repomix（repomix 工具）

Repomix — Codebase Packer & Analyzer

Pack entire codebases into a single, AI-friendly file for analysis. Repomix intelligently collects repository files, respects .gitignore, runs security checks, and generates structured output optimized for LLM consumption.

When to Use

• "Analyze this repo" / "Explore this codebase"
• "What's the structure of facebook/react?"
• "Find all authentication-related code"
• "How many tokens is this project?"
• "Pack this repo for AI analysis"
• "Show me the main components of vercel/next.js"

Quick Reference

Pack a Remote Repository

npx repomix@latest --remote <owner/repo> --output /tmp/<repo-name>-analysis.xml

Always output to a temporary directory (/tmp on Unix, %TEMP% on Windows) for remote repositories to avoid polluting the user's working directory.

Pack a Local Directory

npx repomix@latest [directory] --output /tmp/<name>-analysis.xml

Key Options

Workflow

Step 1: Pack the Repository

Choose the appropriate command based on the target:

# Remote repository (always output to /tmp)
npx repomix@latest --remote yamadashy/repomix --output /tmp/repomix-analysis.xml

# Large remote repo with compression
npx repomix@latest --remote facebook/react --compress --output /tmp/react-analysis.xml

# Local directory
npx repomix@latest ./src --output /tmp/src-analysis.xml

# Specific file types only
npx repomix@latest --include "**/*.{ts,tsx,js,jsx}" --output /tmp/filtered-analysis.xml

Step 2: Check Command Output

The command displays:

• Files processed: Number of files included
• Total characters: Size of content
• Total tokens: Estimated AI tokens
• Output file location: Where the file was saved

Note the output file location for subsequent analysis.

Step 3: Analyze the Output

Structure overview:

1. Search for the file tree section (near the beginning of the output)
2. Check the metrics summary for overall statistics

Search for patterns (use the output file path from Step 2):

# Find exports and main entry points
grep -iE "export.*function|export.*class" <output-file>

# Search with context
grep -iE -A 5 -B 5 "authentication|auth" <output-file>

# Find API endpoints
grep -iE "router|route|endpoint|api" <output-file>

# Find database models
grep -iE "model|schema|database|query" <output-file>

Read specific sections using offset/limit for large outputs.

Step 4: Report Findings

• Metrics: Files, tokens, size from command output
• Structure: Directory layout from file tree analysis
• Key findings: Based on pattern search results
• Next steps: Suggestions for deeper exploration

Best Practices

1. Use --compress for large repos (>100k lines) to reduce token usage by ~70%
2. Use pattern search first before reading entire output files
3. Use a temporary directory for output (/tmp on Unix, %TEMP% on Windows) to keep the user's workspace clean
4. Use --include to focus on specific parts of a codebase
5. XML is the default and recommended format — it has clear file boundaries for structured analysis

Output Formats

Error Handling

• Command fails: Check error message, verify repository URL/path, check permissions
• Output too large: Use --compress, narrow scope with --include
• Network issues (remote): Verify connection, suggest local clone as alternative
• Pattern not found: Try alternative patterns, check file tree to verify files exist

Security

Repomix automatically excludes potentially sensitive files (API keys, credentials, .env files) through built-in security checks. Trust its security defaults unless the user explicitly requests otherwise.