Track upstream contribution: Gemini full api_base duplicate route fix

[StatPan]

Summary

Track the upstream LiteLLM contribution that fixes duplicate Gemini model-route construction when a custom api_base already includes the full model route.

Upstream PR

• PR: fix(gemini): avoid duplicate model route for full api_base BerriAI/litellm#26983
• URL: fix(gemini): avoid duplicate model route for full api_base BerriAI/litellm#26983
• Title: fix(gemini): avoid duplicate model route for full api_base
• Base: litellm_oss_staging
• Branch: fix/gemini-full-api-base-duplicate-path-clean

Upstream Issue

• Issue: [Bug]: Gemini custom api_base with full model route gets duplicate path appended BerriAI/litellm#26979
• URL: [Bug]: Gemini custom api_base with full model route gets duplicate path appended BerriAI/litellm#26979

Scope

This issue tracks the upstream contribution only. It does not track downstream adoption or release timing.

The upstream PR covers:

• Avoid appending duplicate /models/{model}:{endpoint} when Gemini api_base already includes the full route
• Handle custom api_base values ending at /models/{model}
• Add focused regression tests in test_vertex_llm_base.py

Current upstream status

• State: open
• Mergeability: mergeable
• CI: passing as of 2026-05-14
• Earlier routing attempts closed as superseded: fix(gemini): avoid duplicate model route for full api_base BerriAI/litellm#26980, fix(gemini): avoid duplicate model route for full api_base BerriAI/litellm#26982

Close condition

Close this issue when BerriAI#26983 is merged, rejected, or explicitly superseded. Merge is sufficient for this contribution-tracking issue.