fix(gemini): avoid duplicate model route for full api_base

[StatPan]

Summary

• avoid appending a duplicate /models/{model}:{endpoint} route when Gemini api_base already includes the full route
• handle api_base values that already point at /models/{model} prefixes without duplicating the model path
• add regression coverage for both full-route and full-model-prefix custom api_base handling

Validation

• uv run pytest tests/test_litellm/llms/vertex_ai/test_vertex_llm_base.py -k 'custom_proxy or full_route or api_base or credential_project_validation'
• uv run ruff check litellm/llms/vertex_ai/vertex_llm_base.py tests/test_litellm/llms/vertex_ai/test_vertex_llm_base.py

Context

This is the clean OSS-staging re-open of the same fix after identifying that the previous PR routing was wrong and the branch ancestry pulled in unrelated diff.

Related:

• Closes [Bug]: Gemini custom api_base with full model route gets duplicate path appended #26979
• Continues fix(gemini): prevent duplicate model path generation when api_base includes full route #24786
• [Bug]: Gemini models always 404 (GeminiException) - base_url gets doubled OpenHands/OpenHands#14028
• fix(llm): skip Gemini full endpoint URL as base_url to prevent doubled path 404 OpenHands/OpenHands#14029
• docs(sdk): add PostgreSQL-backed persistence guide and API reference OpenHands/docs#465
• Resolves Track upstream contribution: Gemini full api_base duplicate route fix StatPan/litellm#1

[greptile-apps]

Greptile Summary

This PR fixes a URL-construction bug in _check_custom_proxy where a Gemini api_base that already contained the full /models/{model}:{endpoint} route (or the model-only prefix) would have an extra segment appended. The fix adds two early-exit branches and a rstrip("/") improvement. Tests are well-targeted: unused imports are removed, a pre-existing missing assertion is added, and three new parametrized cases cover the new logic.

Confidence Score: 4/5

Safe to merge; the fix is minimal and correct for the documented cases, with one unhandled edge case left as a P2 suggestion.

Only P2 findings present. The three-branch guard correctly handles all cases described in the PR. One unhandled edge case (full-route api_base with a mismatched endpoint) could silently produce a bad URL but is very unlikely in practice and is no worse than the pre-fix behaviour.

No files require special attention.

Important Files Changed

Filename	Overview
litellm/llms/vertex_ai/vertex_llm_base.py	Adds three-way branch in _check_custom_proxy to avoid duplicating /models/{model}:{endpoint} when the Gemini api_base already includes the full route or the model prefix; also adds rstrip("/") to the base-URL case.
tests/test_litellm/llms/vertex_ai/test_vertex_llm_base.py	Removes unused imports, converts an incomplete print statement into a proper assertion, and adds a parametrized regression test for the three new api_base shapes.

_{Reviews (1): Last reviewed commit: "test(vertex): restore credential project..." | Re-trigger Greptile}

[greptile-apps]

If a user sets api_base to a full route that already contains a different endpoint than the one being requested (e.g. api_base ends with :generateContent but the call needs :streamGenerateContent), none of the first two branches match and the else branch appends /models/{model}:{endpoint} onto an already-fully-routed base, producing a malformed URL. A guard on the broader /models/{model}: pattern would prevent this from silently constructing a wrong URL.

Suggested change

	if api_base.endswith(f"/models/{model}:{endpoint}"):
	url = api_base
	elif api_base.endswith(f"/models/{model}"):
	url = f"{api_base}:{endpoint}"
	else:
	url = "{}/models/{}:{}".format(
	api_base.rstrip("/"), model, endpoint
	)
	if api_base.endswith(f"/models/{model}:{endpoint}"):
	url = api_base
	elif api_base.endswith(f"/models/{model}"):
	url = f"{api_base}:{endpoint}"
	elif f"/models/{model}:" in api_base:
	# api_base already contains a routed path for this model but
	# with a different endpoint – swap the endpoint rather than
	# appending a second /models/… segment.
	url = api_base[: api_base.rfind(f"/models/{model}:")] + f"/models/{model}:{endpoint}"
	else:
	url = "{}/models/{}:{}".format(
	api_base.rstrip("/"), model, endpoint
	)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[StatPan]

Quick follow-up: I closed the earlier routing attempts (#26980, #26982) so this PR is now the only active PR for the Gemini custom api_base fix.

This is the clean review path:

• base: litellm_oss_staging
• diff: 2 files (vertex_llm_base.py + focused tests)
• CI: green
• Codecov patch coverage: covered
• fixes [Bug]: Gemini custom api_base with full model route gets duplicate path appended #26979

The bug breaks deployments where api_base already includes the full Gemini model route because LiteLLM appends /models/{model}:{endpoint} a second time. Would appreciate a maintainer review when possible.

[oss-pr-review-agent-shin]

🤖 litellm-agent: This PR is currently BLOCKED from merge.

Score: 4/5 ❌

Why blocked:

• 1 unresolved reviewer concern (greptile) (unresolved_concern, -1 pts)

Details: Score docked for: 1 unresolved reviewer concern (greptile).

Fix the issues above and push an update — the bot will re-review automatically.

Note: This bot is still in beta and might not always work as expected. Please share any feedback via Slack.

[StatPan]

Hi maintainers, gentle ping on this bug fix.

This PR is now the only active PR for #26979 after closing the earlier routing attempts. It is a small 2-file change against litellm_oss_staging, CI is green, and it fixes Gemini custom api_base values that already include the full model route.

Could someone take a look when possible, or let me know if this should be routed differently?