Fix get_model_from_request() to extract model ID from Vertex AI passthrough URLs

[krisxia0506]

Title

Fix get_model_from_request() to extract model ID from Vertex AI passthrough URLs

Relevant issues

Pre-Submission checklist

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

• I have Added testing in the tests/litellm/ directory, Adding at least 1 test is a hard requirement - see details
• I have added a screenshot of my new test passing locally
• 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

Type

🐛 Bug Fix

Changes

Summary

This PR enhances the model extraction logic in the proxy authentication module to properly handle Vertex AI passthrough routes. Previously, the get_model_from_request() function could not extract model IDs from Vertex AI passthrough URLs, which follows a distinct pattern compared to OpenAI deployments.

Problem

Vertex AI passthrough endpoints have a URL structure like:

/vertex_ai/v1/projects/{project}/locations/{location}/publishers/google/models/{model_id}:{method}

The existing model extraction logic only handled OpenAI deployments (/openai/deployments/{model}/), leaving Vertex AI models unidentified.

Solution

Added regex-based model extraction for Vertex AI routes:

• Pattern: /vertex_ai/.../models/{model_id}:*
• Extracts model ID using the regex pattern: /models/([^/:]+)
• Handles both versioned endpoints (v1, v1beta1) and different Google Cloud regions
• Maintains precedence: request body model > extracted from URL
• Does not interfere with existing OpenAI deployment extraction

Testing

Added comprehensive test coverage with 7 parametrized test cases in test_get_model_from_request_vertex_ai_passthrough():

1. ✅ Vertex AI v1 endpoint with generateContent method
2. ✅ Vertex AI v1beta1 endpoint with streamGenerateContent method
3. ✅ Different region locations (us-central1, asia-southeast1)
4. ✅ Model extraction without method suffix (no colon)
5. ✅ Request body model takes precedence over URL pattern
6. ✅ Non-Vertex routes are not affected
7. ✅ Azure deployment pattern still works as expected

Files Changed

• litellm/proxy/auth/auth_utils.py: Added Vertex AI model extraction logic
• tests/local_testing/test_auth_utils.py: Added parametrized test cases

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Review	Updated (UTC)
litellm	Ready	Preview, Comment	Dec 17, 2025 3:35am