[Bug] vllm reranker error
Closed this issue · 9 comments
Contact Information
No response
MaxKB Version
版本 社区版 版本号 v2.1.0 (build at 2025-09-03T16:41, commit: d94871b)
Problem Description
添加vllm部署的重排模型vllm-bge-reranker-v2-m3的时候报错
认证失败，请检查参数是否正确：headers: {'date': 'Thu, 04 Sep 2025 06:04:58 GMT', 'server': 'uvicorn', 'content-length': '22', 'content-type': 'application/json'}, status_code: 404, body: {'detail': 'Not Found'}
用FastGPT是可以正常添加的
Steps to Reproduce
无
The expected correct result
No response
Related log output
Additional Information
No response
请检查基础模型、模型url是否正确
Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑🤝🧑👫🧑🏿🤝🧑🏻👩🏾🤝👨🏿👬🏿
Please check if the model url is correct
请按照如图模型名称和URL格式检查一下是否对应
@Shenguobin0102 @zhanweizhang7 @shaohuzhang1
各位 经过测试在MaxKB中添加嵌入、推理模型的时候端口号后加/v1是可以的 添加重排模型的时候不能增加/v1才行
在FastGPT等项目中添加重排模型的时候添加/v1也是可行的请知悉
Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑🤝🧑👫🧑🏿🤝🧑🏻👩🏾🤝👨🏿👬🏿
@Shenguobin0102 @zhanweizhang7 @shaohuzhang1
Dear everyone After testing, when adding embedding and inference models to MaxKB, it is OK to add /v1 after the port number. When adding the reordering model, you cannot add /v1.
Adding /v1 when adding re-arrangement models to projects such as FastGPT is also feasible. Please know
感谢反馈，后续我们会优化一下，同一个供应商的模型URL格式尽量保持统一。
Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑🤝🧑👫🧑🏿🤝🧑🏻👩🏾🤝👨🏿👬🏿
Thanks for the feedback, we will optimize it in the future, and try to keep the model URL format of the same supplier as uniform as possible.
v2.2.0已发布，请更新使用！
Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑🤝🧑👫🧑🏿🤝🧑🏻👩🏾🤝👨🏿👬🏿
v2.2.0 has been released, please update and use it!