grepai-search-basics_skill

Overview

This skill teaches the fundamentals of semantic code search using GrepAI. It explains how semantic search differs from traditional text grep, how to run basic queries, interpret scores, and troubleshoot common issues. You'll learn practical commands and query patterns to find relevant code quickly.

How this skill works

GrepAI indexes code into semantic embeddings and returns ranked matches by meaning rather than exact text. You run searches with natural language queries (grepai search "your query"), and results show a similarity score, file path, line range, and code snippet. Adjust limits, check index status, and re-index when files are missing.

When to use it

• When you want to find code by intent or behavior rather than exact names
• While learning how semantic search differs from traditional grep
• To locate entry points, error handling, data access, or business logic
• When triaging code you don’t know the exact function names for
• Before debugging or refactoring to discover related code across the repo

Best practices

• Write queries describing intent or behavior (e.g., "validate user credentials")
• Use 3–7 words to give enough context without being verbose
• Prefer phrases that describe what the code does instead of exact names
• Iterate queries if results are irrelevant and try synonyms
• Check index status (grepai status) and re-run grepai watch if needed

Example use cases

• Search for authentication flows: grepai search "user authentication flow"
• Find database access points: grepai search "fetch user from database"
• Locate error handling patterns: grepai search "error handling and logging"
• Discover startup code: grepai search "application startup"
• Filter top results: grepai search "send email notification" --limit 3