refactor(intrinsics): requirement_check migration — loud schema-mismatch, no more silent False (Epic #929 Phase 1) #1138

Parent epic: #929
Design proposal: PR #1080
Phase: 1 (Wave 3 — start after 1.A merges)
Depends on: 1.A (and 0.2 for catalogue deduplication background)
Independent of: 1.B, 1.D, 2.1

⚠️ Pre-flight dependency: PR #1078 must be merged before starting this issue. It adds canned I/O test data for requirement-check and uncertainty formatters that this issue's tests build on.

Problem

requirement_check and requirement_check_to_bool currently route through the old API. PR #1008 is the canonical evidence of silent schema drift: the output changed from {"requirement_likelihood": 0.9} to {"requirement_check": {"score": 0.9}} upstream, and requirement_check_to_bool returned False for every call until someone noticed.

Current state (verified main 2026-05-21): requirement_check_to_bool uses the post-#1008 schema req_dict["requirement_check"]["score"] and returns False with a warning on schema mismatch — still a silent failure mode.

Agreed design

Follows the per-helper-file migration template from PR #1080. File-specific deltas:

• Output contract: post-fix: move test_huggingface.py to granite4.1; and small rag intrinsic … #1008 shape {"requirement_check": {"score": float}}. Implementer confirms against current weights.
• Schema-mismatch is loud here. This helper is the design's worked example — callers must see AdapterSchemaMismatchError immediately, not silently-wrong booleans.
• requirement_check_to_bool stops returning silent False on parse failure — it propagates AdapterSchemaMismatchError. This is an intentional breaking change.
• Catalogue deduplication: by the time this issue is worked, 0.2 will have collapsed the requirement_check/requirement-check duplicate. This issue resolves whichever canonical entry the role-based lookup (from 1.A) surfaces.

Scope

• mellea/stdlib/components/intrinsic/ — wherever requirement_check lives
• mellea/stdlib/requirements/requirement.py — requirement_check_to_bool
• Tests covering both functions

Out of scope

Other helpers (1.B, 1.D), auto-bumping catalogue revision, catalogue deduplication (0.2).

Acceptance criteria

• requirement_check(...) constructs via new Adapter types (no IntrinsicAdapter(...))
• requirement_check(...) accepts model_options: dict | None = None
• requirement_check has a declared output contract in its docstring
• requirement_check_to_bool(...) returns bool on happy path
• requirement_check_to_bool(...) propagates AdapterSchemaMismatchError on schema-break — NOT silent False
• Synthetic output {"requirement_check": {"score": 0.9}} parses correctly
• Synthetic output {"requirement_likelihood": 0.9} (pre-fix: move test_huggingface.py to granite4.1; and small rag intrinsic … #1008 shape) raises AdapterSchemaMismatchError
• Existing happy-path tests pass
• ruff format, ruff check, mypy clean

Test plan

Common per-helper tests (see template in 929-issue-plan.md). Plus:

• test_requirement_check_to_bool_raises_on_schema_mismatch — replaces silent False with explicit AdapterSchemaMismatchError (regression test for fix: move test_huggingface.py to granite4.1; and small rag intrinsic … #1008 behaviour; name it after fix: move test_huggingface.py to granite4.1; and small rag intrinsic … #1008)
• test_pre_1008_schema_raises — specifically: {"requirement_likelihood": 0.9} input raises

Breaking changes

• requirement_check_to_bool no longer returns silent False on parse failure — now raises AdapterSchemaMismatchError. Callers that wrap this in try/except treating failure as "requirement not met" must update. Surface in changelog.

References

• PR docs: design proposal for adapter lifecycle (#929) — review channel, do not merge #1080 §14.1 Easier RCA for failed requirements #2 (silent schema drift worked example)
• PR fix: move test_huggingface.py to granite4.1; and small rag intrinsic … #1008 — the original schema flip
• Jake req 4
• update intrinsic tests #1029 — open (close when PR fix: intrinsic tests and add some safeguards for future adapters changes  #1078 merges; fix: intrinsic tests and add some safeguards for future adapters changes  #1078 is the pre-flight for this issue)