TechDocAgent

AI-powered technical documentation generation for codebases

Features • Installation • Quick Start • Documentation • Examples

---

Overview

TechDocAgent is an intelligent documentation generation system that automatically creates comprehensive technical documentation from your codebase. Powered by Google's Gemini AI and advanced code analysis techniques, it transforms source code into human-readable documentation with minimal effort.

The project offers two implementations:

• Basic (techdocagent): Lightweight, straightforward documentation generation for simple projects
• Advanced (techdocagent_advanced): Enterprise-grade system with memory, semantic search, incremental updates, and feedback learning

---

Features

Feature	Description
Vector Search	FAISS-powered semantic code search using embeddings
Language Detection	Automatic detection of 15+ programming languages
Change Detection	Git and hash-based change tracking with dependency analysis
Incremental Updates	Smart regeneration of only affected documentation
Feedback Loop	Rating system, corrections, and continuous improvement
Multi-Format	9+ documentation types (README, API, Onboarding, etc.)
AST Analysis	Advanced Abstract Syntax Tree parsing for deep code understanding

---

Installation

Prerequisites

• Python 3.8 or higher
• Google Gemini API key (Get one here)

Basic Installation

# Clone the repository
git clone https://github.com/smilingprogrammer/TechDocAgent.git
cd TechDocAgent

# Install dependencies
pip install -r requirements.txt

# Set your API key
export GEMINI_API_KEY="your-api-key-here"

Development Installation

# Install with development dependencies
pip install -e ".[dev]"

# Install with CLI tools
pip install -e ".[cli]"

# Install everything
pip install -e ".[dev,cli]"

---

Quick Start

Basic Usage

from techdocagent.ingestion import ingest_codebase
from techdocagent.analysis import analyze_file
from techdocagent.llm import generate_documentation

# Ingest your codebase
files = ingest_codebase('./your-project-path')

# Analyze files
metadata = [analyze_file(f) for f in files]

# Generate documentation
docs = generate_documentation(files, metadata)
print(docs)

Advanced Usage

from techdocagent_advanced import TechDocAgent

# Use context manager for automatic cleanup
with TechDocAgent() as agent:
    # Analyze codebase and build embeddings
    summary = agent.analyze_codebase()
    print(f"Analyzed {summary['total_files']} files")

    # Generate README
    readme = agent.generate_documentation('README')

    # Search code semantically
    results = agent.search_code("authentication logic")
    for chunk, score in results[:3]:
        print(f"  - {chunk['name']}: {score:.2f}")

    # Collect feedback for continuous improvement
    agent.collect_feedback(doc_id=1, rating=5, comment="Excellent!")

Command Line (Advanced)

# Generate documentation
techdoc-advanced generate --type README --output docs/

# Update documentation after code changes
techdoc-advanced update --type README

# Search codebase
techdoc-advanced search "error handling"

# View statistics
techdoc-advanced stats

---

Documentation Types Supported

Type	Description
README	Comprehensive project documentation
API	API reference with endpoints and examples
ONBOARDING	Developer onboarding and setup guides
CHANGELOG	Automated changelog from git history
ARCHITECTURE	System architecture and design docs
MODULE	Per-module/file documentation
FUNCTION	Function-level detailed documentation
TEST	Test suite documentation
DEPLOYMENT	Deployment and operations guides

Configuration

Create a .techdocagent.json file in your project root:

{
  "gemini_api_key": "your-api-key-here",
  "project_root": ".",
  "output_dir": "./docs",
  "features": {
    "embeddings": true,
    "change_detection": true,
    "feedback_loop": true,
    "ast_analysis": true
  },
  "temperature": 0.7,
  "max_tokens": 8000
}

Or use environment variables:

export GEMINI_API_KEY="your-api-key"
export TECHDOC_OUTPUT_DIR="./docs"
export TECHDOC_PROJECT_ROOT="."

---

Examples

Example 1: Generate Multiple Documentation Types

from techdocagent_advanced import TechDocAgent

with TechDocAgent() as agent:
    # Analyze once
    agent.analyze_codebase()

    # Generate multiple doc types
    doc_types = ['README', 'API', 'ONBOARDING', 'ARCHITECTURE']
    for doc_type in doc_types:
        doc = agent.generate_documentation(doc_type)
        print(f"{doc_type} generated ({len(doc)} chars)")

Example 2: Incremental Updates

from techdocagent_advanced import TechDocAgent

with TechDocAgent() as agent:
    # Initial documentation
    agent.analyze_codebase()
    agent.generate_documentation('README')

    # ... make code changes ...

    # Smart update - only regenerates if changed
    updated = agent.update_documentation('README')
    print("Documentation updated!")

Example 3: Semantic Code Search

from techdocagent_advanced import TechDocAgent

with TechDocAgent() as agent:
    agent.analyze_codebase()

    # Natural language search
    queries = [
        "database connection logic",
        "error handling middleware",
        "user authentication"
    ]

    for query in queries:
        results = agent.search_code(query, top_k=3)
        print(f"\n Query: {query}")
        for chunk, score in results:
            print(f"  - {chunk['name']} (similarity: {score:.2f})")

You can find more examples here

Roadmap

Current Status

• ✅ Basic documentation generation
• ✅ Advanced memory system
• ✅ Vector embeddings and search
• ✅ Change detection
• ✅ Feedback loop
• ✅ Multi-format documentation

Future Plans

• 🔜 CLI interface with rich output
• 🔜 Web UI for interactive documentation
• 🔜 GitHub Actions integration
• 🔜 Support for more output formats (HTML, PDF)
• 🔜 Plugin system for custom doc types
• 🔜 Multi-language documentation (i18n)
• 🔜 Custom LLM backend support (OpenAI, Claude, etc.)

---

⭐ Star it GitHub — it motivates me a lot!