Numtracker should provide a helpful error message when the instrument session is bad

[EmsArnold]

Currently, if an instrument session is typed incorrectly, a very unhelpful error message is given. A few scenarios were tried, including:

• An incorrectly formatted session id (cm1234-1): RuntimeError: Numtracker error: [{'message': 'Authentication failed', 'locations': [{'line': 3, 'column': 5}], 'path': ['scan']}]
• A correctly formatted but non-existent session id (cm12345-1): RuntimeError: Numtracker error: [{'message': 'Authentication failed', 'locations': [{'line': 3, 'column': 5}], 'path': ['scan']}]
• An correctly formatted but non-existant session id with the correct proposal number and session number, but incorrect prefix (sm44191-3):
  • With no file-writing detector: A "successful" task run, with failure to write a nexus file - {"timestamp":"2026-06-15T08:14:56.496","level":"ERROR","thread":"pool-1-thread-4","logger":"uk.ac.diamond.daq.services.BlueskyNexusFileWriterService","message":"Unable to find File for ScanID 3b700b88-21dd-420d-89e2-281c2ba28d93","context":"default","beamline":"b01-1","application_name":"Document to Nexus Converter"}
  • With a file writing detector: A "failed" task run, with the file-writing detector throwing an error - <WatchableAsyncStatus, device: spectroscopy_detector, task: <coroutine object AsyncStatusBase.__init__.<locals>.wait_with_error_message at 0x7f8e7808d740>, errored: FileNotFoundError("Path /dls/b01-1/data/2026/sm44191-3/spool/ doesn't exist or not writable!")>

With respect to the first two errors, it would be good if some indication were given as to what was wrong - i.e. something along the lines of "Instrument session / visit was not recognised"

[EmsArnold]

With respect to the third point above (correctly formatted session ID with an incorrect prefix), as the prefix is relied upon later (for file-writing detectors, NeXus writing, and I'm assuming setting the associated session ID for Tiled) its correctness should be checked, and if it is incorrect a task should fail.

[tpoliaw]

The first two should be improved by recent blueapi and numtracker changes.

The third point is more difficult because I don't think the session code is in the data we're using for authz. It might be available in the new data bundle OPA. @NeilSmithDLS?

[NeilSmithDLS]

As you will know, the session reference is formatted as "-" There is nothing in UAS that enforces the format of the proposal number so it can be any number of digits, therefore cm1234-1 in the first bullet point is a valid format, as will cm123456-1 once/if diamond gets to 100,000 proposals (don't ask me why but there are also some proposals with negative proposal numbers, although these are very old). This means that bullet point 1 and bullet point 2 are actually the same problem, of the proposal/session not existing

[NeilSmithDLS]

regarding validation of the proposal category, yes this will be possible via OPA