code-intelligence

Code Intelligence (LanceDB)

Overview

A semantic search and retrieval system powered by LanceDB and KConsole Gemini Embeddings. It allows you to search your codebase by "meaning" rather than just keywords.

When to Use

• You need to find code but don't know the exact filenames or variable names.
• Working across multiple microservices where dependencies are unclear.
• "Remembering" architecture patterns from different parts of the project.

Usage

1. Index your Project

Run this in a background tmux pane to build the initial memory: ```bash python scripts/indexer.py --index ./your-project ```

2. Semantic Search

Ask for a concept, not a keyword: ```bash python scripts/indexer.py --search "How is the user authentication handled?" ```

Best Practices

• Exclude noise: The indexer automatically ignores hidden files and common build artifacts.
• Refresh: Re-index after major refactors or adding new services.
• Combined with tmux: Keep a "Monitor" pane open to run the indexer in the background.

Implementation

The system uses `text-embedding-004` via KConsole AI Gateway and stores vectors in a local LanceDB instance (`.lancedb`).