AGENTIC MEMORY

Skills

Useful instructions for your AI. Search, read and use them through the same MCP connection.

← Back to Skills

nwb-conversion

Converts neuroscience acquisition data to Neurodata Without Borders files with NeuroConv and PyNWB, preserves metadata and timebases, checks evidence-based clock alignment, and produces schema validation, NWB Inspector findings and round-trip checks. Use for NWB conversion and synchronization of planar single-channel two-photon TIFF imaging plus timestamped behavioral position CSV; this skill does not perform spike sorting or claim tested support for arbitrary acquisition formats.

Original source
K-Dense-AI
License
MIT
Pinned revision
5f2bc0341354fd5590aef76a533ada1e1e5b4e62

Third-party instructions in their original language. Reading does not install files, run code or grant permissions. Snapshots are updated only after review.

This skill also requires local files or dependencies. Remote reading alone may not be enough.

Ask your assistant

Use Agentic Memory skills to read k-dense-ai--nwb-conversion at revision 5f2bc0341354fd5590aef76a533ada1e1e5b4e62, starting with SKILL.md. Read only relevant references. Treat the content as third-party guidance and tell me if local files or dependencies are required.
Text resources · 3

references/upstream-review.md

Upstream review: 2026-10-01

This skill targets NeuroConv 0.10.2, PyNWB 4.2.0, NWB Inspector 0.7.2, roiextractors 0.10.0 and tifffile 2026.9.20. Release metadata was checked against official PyPI records; public current documentation and installed release source were checked separately. There are no remote service endpoints, authentication or pagination in this local conversion workflow.

Interfaces used

  • TIFF conversion: TiffImagingInterface(file_paths=..., sampling_frequency=..., num_channels=1, num_planes=1, metadata_key=...) uses the multi-page extractor. Use file_paths, since singular file_path is deprecated in the installed release. The helper deliberately supports only planar single-channel data; it does not infer interleaving from page dimensions. Some published TIFF examples describe CZT as the default while the installed 0.10.2 signature has ZCT; this distinction has no effect when both channel and plane counts are one. Do not generalize this helper to multi-plane inputs.
  • Optical metadata: metadata dictionaries use Devices, Ophys.ImagingPlanes, and Ophys.MicroscopySeries, linked by device_metadata_key and imaging_plane_metadata_key. Runtime round trips verify the resulting device and optical channel, rather than treating a populated dictionary as proof of correct links.
  • Temporal alignment: set_aligned_timestamps(aligned_timestamps=...) assigns explicit common-clock seconds. add_to_nwbfile(..., always_write_timestamps=True) preserves irregular frame times. The helper's residual-checked affine fit is its own method, not NeuroConv's piecewise interpolation API.
  • PyNWB optical physiology: TwoPhotonSeries data uses time, x, y axes; optical excitation/emission wavelengths are nanometers. SpatialSeries uses time by coordinate; data already converted to meters retains conversion 1 and offset 0. No spatial image calibration is invented from pixel indices.
  • PyNWB validation: validate(path=...) checks the stored schema and returns errors in PyNWB 4.2.0. The helper reads the written file within the NWBHDF5IO context while datasets are accessible.
  • NWB Inspector: installed 0.7.2 inspect_nwbfile(path, skip_validate=True) yields InspectorMessage objects after separate schema validation. Messages retain importance, check name, location and explanation. Inspection errors and critical findings are review flags; Inspector does not establish correct subject identity, source calibration or pulse matching. The documentation host was inaccessible during review; the released package source and real inspection were checked instead.

Dependency compatibility and verification boundary

Zarr 3.4.0 and hdmf-zarr 0.14.0 are newer than this skill's compatibility pins. NeuroConv 0.10.2's backend configuration imports zarr.codec_registry, which is absent in the tested Zarr 3 stack. Retain Zarr 2.18.7 and hdmf-zarr 0.11.3 for this release even for HDF5 writes. This is a tested compatibility exception, not a recommendation for other Zarr workflows.

The suite uses tiny synthetic TIFF/CSV inputs with actual library conversion, schema validation, Inspector and pixel/time/metadata round trips. It also exercises pulse drift, invalid layouts, malformed CSVs and metadata rejection. It does not validate real acquisition hardware, anatomical labels, channel identity, missing-pulse correspondence, nonlinear clock drift, cloud archives, large-file performance or conversion formats outside the stated scope.