Gemma 4 models not supported in ADK model registry #5237

Description

Gemma 4 models (e.g. gemma-4-31b-it) are available via the Gemini API but cannot be used through ADK because the model registry does not match them.

The Gemma class in google/adk/models/gemma_llm.py registers only r'gemma-3.*' as its supported model pattern. This means any Gemma 4 model name fails with:

ValueError: Model gemma-4-31b-it not found.

Similarly, the Gemma3Ollama subclass only registers r'ollama/gemma3.*'.

Steps to reproduce

1. Set gemma-4-31b-it as the model for an LlmAgent:

   agent = LlmAgent(model="gemma-4-31b-it", name="test")

2. Run the agent
3. ADK raises ValueError: Model gemma-4-31b-it not found. from registry.py:124

Expected behavior

Gemma 4 models should be resolved by the registry and handled by Gemma, just like Gemma 3 models are.

Environment

• google-adk 1.29.0 (also tested with 1.28.0)
• Python 3.14
• The model works fine when called directly via the REST API:

  curl "https://generativelanguage.googleapis.com/v1beta/models/gemma-4-31b-it:generateContent?key=$API_KEY" \
      -H 'Content-Type: application/json' \
      -X POST \
      -d '{"contents": [{"parts":[{"text": "Hello"}]}]}'

Root cause

In gemma_llm.py, Gemma.supported_models() returns:

return [
    r'gemma-3.*',
]

And Gemma3Ollama.supported_models() returns:

return [
    r'ollama/gemma3.*',
]

Neither pattern matches gemma-4-*.

Suggested fix

Broaden the regex to cover current and future Gemma versions:

# Gemma.supported_models()
return [
    r'gemma-\d+.*',
]

# Gemma3Ollama.supported_models() (consider renaming class to GemmaOllama)
return [
    r'ollama/gemma\d+.*',
]

Or, if Gemma 4 needs its own class, add a Gemma4 class with r'gemma-4.*' and register it alongside the existing Gemma class.

It would be nice to have Gemma 4 models supported soon in ADK.