fix(proxy): hydrate wildcard discovery credentials (#28284) - CCI Run

[yuneng-berri]

• fix(proxy): hydrate wildcard discovery credentials

• fix(proxy): constrain wildcard credential hydration

Relevant issues

Linear ticket

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/test_litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Screenshots / Proof of Fix

Type

🆕 New Feature
🐛 Bug Fix
🧹 Refactoring
📖 Documentation
🚄 Infrastructure
✅ Test

Changes

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[greptile-apps]

Greptile Summary

This PR fixes wildcard discovery for team-scoped BYOK deployments by threading team_id through the wildcard expansion path and introducing _hydrate_litellm_credential_name to resolve stored credentials into LiteLLM_Params before provider-model discovery calls are made.

• _hydrate_litellm_credential_name in model_checks.py looks up the named credential from litellm.credential_list, copies only the fields present in CredentialLiteLLMParams (preserving any deployment-level overrides), then clears litellm_credential_name so callers don't re-resolve it.
• _get_wildcard_models and get_complete_model_list now accept and forward team_id to llm_router.get_model_list, which enables the router's existing team-pattern-router logic to surface team-scoped deployments during model enumeration.
• get_available_models_for_user in utils.py computes effective_team_id = team_id or user_api_key_dict.team_id and passes it into the model-list pipeline.

Confidence Score: 4/5

Safe to merge; the change only affects wildcard model listing, not the request routing or auth hot-path, and all new behavior is covered by mock-only unit tests.

The core fix is well-scoped and the router's should_include_deployment logic already handles global vs. team-specific deployments correctly, so existing wildcard users without team credentials are unaffected. The one observation is that litellm_credential_name is cleared unconditionally even when no credential fields are actually hydrated, which would silently drop the name without any diagnostic signal — worth a debug log but not blocking.

litellm/proxy/auth/model_checks.py — the new _hydrate_litellm_credential_name helper and its unconditional litellm_credential_name = None clear warrant a second look.

Important Files Changed

Filename	Overview
litellm/proxy/auth/model_checks.py	Adds _hydrate_litellm_credential_name to resolve stored credentials before wildcard expansion, and threads team_id through _get_wildcard_models/get_complete_model_list so team-scoped BYOK router deployments are matched correctly. Logic is sound; minor edge noted around unconditional credential-name clearing.
litellm/proxy/utils.py	Computes effective_team_id = team_id or user_api_key_dict.team_id and passes it to get_complete_model_list. Minimal, targeted change; correct precedence.
tests/test_litellm/proxy/auth/test_model_checks.py	Adds four focused mock-only tests covering credential hydration, field preservation, missing-credential passthrough, and the end-to-end get_available_models_for_user path. No real network calls.

_{Reviews (1): Last reviewed commit: "fix(proxy): hydrate wildcard discovery c..." | Re-trigger Greptile}

[greptile-apps]

litellm_credential_name is unconditionally cleared to None whenever the named credential exists, even if none of its keys match any CredentialLiteLLMParams field. In that edge case the credential lookup silently "consumes" the name and leaves every auth field still None, which would cause provider calls to fail without any indication that the credential name was resolved but produced no usable values. A guard or a debug log before clearing would make this easier to diagnose.

Suggested change

	litellm_params = litellm_params.model_copy()
	for key, value in credential_values.items():
	if (
	key in _CREDENTIAL_LITELLM_PARAM_FIELDS
	and getattr(litellm_params, key, None) is None
	):
	setattr(litellm_params, key, value)
	litellm_params.litellm_credential_name = None
	return litellm_params
	litellm_params = litellm_params.model_copy()
	hydrated_any = False
	for key, value in credential_values.items():
	if (
	key in _CREDENTIAL_LITELLM_PARAM_FIELDS
	and getattr(litellm_params, key, None) is None
	):
	setattr(litellm_params, key, value)
	hydrated_any = True
	if not hydrated_any:
	verbose_proxy_logger.debug(
	"litellm_credential_name=%s resolved but no matching CredentialLiteLLMParams fields found",
	litellm_params.litellm_credential_name,
	)
	litellm_params.litellm_credential_name = None
	return litellm_params

[veria-ai]

PR overview

Low: Unvalidated team_id can steer wildcard discovery

This PR hydrates stored credentials for wildcard model discovery and threads team_id into router lookups. The new effective team selection trusts a query-provided team_id even on paths where membership validation is skipped.

Security review

• 1 new security issue(s) were flagged in the latest review.
• 1 issue(s) remain open on this pull request.

Risk: 4/10

[veria-ai]

Low: Unvalidated team_id controls wildcard expansion

When the DB/cache helpers are not initialized, the membership check above is skipped, but this line still trusts the query-string team_id. A caller can request /v1/models?team_id=<other-team> and have wildcard discovery use another team's router deployment/credential, exposing that team's discovered model IDs. Only pass the query team into get_complete_model_list after validate_membership has run; otherwise fall back to user_api_key_dict.team_id.