Support Gemini thinking budget param & other thinking params

[micke754]

Context

Hi Team,

I realized that there was no way to specify the ThinkingBudget parameter when making requests to Gemini models. This limited the ability to control model reasoning steps. For Flash and Flash-Lite, this means I can't enable it either.

Referencing the Gemini API and looking at the curl example, we should an additional optional alongside specifying token size, temp etc. Below is the example:

curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-pro:generateContent?key=$GOOGLE_API_KEY" \
-H 'Content-Type: application/json' \
-X POST \
-d '{
  "contents": [
    {
      "parts": [
        {
          "text": "Provide a list of 3 famous physicists and their key contributions"
        }
      ]
    }
  ],
  "generationConfig": {
    "thinkingConfig": {
          "thinkingBudget": 1024
          # Thinking off:
          # "thinkingBudget": 0
          # Turn on dynamic thinking:
          # "thinkingBudget": -1
    }
  }
}'

This is a bit longer but the link shows an example of how we can include the IncludeThoughts flag as well: https://ai.google.dev/gemini-api/docs/thinking#go_1

Proposal

• Support for ThinkingBudget: Add support in the codebase to allow users to specify ThinkingBudget for Gemini models via configuration or CLI.
• Support for IncludeThoughts:* Add support for the the IncludeThoughts flag to make troubleshooting if thinking is working as intended.

Rationale

These changes will enable advanced configuration for Gemini users.

Request for Feedback

Is there interest or concern in supporting the Gemini ThinkingBudget parameter? Is this functionality already available and I've somehow missed it? If so apologies for the hassle.

I've also submitted a PR with some proposed changes.

Would love to hear your feedback :)