code-digest 😋

Transform your entire codebase into a single, well-formatted Markdown file optimized for LLM context windows. Similar to gitingest, but faster and with built-in Gemini CLI integration.

Why code-digest?

Leverage Gemini's massive context window to understand your entire codebase at once. This tool gives AI assistants like Claude Code superpowers by enabling them to:

• 🏗️ Plan architectural changes with full visibility of your codebase
• 🔍 Answer complex questions about how different parts interact
• 📊 Analyze patterns and suggest improvements across your entire project
• 🚀 Make informed decisions when they need the big picture

Simply put: feed your entire repo to Gemini and let Claude Code have intelligent conversations about your code architecture.

🚀 Quick Start

Ask Questions About Your Codebase

# Analyze local architecture
code-digest "What are the main architectural patterns used in this codebase?"

# Analyze remote GitHub repository
code-digest --repo https://github.com/owner/repo "How is this project structured?"

# Understand dependencies
code-digest "How does the authentication system interact with the database?"

# Find improvement opportunities
code-digest "How can I implement the new feature X?"

🎯 Key Features

• 🚄 Blazing Fast: Built in Rust with parallel processing
• 🤖 Gemini Integration: Direct piping to Gemini CLI for instant AI analysis
• 🔗 Remote Repository Support: Analyze any GitHub repository directly without cloning
• 📊 Smart Token Management: Accurate token counting using tiktoken
• 🎯 Intelligent Prioritization: Automatically prioritizes important files when hitting token limits
• 🔍 Git-Aware: Respects .gitignore and custom .digestignore patterns

📦 Installation

Prerequisites

Install code-digest

# Using Cargo
cargo install code-digest

# Or from source
git clone https://github.com/matiasvillaverde/code-digest.git
cd code-digest
cargo install --path .

Install Gemini CLI (optional):

npm install -g @google/gemini-cli

Generate Context Files

# Process current directory
code-digest

# Save to file for later use
code-digest -o context.md

# Process specific directory with token limit
code-digest -d /path/to/project --max-tokens 100000

📋 Configuration

.digestignore

Exclude files from processing:

# Dependencies
node_modules/
target/
vendor/

# Build artifacts
dist/
build/
*.pyc

# Sensitive files
.env
secrets/

.digestkeep

Prioritize important files:

# Core functionality
src/main.*
src/core/**/*.rs

# Important configs
Cargo.toml
package.json

Configuration File (.code-digest.toml)

[defaults]
max_tokens = 150000
progress = true

[[priorities]]
pattern = "src/**/*.rs"
weight = 100

[[priorities]]
pattern = "tests/**/*.rs"
weight = 50

🔧 CLI Reference

code-digest [OPTIONS] [PROMPT]

Arguments:
  [PROMPT]  The prompt to send to Gemini

Options:
  -d, --directory <PATH>      Directory to process [default: .]
      --repo <URL>            GitHub repository URL (e.g., https://github.com/owner/repo)
  -o, --output <FILE>         Output to file instead of stdout
      --max-tokens <N>        Maximum tokens for output
  -q, --quiet                 Suppress output except errors
  -v, --verbose               Enable verbose logging
  -c, --config <FILE>         Path to config file
      --progress              Show progress indicators
  -h, --help                  Print help
  -V, --version               Print version

🧪 Common Use Cases

Architecture Review

code-digest "Create a high-level architecture diagram of this codebase"

Security Audit

code-digest "Identify potential security vulnerabilities in this codebase"

Analyze Remote Projects

# Analyze any public GitHub repository
code-digest --repo https://github.com/rust-lang/rust "How is the compiler structured?"

# Compare architectures
code-digest --repo https://github.com/owner/project "Compare this architecture to best practices"

Documentation Generation

code-digest "Generate comprehensive API documentation for all public functions"

Code Quality Analysis

code-digest "What code smells or anti-patterns exist in this project?"

🤝 Contributing

We welcome contributions! Please see our Contributing Guidelines for details.

Development

# Run tests
make test

# Run all checks
make validate

🐛 Troubleshooting

gemini not found

• Ensure Gemini CLI is installed: pip install gemini-cli
• Verify it's in your PATH: which gemini

Token count exceeded

• Use --max-tokens to set a limit
• Configure file priorities in .digestkeep
• Exclude unnecessary files in .digestignore

🚧 Roadmap

• Support for more tokenizers (GPT-4, Claude, etc.)
• Custom output templates
• Integration with more LLM CLIs

Made with ❤️ and Rust