[Bug] Vertex AI: thinkingConfig.includeThoughts sent unconditionally, breaks non-thinking Gemini models (2.0-flash, 2.5-flash-lite, etc.)

[uncle-rennus]

Description

Describe the bug

When using @ai-sdk/google-vertex as the provider, OpenCode unconditionally adds thinkingConfig: { includeThoughts: true } to every request, regardless of whether the selected model supports thinking. This causes a Vertex AI 400 error for models that don't support it:

Unable to submit request because Thinking_config.include_thoughts is only enabled when thinking is enabled.

Root cause (exact file + lines)

In packages/opencode/src/provider/transform.ts, inside the options() export function:

if (input.model.api.npm === "@ai-sdk/google" || input.model.api.npm === "@ai-sdk/google-vertex") {
  result["thinkingConfig"] = {
    includeThoughts: true,
  }
  if (input.model.api.id.includes("gemini-3")) {
    result["thinkingConfig"]["thinkingLevel"] = "high"
  }
}

This block runs for every Gemini model on Google/Vertex, with no check for whether the model's capabilities.reasoning is true. So gemini-2.0-flash, gemini-2.5-flash-lite, and any other non-thinking model all get includeThoughts: true injected — which Vertex rejects.

Proposed fix

Gate it on capabilities.reasoning:

if (input.model.api.npm === "@ai-sdk/google" || input.model.api.npm === "@ai-sdk/google-vertex") {
  if (input.model.capabilities.reasoning) {
    result["thinkingConfig"] = {
      includeThoughts: true,
    }
    if (input.model.api.id.includes("gemini-3")) {
      result["thinkingConfig"]["thinkingLevel"] = "high"
    }
  }
}

Affected models on Vertex

• gemini-2.0-flash / gemini-2.0-flash-001
• gemini-2.5-flash-lite-preview-0514
• Any Gemini model without capabilities.reasoning = true

Working models (not affected)

• gemini-2.5-pro (has capabilities.reasoning, thinking works fine)

Note on smallOptions()

The smallOptions() function handles the google providerID correctly (sends thinkingBudget: 0 to suppress thinking), but this fix is never applied to Vertex because smallOptions() only checks model.providerID === "google", not @ai-sdk/google-vertex.

Plugins

N/A

OpenCode version

1.2.10

Steps to reproduce

1. Configure OpenCode with Google Vertex AI provider + service account (GOOGLE_APPLICATION_CREDENTIALS)
2. Select any non-thinking model: gemini-2.0-flash, gemini-2.5-flash-lite-preview-0514, etc.
3. Send any message
4. → Vertex returns 400: Thinking_config.include_thoughts is only enabled when thinking is enabled

Screenshot and/or share link

No response

Operating System

Windows 11

Terminal

Windows Terminal