IssueHuntlobehub/lobe-chat
The issue has been closed
[Bug] 集成ollama调用llava多模态模型时无法识别图片 #1482
yumcc-com posted onGitHub
💻 系统环境
Windows
📦 部署环境
Docker
🌐 浏览器
Chrome
🐛 问题描述
docker运行lobe-chat，且集成ollama，切换成llava模型，正常文字可以识别，上传图片会报错。
{
"error": {
"message": "json: cannot unmarshal array into Go struct field Message.messages.content of type string",
"type": "invalid_request_error",
"param": null,
"code": null
},
"endpoint": "http://host.do****er.internal:****/v1",
"provider": "ollama"
}
🚦 期望结果
上传图片后能正常识别解析
📷 复现步骤
No response
📝 补充信息
No response