Z.AI GLM-4.7 cannot read images from Read tool (model supports vision) #13310

Bug Description

GLM-4.7 is a multimodal model with vision support, but when using the Read tool to read image files with Z.AI (zai-coding-plan), the model says "I cannot read the image".

Expected Behavior

The model should be able to read and analyze image content, similar to how Claude, GPT-4, and other vision-capable models work.

Actual Behavior

When asking GLM-4.7 to read an image file using the Read tool:

• Error message: "Cannot read "/path/to/image.jpg" (this model does not support image input). Inform the user."
• Pasting images directly in messages works fine
• GLM-4.7 itself supports vision (can read images in other contexts)

Root Cause

The `supportsMediaInToolResults` function in `packages/opencode/src/session/message-v2.ts` doesn't include `@ai-sdk/openai-compatible` (used by Z.AI).

Current supported SDKs:

• `@ai-sdk/anthropic`
• `@ai-sdk/openai`
• `@ai-sdk/amazon-bedrock`
• `@ai-sdk/google-vertex/anthropic`
• `@ai-sdk/google` (Gemini-3 only)

Suggested Fix

Add `@ai-sdk/openai-compatible` to the `supportsMediaInToolResults` list:

```typescript
if (model.api.npm === "@ai-sdk/openai-compatible") {
// Z.AI GLM models support vision
return true
}
```

Related Issues

• Similar issue was fixed for Kimi K2.5 in PR fix(opencode): Fixes image reading with OpenAI-compatible providers like Kimi K2.5. #11323
• Z.AI provider documentation: https://opencode.ai/docs/providers#zai

Environment

• OpenCode version: 1.1.60
• Provider: Z.AI (GLM Coding Plan)
• Model: GLM-4.7