semantic-search

Search codebases by concept and intent using vector embeddings in Weaviate.

1|1|Updated Nov 25, 2025
One-click install
npx skills add https://github.com/PearlThoughts/CodeCompass --skill semantic-search
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: semantic-search
Source: https://github.com/PearlThoughts/CodeCompass/tree/main/.claude/skills/semantic-search
Command: npx skills add https://github.com/PearlThoughts/CodeCompass --skill semantic-search

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Traditional keyword-based search (grep) often fails to find relevant code when you don't know exact names or when concepts are expressed differently. This Skill enables finding code by its meaning and intent, drastically improving code discovery.

Core Features & Use Cases

  • Concept-Based Queries: Search for code based on its purpose (e.g., "payment processing logic") rather than exact function names.
  • Intent-Based Discovery: Understand how features are implemented (e.g., "how is user authentication implemented?") even in unfamiliar codebases.
  • Fuzzy & Exploratory Search: Ideal for exploring new projects or when exact terms are unknown, finding similar patterns across files and languages.
  • Use Case: You need to understand how a legacy system handles customer refunds but don't know the exact class or function names. Use semantic search to find all relevant code snippets by querying "logic for processing customer refunds."

Quick Start

Use the semantic-search skill to find all code related to "user profile management and data updates" in the attached codebase.

Frequently Asked Questions about semantic-search

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I find code by meaning instead of exact keywords?

Semantic search finds code by its intent and purpose rather than keyword matching. It uses embeddings to understand what code does, letting you query concepts like "payment processing logic" and discover relevant snippets across your codebase even when exact function names are unknown.

When should I use semantic search instead of grep?

Use semantic search when keyword-based grep fails—especially for concept-based queries, unfamiliar codebases, cross-language patterns, or when implementation details vary. It excels at intent-based discovery where traditional grep returns no results or noise.

What do I need to set up semantic search in my codebase?

Semantic search requires an indexed codebase with embeddings generated via Ollama mxbai-embed-large (1024 dimensions) and vector similarity search powered by Weaviate. Once configured, query through codecompass search:semantic for standardized results.

Can semantic search work across multiple files and programming languages?

Yes, semantic search enables cross-file and cross-language exploration by matching code meaning rather than syntax. Query once to find similar patterns, implementations, or concepts regardless of file type or language used.

What's the best way to formulate queries for semantic code discovery?

Express queries as natural language descriptions of intent or functionality—"how is user authentication implemented?" or "logic for processing customer refunds." The semantic search engine interprets meaning, returning relevant code without requiring exact naming conventions.