4.1.0

• Allow web socket messages of any size
• Improve reporting of more types of errors, including cachemachine errors
• Only number code cells in error reporting of notebooks
• Increase user token expiration to one year so that it stops expiring during runs
• Make default spawn timeout more closely match JupyterHub's
• Stop using the FastAPI Docker image as a base and build our own
• Drop support for Python 3.8

4.0.1

• Use the correct token to make cachemachine calls

4.0.0

• Query cachemachine for images rather than requiring configuring all the spawner form fields
• Allow testing the latest weekly by asking cachemachine what it is
• Report the image name in error messages on Slack
• Truncate errors and code if they are over 3000 characters, since Slack has a limit on block size

3.0.0

• Overhaul TAP testing, now called TAPQueryRunner, and update the queries it runs
• Add an API for a summary of flocks
• Post a periodic status summary to Slack
• Log additional information when lab deletion fails

2.4.0

• Add support for setting the working directory when running a notebook
• Add support for getting the node on which a notebook is running and reporting it in alerts
• Catch and translate the poor aiohttp exception from a closed WebSocket
• Improve the way we determine what notebooks to run
• Report the cell of the notebook that was being executed when an alert fires
• Simplify the timing data that's gathered and reported
• Add divider after unknown exception reports to make them easier to read
• Update dependencies

2.3.0

• Use the Jupyter Hub progress and user APIs instead of interpreting HTML UIs
• Fix alert reporting for unknown exceptions
• Handle WebSocket exceptions properly
• Increase delete timeout
• Re-add a settle time for the lab after creation
• Update dependencies

2.2.0

• Use the Slack block API to format the alert messages
• Accumulate alert information from the running timer
• Fix display of the lab timeout error message

2.1.1

• Overhaul how lab status detection is done
• Wait for lab deletion before continuing
• Clean up logging
• Update dependencies

2.1.0

• Add proper shutdown support for monkeys using a queue to coordinate
• Refactor the business classes more to reuse more code and better document overrides
• Make shutting down the lab between notebook executions optional
• Use max_interations for NotebookRunner as well, instead of its own config variable
• Lift more infrastructure into the base business class
• Reauthenticate to the lab before deleting a session
• Poll the hub URL instead of the pending spawn URL to see if the lab is up since the latter sometimes 503s
• Make the default JupyterPythonLoop code produce cleaner logs
• Fix desynchronization of the web socket that caused the results of code execution to be lost
• Remove the periodic hub reauth code (should now be unnecessary)
• Wait for the lab to shut down after deletion
• Be more precise when checking if the lab is running

2.0.0

• Refactor to use FastAPI instead of the aiohttp web application framework
• Redo the API to treat a flock of monkeys as the top-level object
• Use a JupyterLab session instead of a bare kernel to run code