refactor(taxonomy): pin the CLI registry == schema both ways; drop dead instance_segmentation (#1005)

[LukasWodka]

Summary

The task taxonomy lives in 5 hand-synced copies across 3 repos (#1005). The only actual divergence is instance_segmentation — in the CLI registry (16th entry) but deliberately absent from the ingest.v1 schema and the ingestor registry.

The "decide first" question (planned vs dead) is settled by the ingestor:instance_segmentation is dead — it "briefly shipped with no validators and no file transfer, so configs half-ingested (DB rows + API records, zero files staged)" and was removed (data-ingestors constants.py, #240/#99). So this drops it from the CLI rather than keeping a misleading "not implemented" placeholder.

What changed (CLI third of #1005)

• Remove instance_segmentation from categoryRegistry — registry is now == the schema enum (15 == 15) — and from the two tests + code comments that named it. It now hits the "unrecognized category" gate like any other non-category (still exit 2).
• Add TestRegistryWithinSchema — the reverse of the existing TestRegistryCoversSchemaCategories (schema ⊆ registry). Together they pin registry == schema both ways, so the registry can neither fall behind the schema (a valid --category wrongly rejected as "unrecognized" — the token_classification RC drift) nor carry an extra the ingestor won't accept (the instance_segmentation half-ingest class).
• A future known-but-unschema'd placeholder must be declared in the new registryAliases allow-list (empty today) — an intentional superset is explicit, never silent. Exactly the #1005 proposal. Factored the schema-enum parse into a shared helper.
• Mutation-verified the new gate bites: re-adding instance_segmentation fails TestRegistryWithinSchema with an actionable message.

Scope

• data-ingestors already pins registry == schema (test_category_congruence.py + test_schema_validation). ✔ Nothing needed.
• backend copies (UserDataSet.CATEGORY_CHOICES + global_meta/constants.py) still carry instance_segmentation and aren't tied to the schema — a separate PR: it needs a choices migration, and global_meta is slated for deletion. #1005 stays open until that lands.

Type

Tech-debt / correctness · cli · RFC-0002 foundation. Part of backend#1005 (ties RFC-0002 §9 / cli#174). Part of the data-ingest epic backend#1008.

---

Note

Low Risk
Category taxonomy and validation messaging only; no auth, staging, or submit path changes. Slight behavior change for the removed dead category string.

Overview
Aligns the CLI category registry with the embedded ingest.v1 schema enum by removing dead instance_segmentation (not in schema/ingestor; previously caused half-ingest risk) and tightening parity tests.

Registry & behavior:instance_segmentation is gone from categoryRegistry and related docs/comments. --category=instance_segmentation now follows the unrecognized category path (exit 2), not the known-but-unsupported message. README roadmap/status no longer lists it as a future CLI task.

Tests: Adds TestRegistryWithinSchema (registry ⊆ schema, with an explicit empty registryAliases allow-list) alongside the existing schema ⊆ registry test, so registry == schema both ways. Shared schemaCategoryEnum helper reads the embedded schema; hand-maintained expected lists and ingest gate tests are updated accordingly.

^{Reviewed by Cursor Bugbot for commit 6e40580. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 6e40580. Configure here.}

[saadqbal]

Clean, well-reasoned PR 👍 Verified the core invariant directly against the embedded schema — enum is exactly 15, instance_segmentation absent, registry now matches both ways, so all three parity tests hold. The behavior change stays exit-2 (just a better message).

Only nit, non-blocking: the RFC (docs/rfcs/0002) still frames planned-vs-dead for instance_segmentation as an open question and calls it "registry-only, unchecked" — this settles it. A one-liner "resolved in #1005" would keep it from reading stale.