Drop unused robot_name from TableRow (DB has only embodiment)

[ElmoPA]

robot_name was declared in TableRow but app.episodes has only the embodiment column.
The strict column check in episode_hash_to_table_row raised on it, crashing any
conversion that reaches it (e.g. force re-converts). It was never populated as a real
column; add_raw_data_to_table already omits it. Removed the field and the only
constructor still passing it (external/scale/scripts/sfs_zarr_pipeline.py).

Co-Authored-By: Claude Opus 4.8 (1M context) noreply@anthropic.com

[ElmoPA]

• Drop unused robot_name from TableRow (DB has only embodiment) #508 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[github-actions]

Claude Code Review

Summary

Removes the unused robot_name field from TableRow since the SQL schema (app.episodes) only has an embodiment column, and the strict column check in episode_hash_to_table_row was crashing on it.

Key concerns

1. Verify no other callers pass robot_name. The PR description claims the Scale pipeline was the "only constructor still passing it," but a repo-wide check should confirm — any other call site (e.g., Aria/Eva/Mecka conversion scripts, test fixtures, notebooks under egomimic/scripts/) would now hit a TypeError: unexpected keyword argument. Worth grepping for robot_name= and TableRow( before merging.

2. Historical data with robot_name in serialized form. If TableRow is ever pickled/JSON-serialized to disk or cached (e.g., in dataset manifests), older artifacts containing robot_name could fail to deserialize. Likely a non-issue since it was never a real column, but worth a quick check.

3. Embodiment string mismatch. The removed Scale row used embodiment="scale" but robot_name="scale_bimanual". Per conventions, embodiment strings should match the canonical enum. If the correct embodiment is actually scale_bimanual, this PR silently drops that information rather than fixing it. Please confirm "scale" (not "scale_bimanual") is the intended embodiment for Scale data — otherwise existing Scale rows in the DB may be mislabeled and this PR makes that harder to notice later.

Suggestions

• Run rg "robot_name" egomimic/ external/ tests/ to confirm full removal.
• Add a brief regression test (or extend an existing one) that round-trips a TableRow through episode_hash_to_table_row for a force re-convert path — that's the exact bug being fixed and a test prevents reintroduction.
• Clarify in the commit message whether embodiment="scale" is correct, or open a follow-up to reconcile Scale's embodiment enum.

Verdict: Approve (with comment)

Small, well-scoped cleanup that fixes a real crash. Please verify the grep and the Scale embodiment value before merging.

---

Reviewed by Claude · Review workflow