Claude Pytorch Treehugger

🚀 PyTorch HUD API Python Library with MCP Support

A Python library for interacting with the PyTorch HUD API and MCP server, providing support for CI/CD data, job logs, and analysis.

🚀 Quick Start

This project offers the following tools for PyTorch's continuous integration/delivery (CI/CD) data analysis:

• Retrieve data on workflows, jobs, and test runs.
• Efficiently analyze large CI logs.
• Integrate ClickHouse queries for analysis.
• Provide resource utilization metrics.

📦 Installation

For Users

# Install from the GitHub repository
pip install git+https://github.com/izaitsevfb/claude-pytorch-treehugger.git

claude mcp add hud pytorch-hud

For Developers

# Install dependencies (if not installed via pip)
pip install -r requirements.txt

# Start the MCP server
python -m pytorch_hud

✨ Features

Data Access

• get_commit_summary: Retrieve basic commit information without jobs.
• get_job_summary: Get aggregated job status statistics.
• get_filtered_jobs: Filter jobs by status, workflow, or name.
• get_failure_details: Obtain detailed failure information for failed jobs.
• get_recent_commit_status: Get the status of recent commits and job statistics.

Log Analysis

• download_log_to_file: Download logs to local storage.
• extract_log_patterns: Find error, warning, and other information.
• extract_test_results: Parse test execution results.
• filter_log_sections: Extract specific log sections.
• search_logs: Search across multiple logs.

📚 Documentation

• CLAUDE.md: Detailed usage instructions, code style, and implementation notes.
• mcp-guide.md: General information about the MCP protocol.

💻 Usage Examples

For Developers

# Run tests
python -m unittest discover test

# Type checking
mypy -p pytorch_hud -p test

# Code review
ruff check pytorch_hud/ test/

📄 License

MIT