fix: resolve NP2004 probe model correctly from SpikeGLX metadata

[tabedzki]

🤖 Human guided, AI assisted PR (using this skill). AI text below. 🤖

Summary

A user has a Neuropixels probe with imDatPrb_pn=NP2004 whose SpikeGLX metadata reports imDatPrb_type=2003 — a numeric type code not covered by the existing 1100/21/24 branches in SpikeGLXMeta. The fallback branch stringified the raw type code (producing "2003"), which never matches any registered ProbeType. This caused ingestion to fail with either:

• NotImplementedError: Processing for neuropixels probe model 2003 not yet implemented, or
• datajoint.errors.IntegrityError: Cannot add or update a child row: a foreign key constraint fails ... FOREIGN KEY (probe_type) REFERENCES #probe_type (probe_type)

depending on which insertion path is hit.

Changes

• element_array_ephys/readers/spikeglx.py:
  • Fall back to self.probe_PN (the part number, e.g. "NP2004") instead of str(probe_model) when imDatPrb_type doesn't match a known code. probe_geometry.M already has geometry defined for NP2004.
  • Pull in upstream's imMaxInt-based IMAX lookup (SpikeGLX.get_channel_bit_volts), reading gain/max-int directly from the meta file when present, falling back to the IMAX dict only when absent. Without this, IMAX["NP2004"] would KeyError since IMAX is only keyed by the four generic probe model names, not part numbers.
  • imDatPrb_type default changed from 1 to 0, matching upstream, so meta files missing this field correctly enter the < 1 branch.
• element_array_ephys/probe.py:
  • Register "NP2004" in create_neuropixels_probe_types() so the corresponding ProbeType row is created automatically on schema activation, rather than requiring a manual probe.create_neuropixels_probe("NP2004") call.

These changes were verified against this exact metadata:

imDatPrb_pn=NP2004
imDatPrb_type=2003

which now correctly resolves probe_model to "NP2004".

Test plan

• Re-run ingestion against a session containing the NP2004 probe metadata above and confirm ProbeInsertion/EphysRecording populate without error
• Confirm existing probe types (1.0 3A/3B, UHD, 2.0 SS/MS) still resolve and ingest correctly (regression check on the IMAX/probe_model changes)

Assisted-by: ClaudeCode:claude-sonnet-5