Rocket.Chat Code Analyzer

This tool helps AI agents understand the Rocket.Chat codebase. It scans TypeScript files and creates compressed "skeletons" (summaries) of the code. This allows you to analyze large parts of the application without using too many tokens.

Features

• Smart Scanning: Finds relevant files based on your question.
• Skeleton Generation: Removes function bodies to save space, keeping only structure and types.
• Module Scoping: Can focus on specific Rocket.Chat modules to reduce noise.
• Hybrid Retrieval: Combines lexical, structural, and dependency-graph analysis for smarter file ranking.
• Grounded Responses: Agent answers include evidence citations linking to specific files and symbols.
• Caching: Hybrid in-memory + persistent disk cache for repeated queries.

Performance

Tests

• Status: ✅ Passing (Core scanner, Scope enforcement, Agent tools, Ranking)

Benchmarks

Tested on Rocket.Chat (apps/meteor) with real-world queries (Messaging, Auth, E2E):

Metric	Result
Token Reduction (vs Full Scan)	~96.5%
Token Reduction (vs Top-N)	~78.2%
Avg top-1 keyword coverage	~75%
Directory diversity (top-5)	~80%
Average Execution Time	~1.2s

Ranking Strategy (Hybrid):

• Lexical: Path and content token matching
• Structural: Export/import graph and symbol density
• Dependency: Graph centrality from import relationships
• Cross-Rerank: Phrase-level relevance pass on top candidates

Usage

Option 1: CLI (Direct Usage)

You can run the tool using npm run cli.

Basic Example

npm run cli -- --question "how are messages sent?"

Using Modules (Recommended)

To get better results, tell the tool which part of Rocket.Chat to look at:

npm run cli -- --question "permission checks" --modules authorization

Available Modules:

• lib-server-functions
• authorization
• e2e
• file-upload

All Options

• --question: (Required) The question you want to answer.
• --root: The folder to scan (default: current folder).
• --limit: How many files to return (default: 10).
• --modules: Comma-separated list of modules to filter by.
• --with-mapping: Creates a map of symbols to find definitions later.
• --skip-cache: Force full analysis without caching.

Caching

The tool caches analysis results for faster repeated queries. Cache is stored in .code-analyzer-cache/.

Cache management:

# View cache statistics
npm run cli -- --cache-stats

# Clear cache
npm run cli -- --cache-clear

# Skip cache for current query
npm run cli -- --question "how are messages sent?" --skip-cache

Cache Details:

• Strategy: Hybrid (in-memory + persistent disk cache)
• Invalidation: MD5 file hashing (detects changes automatically)
• TTL: 24 hours
• Location: .code-analyzer-cache/

Agent Mode (Interactive)

Run the tool in interactive mode with symbol lookup and dependency analysis:

npm run cli -- --interactive --root /path/to/project

Available Commands:

• /help - Show command list
• /files - List currently retrieved files
• /symbols <query> - Search for symbols in the codebase
• /symbol <symbolId> - Get full implementation of a symbol
• /deps <filePath> - Show imports and symbols for a file
• /refresh - Re-analyze project with new context
• /stats - Show mapping statistics
• /exit - Quit

Grounded Responses: Answers include an "Evidence" section that cites the exact files and symbols used, making it easy to verify answers and trace to source code.

Option 2: Gemini CLI Integration (MCP Server)

Use this tool as an MCP (Model Context Protocol) server with Gemini CLI for AI-powered code analysis and benchmarking.

Setup

1. Build the extension:

   npm run build
   cd gemini-extension && npm install && npm run build

2. Configure Gemini CLI:

   Create or update ~/.gemini/config.json:

   {
     "mcpServers": {
       "rc-code-analyzer": {
         "command": "node",
         "args": ["/path/to/code-analyzer/gemini-extension/dist/server.js"]
       }
     }
   }

3. Run Gemini CLI:

   cd /path/to/code-analyzer
   gemini

Example Queries with Gemini CLI

How are messages sent in Rocket.Chat?
How does user authentication work?
How are permissions checked?
What is the E2E encryption flow?

Gemini CLI will automatically use the scan_codebase tool to analyze relevant files and provide insights with token savings metrics.

Benefits

• AI-Powered Analysis: Gemini CLI provides context-aware insights and suggestions.
• Token Efficiency: Skeletons reduce token usage by ~96.5% vs full scans.
• Easy Integration: Works seamlessly with Gemini CLI's MCP protocol.
• Benchmarking: Track token stats and performance across queries.

Install

npm install
npm run build

For Gemini CLI integration:

cd gemini-extension
npm install
npm run build

Testing

Run Tests

npm test

Test MCP Server (Optional)

To verify the Gemini CLI extension works correctly:

npm run build && node test-gemini-server.mjs

This tests tool discovery and the scan_codebase tool without requiring Gemini CLI.