and discovery
Context provision for code generation	Context injection for code explanation and documentation	Duplicate and similar code detection
Multi-step reasoning over large codebases	Semantic code search across repositories	Legacy codebase analysis and understanding
Tool integration for agent frameworks	Parent-child relationship tracking for complete context	MCP-compliant async architecture

---

Getting Started

Prerequisites

• Python 3.11 or higher
• Docker (for ChromaDB)
• OpenAI API key (for embeddings)

Installation

pip install contextinator

Verify the installation (requiers chromadb & openai api key setup):

contextinator --help

For detailed setup and configuration, see USAGE.md

Getting Started

1. Index a repository:

contextinator chunk-embed-store-embeddings \
  --repo-url https://github.com/user/repo \
  --save \
  --collection-name MyRepo

2. Search your codebase:

# Natural language semantic search
contextinator search "authentication logic" -c MyRepo

# Find specific functions
contextinator symbol authenticate_user -c MyRepo

# Export results in TOON format for LLM consumption
contextinator search "error handling" -c MyRepo --toon results.json

For comprehensive CLI and Python API documentation, see USAGE.md

Acknowledgements

Built with and inspired by amazing open-source projects:

Core Technologies

• tree-sitter - Incremental parsing system for AST generation
• ChromaDB - AI-native embedding database
• OpenAI - Embedding generation API

Inspired By

• Serena - Code intelligence and semantic search
• Continue - AI-powered code assistant
• Tabby - Self-hosted AI coding assistant
• Semantic Code Search - Code search and retrieval
• Aider - AI pair programming in the terminal
• VS Code Copilot Chat - Conversational AI for code

License

Licensed under the Apache License, Version 2.0. See LICENSE for details.

TL;DR

Contextinator is a code intelligence tool that uses Abstract Syntax Tree (AST) parsing to extract semantic code chunks, generates embeddings, and stores them in a vector database. This enables AI systems to understand, navigate, and reason about codebases with precision.

Star History