fix(models): Prefix digit-leading upstream model ids

[matthewgrossman]

What

When the models controller reconciler discovers models from an external provider, it normalizes each upstream id to an internal entity name that satisfies the entity store's RFC 1035-style NAME_PATTERN (leading [a-z], length 2-63, only [a-z0-9-]). Upstream catalogs occasionally publish digit-leading ids — e.g. NVIDIA Build's 01-ai/yi-large — which the normalizer rejects because 01-ai-yi-large doesn't start with a letter. The reconciler then logs Skipped 1 model(s) with invalid names ... on every cycle (~every 5s) and the model never becomes routable.

This PR teaches normalize_model_entity_name to prepend an internal m- prefix when the cleaned form starts with a digit, so digit-leading upstream ids become valid entity names instead of being skipped.

Why this approach

• Internal-only, not user-facing. The normalized string is the entity store's model_entity_id. The served_model_name mapping still preserves the original id (provider_reconciler.py:707), so the IGW continues to forward 01-ai/yi-large (the original) to the backend. End users keep calling the model exactly as the upstream advertises it.
• Minimal & deterministic. A constant m- prefix is mechanical, idempotent (m-01-ai-yi-large → m-01-ai-yi-large), and preserves the source id verbatim after the prefix, so the entity name still reads like the upstream id.
• Length-safe. The prefix is applied before the existing 63-char truncate-with-hash step, so long digit-leading names still produce length-bounded results via the existing machinery.

Considered and rejected: provider-derived prefixes (collide with the existing workspace/name convention, lengthen names), spelling out leading digits (lossy/weird), dropping leading digits (collisions across ids that differ only in the digit segment), hash prefixes (ugly, inconsistent with the existing policy of hashing only on overflow).

Verification

• New positive test cases for 01-ai/yi-large, 9model, 123, 2pac/rapper-model.
• New dedicated tests for idempotence and the digit-leading + 63-char-overflow interaction.
• Updated test_entity_name_from_discovered_model_invalid_raises and added test_entity_name_from_discovered_model_digit_leading_gets_prefix so the reconciler's caller-side contract is also pinned.
• Full services/core/models/tests/unit suite: 992 passed, 1 skipped.
• ruff check / ruff format --check clean. ty clean for the change (5 pre-existing diagnostics in unrelated functions in the same file).

Tracking

• Linear: AIRCORE-678
• NVBug: 6215677

Signed-off-by: Matthew Grossman mgrossman@nvidia.com

Summary by CodeRabbit

• Bug Fixes

  • Improved normalization of model identifiers that start with digits to ensure proper system compatibility.
  • Updated validation logic to handle digit-leading model IDs with automatic prefix handling.

• Tests

  • Added test coverage for digit-leading model identifier normalization and edge cases.

[coderabbitai]

📝 Walkthrough

Walkthrough

The PR updates model identifier normalization to handle digit-leading IDs by prefixing them with m- before truncation. Implementation, unit tests, and integration tests all reflect this new behavior; digit-leading identifiers are no longer treated as invalid but normalized with the internal prefix while preserving the original as a user-facing handle.

Changes

Digit-leading Normalization with m- Prefix

Layer / File(s)	Summary
Normalization logic and documentation services/core/models/src/nmp/core/models/app/utils.py	normalize_model_entity_name adds m- prefix when normalized output starts with a digit, applied before truncation/validation. Docstring updated with new behavior description and digit-leading example ("01-ai/yi-large" → "m-01-ai-yi-large").
Unit tests for digit-leading normalization services/core/models/tests/unit/common/test_utils.py	Parametrized test matrix extended with digit-leading inputs now expecting m- prefix in normalized output. Invalid-input test matrix adjusted to remove digit-leading cases. New test cases verify idempotency and correct truncation behavior for long digit-leading identifiers with hash suffix.
Integration tests for discovered models services/core/models/tests/unit/controllers/test_provider_reconciler.py	Invalid-entity-name test narrowed to single-character ID case. New test added verifying digit-leading discovered model IDs (e.g., "01-ai/yi-large", "123") receive m- prefix in resulting entity names.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	Title directly describes the main change: handling digit-leading upstream model IDs by adding m- prefix, which is the core fix across documentation, implementation, and tests.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix-yi-large-logs/mgrossman

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

Suite	Lines Covered	Line Rate	Branch Rate
Unit Tests	18245/24195	75.4%	61.8%
Integration Tests	11665/22977	50.8%	25.9%

[mckornfield]

lmao jinx #67

[mckornfield]

m- is probably better than md- and mine has lint failures