This PyTorch HUD API implementation provides a Python library and MCP server for accessing PyTorch's CI/CD analytics data, enabling developers to investigate build failures and trunk health issues. Built with FastMCP, it offers tools for retrieving workflow and job information, analyzing large log files efficiently, executing ClickHouse queries against CI metrics, and monitoring resource utilization. The server exposes both synchronous and asynchronous functions through a standardized interface, making it valuable for PyTorch contributors debugging CI failures, investigating test flakiness, or analyzing performance trends across the CI infrastructure.
No reviews yet. Be the first to review!
Sign in to join the conversation
Retrieve basic commit info without jobs.
Get aggregated job status counts.
Retrieve jobs with filtering by status, workflow, or name.
Obtain detailed failure information for failed jobs.
Fetch status for recent commits along with job statistics.
Download logs to local storage.
Find errors, warnings, and other patterns in logs.
Parse and retrieve test execution results.
Extract specific sections from logs.
Search across multiple logs for specific entries.
