runopsy 0.1.0

Find where an AI agent run started going wrong, not just where it stopped

runopsy

Find where an AI agent run started going wrong — not just where it stopped.

uv tool install runopsy        # or: pipx install runopsy
runopsy --help

This is the meta-distribution: it installs the CLI, the analysis engine, the collector, replay, the runtime adapters and the local web view. Everything works offline with no provider key.

runopsy record -s "make" -s "pytest"   # wrap any pipeline
runopsy diagnose latest                # where it started going wrong
runopsy ui                             # the timeline and failure map

Extras:

pip install "runopsy[inspect]"   # read Inspect AI eval logs

If you are writing code against Runopsy rather than using the CLI, depend on the piece you need — runopsy-core is the framework-agnostic engine and pulls in nothing but Pydantic.

Full documentation: https://github.com/vahit19/runopsy