Do I need to manually index every project?

No, the skill features auto-discovery; searching a new project directory automatically registers and begins indexing it in the background.

What are the technical prerequisites?

You will need Docker installed to run the pgvector container, Python 3.10 or higher, and a Gemini API key to generate the code embeddings.

Does indexing impact my machine's performance?

The indexer runs as a background daemon with configurable watcher fallbacks, ensuring your development environment remains responsive while code changes are processed.

Can I search across multiple repositories?

Yes, the skill supports multi-project management and allows you to search specific project paths or the current working directory seamlessly.

How does semantic search differ from using grep?

Unlike grep which finds literal text matches, semantic search uses AI embeddings to find code based on its functional meaning and intent, even if the specific keywords in your query don't appear in the source.

Semantic Codebase Search

Name: Semantic Codebase Search
Author: amoscicki

byamoscicki

0•

开发者工具

Enables high-accuracy semantic code discovery using vector embeddings to find logic by meaning rather than literal keyword matches.

The Semantic Codebase Search skill transforms how developers navigate complex repositories by leveraging pgvector and Gemini embeddings to understand the intent behind code. Instead of relying on traditional string matching, it allows Claude to locate relevant functions, components, and logic patterns through natural language queries. It features zero-config auto-discovery, background indexing daemons for multi-project support, and precise byte-range mapping, making it an essential first step for codebase exploration, refactoring, and impact analysis.

主要功能

01Precise result mapping including file paths, byte ranges, and similarity scores

02Integrated workflow for upstream and downstream impact review

03Automatic project discovery and background indexing daemons

040 GitHub stars

05Natural language code search powered by vector embeddings and pgvector

06Multi-project support with centralized state management

使用场景

01Onboarding to large, unfamiliar codebases by searching for high-level concepts

02Performing impact analysis to identify callers and dependencies before refactoring

03Locating specific implementation patterns when traditional keyword search fails

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add amoscicki/aromatt semantic-search

For use in Claude.ai and ChatGPT

Download Skill