Bug: qwen3.6-plus model in alibaba-coding-plan-cn provider has attachment: false, preventing image input #28690

Description

Environment

• OpenCode version: 1.15.6
• Provider: alibaba-coding-plan-cn
• Model: qwen3.6-plus

Describe the bug
When using the qwen3.6-plus model with the alibaba-coding-plan-cn provider, image attachments are silently ignored and the model cannot understand images.

Root cause
The model registry (~/.cache/opencode/models.json) has "attachment": false for qwen3.6-plus:

"qwen3.6-plus": {
  "id": "qwen3.6-plus",
  "name": "Qwen3.6 Plus",
  "attachment": false,
  "modalities": {"input": ["text", "image", "video"], "output": ["text"]}
}

The modalities.input correctly includes image and video, but attachment: false prevents OpenCode from processing image attachments.

Expected behavior
Since qwen3.6-plus supports vision (modalities include image/video), attachment should be true.

Workaround
Add this to ~/.config/opencode/opencode.jsonc:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "alibaba-coding-plan-cn": {
      "models": {
        "qwen3.6-plus": {
          "attachment": true
        }
      }
    }
  }
}

Related
This is different from #20813 which is about API message format errors. This issue is about the model registry configuration preventing images from being sent at all.