Problem

qwen3.7-plus supports multimodal input (image + video), but the current modality detection logic treats it as text-only.

The defaultModalities() function in modalityDefaults.ts uses ordered regex matching. Since there is no explicit pattern for qwen3.7-plus, it falls through to the catch-all /^qwen/ pattern which returns {} (text-only).

Meanwhile, qwen3.7-max is correctly text-only — it only supports text input.

Expected Behavior

• qwen3.7-plus → { image: true, video: true } (multimodal)
• qwen3.7-max → {} (text-only, no change needed)

Root Cause

Two places need updating:

1. packages/core/src/core/modalityDefaults.ts — MODALITY_PATTERNS is missing an entry for qwen3.7-plus. It falls through to /^qwen/ → {}.
2. packages/core/src/core/openaiContentGenerator/provider/dashscope.ts — VISION_MODEL_PREFIX_PATTERNS doesn't include qwen3.7-plus, so the DashScope provider won't set vl_high_resolution_images: true for it.

Proposed Fix

1. Add [/^qwen3\.7-plus/, { image: true, video: true }] to MODALITY_PATTERNS before the catch-all /^qwen/.
2. Add 'qwen3.7-plus' to VISION_MODEL_PREFIX_PATTERNS in the DashScope provider.
3. Add/update tests to verify the modality defaults for both qwen3.7-plus (multimodal) and qwen3.7-max (text-only).

Impact

Users selecting qwen3.7-plus through any provider (token plan, custom config) will correctly get multimodal support — images and videos will be sent as inline data instead of being downgraded to text placeholders.