多模态模型如何设置识图?光把模型识图开关打开了还不行! #4092
macurph8126
started this conversation in
Show and tell
Replies: 1 comment
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
找到了!正如你分析的,
model_settings里只有image_input: "supported",但缺少modalities声明。Harness 识别不到模型支持图片输入,直接丢弃图片数据。按照你给的方案,需要加上:
看到了一个关键问题 —
modalities字段被 API 忽略了!好的,现在问题完全清晰了。让我总结一下排查结果:
找到了!数据库中
providers表有model_settings列。现在直接修改数据库,给每个模型加上modalities声明。数据库已直接修改成功!
✅ 修复完成
问题根因
如你分析的,Harness(aionrs)的
model_settings缺少modalities声明:image_input: "supported"— 告诉 UI 显示识图开关modalities: {input: ["text", "image"]}— 告诉 Harness 把图片数据塞进 API 请求aioncore config providers update这个 CLI 会静默丢弃modalities字段,所以我绕过 CLI,直接修改了底层 SQLite 数据库。已修改内容
给所有支持视觉的模型添加了
modalities声明:["text", "image"]["text", "image"]["text", "image"]["text", "image"]["text", "image"]["text", "image"]["text", "image"]["text", "image"]["text", "image"]接下来你需要做的
mimo-v2-omni或任意模型All reactions