codesense 0.4.0

AI-powered semantic code search with intelligent parsing for Python projects

CodeSense

AI-powered semantic code search for Python projects. Auto-detects Django, FastAPI, Flask and understands natural language queries.

Key Features

• Smart Search - Natural language queries with auto-intent detection
• Framework Support - Auto-detects Django, FastAPI, Flask patterns
• Python API - Use programmatically in your code
• Fast & Local - FAISS-powered search, all data stored locally
• Incremental Updates - Update only changed files, 10x faster
• Chunk-based - Search functions, classes, methods separately

Installation

pip install codesense

From source:

git clone https://github.com/steliarix/codesense.git
cd codesense
pip install -e .

See Installation Guide for details.

Quick Start

CLI

# Index your project
codesense index ~/projects/my-app --name my_app

# Search with natural language
codesense search "user authentication" --index my_app

# Update after code changes
codesense update my_app

Python API

from codesense import CodeSense

cs = CodeSense(index_name="my_project")
cs.index("/path/to/project")

results = cs.search("user authentication", top_k=5)
for result in results:
    print(f"{result.file_path}:{result.start_line}")

Documentation

• Installation - Installation methods and setup
• CLI Usage - Command reference and options
• Python API - API reference and methods
• Examples - Real-world usage examples

How It Works

1. Index - Parse Python files using AST, extract code chunks
2. Embed - Create semantic embeddings with sentence-transformers
3. Search - Find similar code using FAISS vector similarity

CodeSense understands concepts, not just keywords:

• "verify user identity" → finds authentication code
• "product model" → finds only Django/Pydantic models
• "api endpoint" → finds FastAPI/Flask routes

Technologies

• sentence-transformers - Semantic embeddings
• FAISS - Vector search
• Click - CLI framework

Roadmap

• ✅ v0.4 - Universal framework support + Python API
• ✅ v0.3 - Incremental updates
• ✅ v0.2 - Chunk-based parsing
• 🔜 v0.6 - MCP integration for AI tools
• 🔜 v0.7 - Multi-language support (JS, TypeScript, Java)

See CHANGELOG.md for full version history.

Requirements

• Python 3.9+
• 1GB RAM (for embedding model)
• ~1-5 MB storage per 100 indexed files

License

MIT License - see LICENSE for details.

Links

• GitHub
• Issues
• PyPI

---

Author: Artem | Version: 0.4.0