mirror of
https://github.com/QwenLM/qwen-code.git
synced 2026-05-16 11:15:19 +00:00
Fixes #3616. Adds opt-in `splitToolMedia` flag (default false). When enabled, media parts (image / audio / video / file) returned by MCP tool calls are split into a follow-up `role: "user"` message instead of being embedded in the `role: "tool"` message. Required for strict OpenAI-compatible servers (e.g., LM Studio) that reject non-text content on tool messages with HTTP 400 "Invalid 'messages' in payload". Media from parallel tool responses is accumulated and emitted as a single follow-up user message after all tool messages, preserving OpenAI's contiguity requirement for tool responses. Default behavior is unchanged for permissive providers. |
||
|---|---|---|
| .. | ||
| constants.ts | ||
| content-generator-config.test.ts | ||
| content-generator-config.ts | ||
| index.ts | ||
| modelConfigErrors.ts | ||
| modelConfigResolver.test.ts | ||
| modelConfigResolver.ts | ||
| modelRegistry.test.ts | ||
| modelRegistry.ts | ||
| modelsConfig.test.ts | ||
| modelsConfig.ts | ||
| types.ts | ||