Fix run_extraction TypeError in augment/run CLI commands

[pham-tuan-binh]

Problem

livekit-wakeword augment and livekit-wakeword run crash after augmentation in v0.2.2:

TypeError: run_extraction() missing 1 required positional argument: 'sess_options'

#73 added sess_options to run_extraction as a required argument, but the two CLI callers (cli.py:292, cli.py:398) still call run_extraction(config). CI didn't catch it because no test goes through run_extraction and CI doesn't run mypy.

Fix

• Default sess_options to None in run_extraction, matching WakeWordModel.__init__. The mel/embedding wrappers already handle None via session_options().
• Add a regression test that calls run_extraction(config) without sess_options. It fails with the TypeError before this change and passes after.

Testing

• uv run pytest: 74 passed, 2 skipped
• Verified on an RTX 5090 box: resumed a hey_reachy pipeline that had crashed at step 3, and feature extraction now runs.

[jvzelst-tech]

Thanks for this fix. Flagging that it isn't in a published release yet: PyPI's latest is still 0.3.0, and its attestation shows it was built from 69324ba — the commit immediately before fb92cb3 merged here. So pip install livekit-wakeword still hits this.

It's an expensive one to hit, because run fails at "Step 3/6: Extract features" after generation completes. On an M2 that was ~12 hours of synthesis (the adversarial negatives alone took 7.5h) before the TypeError surfaced.

Would a patch release including fb92cb3 be possible? In the meantime, anyone stuck can resume without re-generating by calling the remaining stages directly:

from livekit.wakeword.data.features import run_extraction
from livekit.wakeword.session import session_options
run_extraction(config, session_options())