Check for existing issues

• I have searched the existing issues and checked that my issue is not a duplicate.

What happened?

When using Gemini 2.5 Flash or Gemini 2.5 Pro on Vertex AI with search tools defined in litellm_params (e.g. enterpriseWebSearch, urlContext) in the config, AND the user sends additional function calling tools (e.g. via MCP), LiteLLM v1.82.x throws a 400 BadRequestError:

"Multiple tools are supported only when they are all search tools."

This is a regression from v1.81.0 where the same configuration worked correctly.

Root cause: The _merge_tools_from_deployment function introduced in v1.82 merges tools from litellm_params in the deployment config with tools from the user request. This results in multiple separate Tool objects being sent to Vertex AI (search tools + function declaration tools), which violates Vertex AI Gemini's constraint that multiple tool objects are only allowed when they are all search tools.

What did you expect to happen?

Function calling tools (from MCP or user request) should work alongside deployment-level search tools defined in litellm_params, as they did in v1.81.0. LiteLLM should either:

1. Not merge deployment-level search tools with request-level function declaration tools, OR
2. Properly combine them into a single tool object that Vertex AI accepts

Steps to Reproduce

1. Set up LiteLLM proxy config with Gemini 2.5 Flash on Vertex AI with search tools in litellm_params:

model_list:
  - model_name: gemini-2.5-flash
    litellm_params:
      model: vertex_ai/gemini-2.5-flash
      vertex_project: your-project
      vertex_location: us-central1
      tools: [{"enterpriseWebSearch": {}}, {"urlContext": {}}]

1. Send a chat completion request with additional function calling tools (e.g. from MCP):

import openai

client = openai.OpenAI(
    api_key="your-litellm-key",
    base_url="http://localhost:4000"
)

response = client.chat.completions.create(
    model="gemini-2.5-flash",
    messages=[{"role": "user", "content": "Search for something"}],
    tools=[
        {
            "type": "function",
            "function": {
                "name": "notion_search",
                "description": "Search Notion",
                "parameters": {"type": "object", "properties": {"query": {"type": "string"}}}
            }
        }
    ]
)

1. Observe 400 BadRequestError from Vertex AI: "Multiple tools are supported only when they are all search tools."

Relevant log output

litellm.BadRequestError: Vertex_ai_betaException BadRequestError - b'{
"error": {
"code": 400,
"message": "Multiple tools are supported only when they are all search tools.",
"status": "INVALID_ARGUMENT"
}
}'

What part of LiteLLM is this about?

Proxy

What LiteLLM version are you on ?

v1.82.rc.3 (regression from v1.81.0)

Twitter / LinkedIn details

No response