Agentic Doc - AI-Powered Documentation Generator

Professional AI-Powered Documentation Generator for Large Codebases

A powerful CLI tool that uses AI (OpenAI GPT or Google Gemini) to automatically generate comprehensive, multi-level documentation for codebases of any size. Features interactive mindmap visualization, dependency analysis, and smart caching.

👨‍💻 About the Creator

Created, Designed, and Developed by Ish Kapoor

Role: Product Manager & Full-Stack Developer

This project was conceived and built from the ground up as a comprehensive solution to the documentation problem in modern software development. Every aspect—from architecture design to implementation—was carefully crafted to deliver a professional-grade tool that scales to enterprise codebases.

Vision: Transform documentation from a manual burden into an automated, AI-powered process that keeps pace with code evolution.

✨ Features

• 🔍 Smart Indexing: Recursive file scanning with .gitignore support
• 🧠 AI-Powered Docs: Uses LLMs (OpenAI, Gemini) to generate high-quality documentation
• 📊 Symbol Graph: Tracks functions, classes, and their relationships
• 🗺️ Interactive Mindmap: Beautiful, interactive visualization of your codebase
• 📁 Multi-Level Docs:
  • File-level: Detailed documentation for each source file
  • Directory-level: Module/package overviews
  • Architecture-level: System-wide architecture analysis
• 💾 Smart Caching: Saves LLM responses to avoid redundant API calls
• ⚡ Incremental Updates: Only re-documents changed files
• 📈 API Request Tracking: Monitor and optimize your API usage
• 🔧 Easy Configuration: Interactive setup wizard for API keys

📸 Screenshots

Interactive Mindmap Visualization

Explore your codebase with force-directed graph visualization showing files, functions, and dependencies

CLI in Action

Beautiful terminal output with progress tracking and detailed statistics

System Architecture

Clean, modular architecture designed for scalability and extensibility

🚀 Quick Start

Installation

# Install from PyPI (when published)
pip install agentic-aish

# Or install from source
git clone https://github.com/ishkapoor2000/agentic-ai.git
cd agentic-ai
pip install -e .

Initial Setup

# 1. Configure your API keys and provider
agentic-doc configure

# 2. Initialize your project
agentic-doc init

# 3. Scan your codebase
agentic-doc scan

# 4. Generate documentation
agentic-doc doc files --limit 10

# 5. Create interactive mindmap
agentic-doc mindmap

📖 CLI Commands

Configuration

# Interactive configuration wizard
agentic-doc configure

# Switch between providers
agentic-doc switch-provider gemini
agentic-doc switch-provider openai

# Check current configuration
agentic-doc status

Documentation Generation

# Generate standard file documentation
agentic-doc doc files --limit 100

# Generate enhanced documentation (with dependencies & usage)
agentic-doc doc files --limit 50 --enhanced

# Generate directory-level documentation
agentic-doc doc dirs

# Generate architecture overview
agentic-doc doc architecture

Visualization

# Generate interactive HTML mindmap
agentic-doc mindmap

# Generate Mermaid diagram
agentic-doc mindmap --format mermaid

# Visualize dependencies
agentic-doc visualize-deps

# Generate interactive Mermaid viewer
agentic-doc visualize-deps --format mermaid-html

# Export graph data
agentic-doc graph --format json
agentic-doc graph --format html

Analysis Tools

Gain deep insights into your codebase structure and usage patterns.

1. Dependency Analysis

Identify the most critical parts of your codebase ("load-bearing" code).

# Generate a "Hot Functions Report" ranking functions by usage frequency
agentic-doc analyze-deps --show-hot

2. Use Case Documentation

Understand how your code is used in practice with real examples.

# Document usage patterns for functions used at least 5 times
agentic-doc use-cases --min-usage 5

3. Function Usage Network

Trace the complete call graph for specific functions.

# Generate detailed "Who calls whom?" reports
agentic-doc function-usage

Incremental Documentation

You can document different subsets of your codebase incrementally:

# Day 1: Document only the API layer
agentic-doc doc files --paths "app/api/**/*.py" --limit 20 --enhanced

# Day 2: Document the utilities
agentic-doc doc files --paths "app/utils/**/*.py" --limit 15 --enhanced

# Day 3: Document the database layer
agentic-doc doc files --paths "app/db/**/*.py" --limit 10 --enhanced

Benefits:

• Process large codebases incrementally
• Focus on specific modules or features
• Avoid redundant API calls (results are cached)
• Spread API usage over multiple days

🔑 API Keys Setup

Google Gemini (FREE)

1. Get your free API key: https://makersuite.google.com/app/apikey
2. Run agentic-doc configure and select Gemini
3. Enter your API key when prompted

OpenAI (PAID)

1. Get your API key: https://platform.openai.com/api-keys
2. Run agentic-doc configure and select OpenAI
3. Enter your API key when prompted

📊 API Request Tracking

Every documentation generation command automatically tracks and displays API usage:

==================================================
API Request Summary
==================================================
Total Requests: 45
Session Duration: 127.3s

By Provider:
  gemini: 45
==================================================

🗺️ Interactive Mindmap

The mindmap feature creates a beautiful, interactive visualization of your codebase:

• Force-directed graph showing files and symbols
• Interactive exploration with zoom, pan, and search
• Dependency visualization showing relationships
• Standalone HTML - no server required

💰 Cost Optimization

1. Use Gemini (FREE): Google's Gemini models are free for moderate usage
2. Start small: Use --limit to test on a subset of files
3. Smart caching: Responses are cached in SQLite
4. Incremental updates: Re-scan only updates changed files

Estimated costs (using gpt-4o-mini):

• Small project (100 files): ~$1-2
• Medium project (1,000 files): ~$10-20
• Large project (15,000 files): ~$150-300 (first run, then incremental)

📁 Output Structure

docs/
├── files/
│   ├── agentic_doc/
│   │   ├── cli.py.md
│   │   ├── config.py.md
│   │   └── ...
│   └── ...
├── dirs/
│   ├── agentic_doc/
│   │   ├── analysis.md
│   │   ├── db.md
│   │   └── ...
│   └── ...
├── architecture/
│   └── overview.md
├── analysis/
│   └── hot-functions.md
└── graph.html  # Interactive mindmap

🛠️ Supported Languages

• ✅ Python: Full AST-based analysis
• ✅ JavaScript/TypeScript: Regex-based extraction
• 🔜 More languages: Coming soon

🔧 Configuration File

.agentic-doc.yml:

root_path: .
model_provider: gemini  # or 'openai' or 'mock'
gemini_model: gemini-2.0-flash-exp
openai_model: gpt-4o-mini
exclude_globs:
  - '**/node_modules/**'
  - '**/.git/**'
  - '**/__pycache__/**'
  - '**/.venv/**'

🏗️ Architecture

For detailed architectural documentation, see docs/ARCHITECTURE.md.

Key Components:

• CLI Interface: Typer-based command-line interface
• File Indexer: AST/regex-based code analysis
• LLM Gateway: Multi-provider abstraction (OpenAI, Gemini)
• Database: SQLite-based caching and state management
• Graph Builder: D3.js-powered visualizations

📦 Distribution

Building the Package

# Install build tools
pip install build

# Build wheel and source distribution
python -m build

# Install locally
pip install dist/agentic_doc-1.0.0-py3-none-any.whl

Publishing to PyPI

# Install twine
pip install twine

# Upload to PyPI
twine upload dist/*

🤝 Contributing

Contributions, issues, and feature requests are welcome! See CONTRIBUTING.md for guidelines.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

Copyright (c) 2025 Ish Kapoor

🆘 Support & Troubleshooting

Common Issues

"No symbols with sufficient usage found" If use-cases returns this error, your dependency index might be empty. Run a forced scan to rebuild it:

agentic-doc scan --force

API Key Errors Ensure your .env file is in the root directory or configure keys globally:

agentic-doc configure

For other issues, questions, or feature requests, please open an issue on GitHub.

🎯 Project Roadmap

✅ Completed

• Core documentation generation
• Multi-provider LLM support
• Interactive mindmap visualization
• Dependency analysis
• Incremental documentation

🚧 In Progress

• PyPI package publishing
• Enhanced CLI features
• Performance optimizations

📋 Planned

• Watch mode (auto-regenerate on changes)
• Local LLM support (Ollama, llama.cpp)
• VSCode extension
• Web UI (optional)
• Plugin system

🌟 Acknowledgments

Concept, Design, and Development: Ish Kapoor

Built with passion to solve real-world documentation challenges in enterprise software development.

Made with ❤️ by Ish Kapoor
Product Manager & Developer