33 lines (27 loc) · 1.22 KB

Instructions for AI Agents

Overview

NDR-python is a faithful Python port of NDR-matlab (Neuroscience Data Reader).

Architecture

Lead-Follow: MATLAB is the source of truth. Python mirrors it exactly.
Bridge Contract: Each sub-package has an ndr_matlab_python_bridge.yaml defining the function names, arguments, and return types.
Naming: Preserve MATLAB names exactly. Use readchannels_epochsamples, not read_channels_epoch_samples.

Key Classes

ndr.reader.base — Abstract base class. All readers inherit from this.
ndr.reader (wrapper) — High-level interface that delegates to a base reader.
ndr.reader.intan_rhd, ndr.reader.axon_abf, etc. — Format-specific readers.

Workflow

Check the bridge YAML in the target package.
If the function is missing, add it based on the MATLAB source.
Record the MATLAB git hash in matlab_last_sync_hash.
Implement the Python code.
Run black and ruff check --fix before committing.
Run pytest to verify.

Testing

Unit tests: pytest tests/
Symmetry tests: pytest tests/symmetry/ (excluded from default run)

Environment

Python 3.10+
NumPy for all numerical data
Pydantic for input validation (@validate_call)