dococtopy 0.2.0

A language-agnostic docstyle compliance & remediation tool

DocOctopy

A language-agnostic docstyle compliance & remediation tool that scans code for docstring/docblock presence and style, reports findings, and can auto-propose LLM-based fixes.

Features

🔍 Comprehensive Scanning

• Python-first with extensible architecture for other languages
• Google-style docstring validation with detailed compliance checking
• Context-specific rules for test functions, public APIs, and exception handling
• AST-based analysis for accurate symbol and signature detection
• Smart caching with incremental scanning for large codebases
• Sophisticated complexity detection using cyclomatic complexity metrics
• Multi-style docstring support (Google, Sphinx, NumPy, Facebook)
• Content quality validation (TODO/placeholders, conflict markers, delimiter style)

📊 Multiple Output Formats

• Pretty console output with Rich formatting
• JSON reports for CI/CD integration
• SARIF format for GitHub Code Scanning
• Configurable exit codes based on severity levels

🤖 LLM-Powered Remediation

• Automatic docstring generation for missing documentation
• Smart fixing of non-compliant docstrings
• Enhancement of existing docstrings with missing elements
• DSPy integration for reliable, structured LLM interactions
• Interactive review mode with diff preview and approval workflow
• Multiple LLM providers (OpenAI, Anthropic, Ollama)

⚙️ Flexible Configuration

• pyproject.toml integration with rule enable/disable switches
• Per-path overrides for different project sections
• Gitignore-style exclusions with pathspec support
• Rule severity customization (error, warning, info, off)

Installation

Basic Installation

pip install dococtopy

With LLM Support

pip install dococtopy[llm]

Note for uv users: uv doesn't support the [extra] syntax yet. Use uv pip instead:

This works with uv

uv pip install dococtopy[llm]

This also works with uv (if you have a dev group in pyproject.toml)

uv pip install --group dev dococtopy[llm]

This doesn't work with uv (known limitation)

uv add dococtopy[llm] --dev # Won't install DSPy correctly

### Development Installation

```bash
git clone https://github.com/CrazyBonze/DocOctopy.git
cd DocOctopy
uv sync --group dev

Development Scripts

The scripts/ directory contains utility scripts for development and testing:

• comprehensive_compare_models.py - Compare LLM models for docstring generation quality and cost
• pre-commit.sh - Pre-commit hook for formatting and linting
• publish.sh - Publishing script for PyPI releases
• See scripts/README.md for detailed usage instructions

Quick Start

1. Scan Your Code

# Scan current directory
dococtopy scan .

# Scan specific paths
dococtopy scan src/ tests/

# Get JSON output
dococtopy scan . --format json --output-file report.json

# Use SARIF for GitHub Code Scanning
dococtopy scan . --format sarif --output-file report.sarif

2. Fix Issues with LLM Assistance

# Dry-run mode (safe, shows what would be fixed)
dococtopy fix . --dry-run

# Interactive mode (review each change)
dococtopy fix . --interactive

# Fix specific rules only
dococtopy fix . --rule DG101,DG202 --dry-run

# Use different LLM provider
dococtopy fix . --llm-provider anthropic --llm-model claude-haiku-3.5

# Use local Ollama server
dococtopy fix . --llm-provider ollama --llm-model codeqwen:latest --llm-base-url http://localhost:11434

3. Configure Your Project

Create a pyproject.toml file:

[tool.docguard]
exclude = ["**/.venv/**", "**/build/**", "**/node_modules/**"]

[tool.docguard.rules]
DG101 = "error"    # Missing docstrings
DG201 = "error"    # Google style parse errors
DG202 = "error"    # Missing parameters
DG203 = "error"    # Extra parameters
DG204 = "warning"  # Returns section issues
DG205 = "info"     # Raises validation
DG301 = "warning"  # Summary style
DG302 = "warning"  # Blank line after summary
DG211 = "info"     # Yields section validation
DG212 = "info"     # Attributes section validation
DG213 = "info"     # Examples section validation
DG214 = "info"     # Note section validation
DG215 = "info"     # Private method docstring recommendation
DG216 = "info"     # Dunder method docstring recommendation
DG303 = "warning"  # Content quality validation
DG304 = "info"     # Docstring delimiter style
DG401 = "warning"  # Test function docstring style
DG402 = "warning"  # Public API function documentation
DG403 = "warning"  # Exception documentation completeness

# Per-path overrides
[[tool.docguard.overrides]]
patterns = ["tests/**"]
rules.DG101 = "off"  # Disable missing docstrings in tests

LLM Setup

Option A: Local Ollama (Recommended for Development)

1. Install Ollama: Download from ollama.ai

2. Pull a model:

   ollama pull codeqwen:latest
   # or
   ollama pull llama3.1:8b
   

3. Test DocOctopy:

   dococtopy fix . --llm-provider ollama --llm-model codeqwen:latest --llm-base-url http://localhost:11434 --dry-run
   

Option B: OpenAI

1. Get API key: OpenAI API Keys

2. Set environment variable:

   export OPENAI_API_KEY="your-api-key"
   

3. Test DocOctopy:

   dococtopy fix . --llm-provider openai --llm-model gpt-5-nano --dry-run
   

Option C: Anthropic

1. Get API key: Anthropic Console

2. Set environment variable:

   export ANTHROPIC_API_KEY="your-api-key"
   

3. Test DocOctopy:

   # Use best Anthropic model (claude-haiku-3.5) - Highest quality score
   dococtopy fix . --llm-provider anthropic --llm-model claude-haiku-3.5 --dry-run
   
   # Or use budget Anthropic option (claude-haiku-3) - Best value
   dococtopy fix . --llm-provider anthropic --llm-model claude-haiku-3 --dry-run
   

Rules Reference

Basic Compliance Rules

• DG101: Missing docstring (functions and classes)
• DG301: Summary first line should end with period
• DG302: Blank line required after summary

Google Style Validation Rules

• DG201: Google style docstring parse error
• DG202: Parameter missing from docstring
• DG203: Extra parameter in docstring
• DG204: Returns section missing or mismatched
• DG205: Raises section validation
• DG206: Args section format validation
• DG207: Returns section format validation
• DG208: Raises section format validation
• DG209: Summary length validation
• DG210: Docstring indentation consistency

Advanced Google Style Rules

• DG211: Generator functions should have Yields section
• DG212: Classes with public attributes should have Attributes section
• DG213: Complex functions should have Examples section
• DG214: Functions with special behavior should have Note section

Content Quality Rules

• DG303: Detect TODO/placeholder content and version control conflict markers
• DG304: Detect single quote delimiters vs double quotes (style consistency)

Recommendation Rules

• DG215: Private methods should have docstrings (recommendation)
• DG216: Standard dunder methods should have docstrings (recommendation)

Context-Specific Rules

• DG401: Test functions should have descriptive docstrings explaining what they test
• DG402: Public API functions should have comprehensive documentation (Args, Returns, Raises sections)
• DG403: Functions should document all exceptions they raise in the Raises section

Context-Specific Rules Details

DG401: Test Function Docstring Style

Ensures test functions have descriptive docstrings that explain what they're testing, improving test readability and debugging.

Examples:

# ✅ Good - Descriptive test docstring
def test_user_authentication_works_correctly():
    """Test that user authentication validates credentials properly."""
    # Test implementation...

# ❌ Bad - Generic or non-descriptive
def test_user_login():
    """Test."""  # Too short and generic
    # Test implementation...

def test_something():
    """Test function."""  # Generic pattern
    # Test implementation...

What it checks:

• Functions starting with test_ or containing "test" in the name
• Docstrings must be descriptive (>20 characters)
• Avoids generic patterns like "Test", "Test function", etc.

DG402: Public API Function Documentation

Requires public API functions to have comprehensive documentation with Args, Returns, and Raises sections.

Examples:

# ✅ Good - Complete public API documentation
def process_data(data, options=None):
    """Process the input data according to the given options.
    
    Args:
        data: The input data to process
        options: Optional configuration options
        
    Returns:
        Processed data result
        
    Raises:
        ValueError: If data format is invalid
    """
    # Implementation...

# ❌ Bad - Missing required sections
def process_data(data, options=None):
    """Process the input data."""  # Missing Args, Returns, Raises
    # Implementation...

What it checks:

• Public functions (non-private, non-test, non-dunder)
• Requires Args, Returns, and Raises sections
• Skips internal/helper functions automatically

DG403: Exception Documentation Completeness

Ensures functions document all exceptions they raise in the Raises section.

Examples:

# ✅ Good - All exceptions documented
def risky_function():
    """Do something risky.
    
    Raises:
        ValueError: If input is invalid
        RuntimeError: If operation fails
    """
    raise ValueError("test")
    raise RuntimeError("test")

# ❌ Bad - Undocumented exceptions
def risky_function():
    """Do something risky."""
    raise ValueError("test")  # Not documented
    raise RuntimeError("test")  # Not documented

What it checks:

• Uses AST analysis to detect raise statements
• Parses docstring Raises sections
• Flags undocumented exceptions with specific names

Interactive Fix Mode

DocOctopy includes an interactive mode that lets you review and approve each proposed change:

dococtopy fix . --interactive

Interactive Features

• Diff preview: See exactly what will be changed
• Change-by-change review: Accept or reject each fix individually
• Rich formatting: Beautiful console output with colors
• Summary statistics: Track approved vs rejected changes

Example Interactive Session

Found 3 changes for src/main.py

Change: process_data (function)
Issues: DG101
Proposed docstring:
    """Process the input data and return results.

    Args:
        data: The input data to process.
        options: Processing options.

    Returns:
        Processed data results.
    """
Show diff? [Y/n]: y
--- Original
+++ Proposed
@@ -15,6 +15,15 @@
 def process_data(data, options):
+    """Process the input data and return results.
+
+    Args:
+        data: The input data to process.
+        options: Processing options.
+
+    Returns:
+        Processed data results.
+    """
     result = []
     for item in data:
         result.append(transform(item, options))
Apply this change? [Y/n]: y
✓ Applied change for process_data

Summary:
- Total changes: 3
- Applied: 1
- Rejected: 1
- Skipped: 1

Recommended Models

DocOctopy supports multiple LLM providers with different models optimized for docstring generation. Based on comprehensive testing with real-world code:

🏆 OpenAI Models (Recommended)

Model	Cost (per 1M tokens)	Quality Score	Quality per Dollar	Best For
gpt-5-nano	$0.45	39/50	39,796	✅ Default choice - Best value
gpt-5-mini	$2.25	41/50	8,367	✅ Premium choice - Enterprise quality
gpt-4.1-mini	$2.00	41/50	9,491	Alternative option
gpt-4.1-nano	$0.50	46/50	42,593	Budget option

Key Findings:

• gpt-5-nano: 5x cheaper than GPT-5-mini with 95% of the quality - exceptional value
• gpt-5-mini: Comprehensive documentation with detailed business logic and examples
• gpt-4.1-mini: Solid alternative with good quality but higher cost than GPT-5-nano
• gpt-4.1-nano: Budget option with good quality-per-dollar ratio

🤖 Anthropic Models

Model	Cost (per 1M tokens)	Quality Score	Quality per Dollar	Best For
claude-haiku-3.5	$0.25	67/50	6,442	✅ Best Anthropic - Highest quality
claude-sonnet-4	$3.00	41/50	1,051	High performance option
claude-haiku-3	$0.25	41/50	12,615	Budget Anthropic option
claude-opus-4.1	$15.00	41/50	210	Premium option (expensive)

Anthropic Highlights:

• claude-haiku-3.5: Highest quality score (67/50) with excellent cost efficiency
• claude-haiku-3: Best quality-per-dollar ratio among Anthropic models
• All Anthropic models provide reliable, consistent docstring generation

💡 Model Selection Guide

Use Case	Recommended Model	Reason
Development	gpt-5-nano	Best value, reliable quality
Testing/CI	gpt-5-nano	Cost-effective for automated runs
Production	gpt-5-mini	Maximum quality for end users
Enterprise	gpt-5-mini	Comprehensive documentation
Budget-Conscious	gpt-5-nano	Excellent quality at low cost
Privacy-First	Ollama codeqwen	Local processing
Anthropic Preference	claude-haiku-3.5	Highest quality score
Free Tier	claude-haiku-3	Good quality-per-dollar ratio

🚀 Quick Start Commands

# Use default (gpt-5-nano) - Best value
dococtopy fix . --rule DG101

# Use premium (gpt-5-mini) - Maximum quality
dococtopy fix . --rule DG101 --llm-model gpt-5-mini

# Use best Anthropic (claude-haiku-3.5) - Highest quality score
dococtopy fix . --rule DG101 --llm-provider anthropic --llm-model claude-haiku-3.5

# Use local (Ollama) - Privacy-first
dococtopy fix . --rule DG101 --llm-provider ollama --llm-model codeqwen:latest

📊 See detailed comparison results: docs/model-comparison/ - Compare actual generated docstrings side-by-side
📋 Quick summary: docs/model-comparison/SUMMARY.md - Decision matrix and recommendations

CLI Reference

dococtopy scan

Scan paths for documentation compliance issues.

dococtopy scan [PATHS...] [OPTIONS]

Options:
  --format {pretty,json,sarif,both}  Output format [default: pretty]
  --config PATH                      Config file path [default: pyproject.toml]
  --fail-level {error,warning,info}  Exit code threshold [default: error]
  --no-cache                        Disable caching
  --changed-only                    Only scan changed files
  --stats                           Show cache statistics
  --output-file PATH                Write output to file

dococtopy fix

Fix documentation issues using LLM assistance.

dococtopy fix [PATHS...] [OPTIONS]

Options:
  --dry-run                         Show changes without applying [default: False]
  --interactive                     Accept/reject each fix interactively
  --rule TEXT                       Comma-separated rule IDs to fix
  --max-changes INTEGER             Maximum number of changes
  --llm-provider {openai,anthropic,ollama}  LLM provider [default: openai]
  --llm-model TEXT                  LLM model to use [default: gpt-5-nano]
  --llm-base-url TEXT               Base URL for LLM provider (for Ollama, etc.)
  --config PATH                     Config file path

CI/CD Integration

Create .github/workflows/docstring-check.yml:

name: Docstring Compliance
on: [push, pull_request]

jobs:
  docstring-check:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: "3.11"
      - run: pip install dococtopy
      - run: dococtopy scan . --format json --output-file report.json --fail-level error
      - name: Upload report
        uses: actions/upload-artifact@v4
        with:
          name: docstring-report
          path: report.json

Architecture

DocOctopy is built with a modular, extensible architecture:

dococtopy/
├── cli/           # Command-line interface
├── core/          # Core engine, discovery, caching
├── adapters/      # Language-specific adapters
├── rules/         # Compliance rules and registry
├── remediation/   # LLM-powered fixing
└── reporters/     # Output formatters

Key Components

• Discovery Engine: Finds files using gitignore-style patterns
• Language Adapters: Parse code and extract symbols/docstrings
• Rule Engine: Applies compliance rules with configurable severity
• Remediation Engine: Uses DSPy for structured LLM interactions
• Caching System: Incremental scanning with fingerprint-based invalidation
• Interactive Reviewer: Handles interactive fix workflows with diff preview

Contributing

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

Development Setup

git clone https://github.com/CrazyBonze/DocOctopy.git
cd DocOctopy
uv sync --group dev
uv run pytest

Development Workflow

We use pre-commit hooks to ensure code quality and prevent CI failures:

# Install pre-commit hooks (one-time setup)
uv run task pre-commit:install

# Run pre-commit checks manually
uv run task pre-commit:run

# Or use the convenience script
./scripts/pre-commit.sh

Pre-commit checks include:

• Black: Code formatting
• isort: Import sorting
• MyPy: Type checking
• Pytest: Fast test suite

Available tasks:

uv run task format          # Format code
uv run task lint            # Run linting
uv run task test:fast       # Run fast tests
uv run task test:cov        # Run tests with coverage
uv run task ci              # Run full CI pipeline

Adding New Rules

1. Create rule class in src/dococtopy/rules/
2. Implement check() method
3. Register with register() function
4. Add tests in tests/unit/

Adding New Languages

1. Implement LanguageAdapter interface
2. Create symbol extraction logic
3. Add language-specific rules
4. Update discovery patterns

Roadmap

MVP (Current)

• ✅ Python docstring compliance checking
• ✅ Google-style validation rules
• ✅ LLM-powered remediation
• ✅ Multiple output formats
• ✅ Configuration system
• ✅ Caching and incremental scanning
• ✅ Interactive fix workflows
• ✅ File writing capabilities
• ✅ Advanced Google style rules (DG211-DG214)
• ✅ Context-specific rules (DG401-DG403)
• ✅ Content quality rules (DG303-DG304)
• ✅ Recommendation rules (DG215-DG216)
• ✅ Sophisticated complexity detection with cyclomatic complexity
• ✅ Multi-style docstring support (Google, Sphinx, NumPy, Facebook)
• ✅ Comprehensive test suite (811 tests, 79% coverage)
• ✅ Enhanced AST utilities with full parameter support

V1 (Next)

• 🔄 GitHub Action and pre-commit hooks
• 🔄 Playground UI for prompt experimentation
• 🔄 Additional Python rules (coverage thresholds, etc.)
• 🔄 Batch processing for large codebases

Future

• 📋 JavaScript/TypeScript support
• 📋 Go documentation checking
• 📋 Rust documentation checking
• 📋 Language server integration
• 📋 Advanced prompt optimization

License

MIT License - see LICENSE file for details.

Acknowledgments

• Built with DSPy for reliable LLM interactions
• Uses docstring-parser for Google-style parsing
• Powered by Typer for CLI interface
• Styled with Rich for beautiful output