ollama配置问题

[2314254971]

Search before asking

• I had searched in the issues and found no similar issues.

Operating system information

Windows

What happened

问题：设置ollama生成模型在聊天时会报错：
执行失败pemja.core.PythonException: <class 'tenacity.RetryError'>: <Future at 0x7f8d69e3fc70 state=finished raised RuntimeError>
at /openspg_venv/lib/python3.8/site-packages/kag/solver/main_solver.invoke(main_solver.py:94)
at /openspg_venv/lib/python3.8/site-packages/kag/solver/logic/solver_pipeline.run(solver_pipeline.py:67)
at /openspg_venv/lib/python3.8/site-packages/kag/solver/implementation/default_reasoner.reason(default_reasoner.py:64)
at /openspg_venv/lib/python3.8/site-packages/kag/solver/execute/default_lf_executor.execute(default_lf_executor.py:239)
at /openspg_venv/lib/python3.8/site-packages/kag/solver/execute/default_lf_executor._execute_lf(default_lf_executor.py:204)
at /openspg_venv/lib/python3.8/site-packages/kag/solver/execute/default_lf_executor._execute_chunk_answer(default_lf_executor.py:154)
at /openspg_venv/lib/python3.8/site-packages/kag/solver/retriever/impl/default_chunk_retrieval.recall_docs(default_chunk_retrieval.py:425)

How to reproduce

我的yaml配置：
openie_llm: &openie_llm
base_url: http://localhost:11434/
model: qwen2_0.5b_instruct:latest
type: ollama

chat_llm: &chat_llm
base_url: http://localhost:11434/
model: qwen2_0.5b_instruct:latest
type: ollama

vectorize_model: &vectorize_model
api_key: empty
base_url: http://localhost:11434/v1/
model: bge-m3:latest #qwen2_0.5b_instruct:latest
type: openai
vector_dimensions: 1024
vectorizer: *vectorize_model

ollama版本0.5.6

Are you willing to submit PR?

• Yes I am willing to submit a PR!

[BBC-9527]

change the url ，like http://host.docker.internal:11434

[2314254971]

host.docker.internal

我在网站上配置时使用的就是http://host.docker.internal:11434/ ，报同样的错

[caszkgui]

host.docker.internal

我在网站上配置时使用的就是http://host.docker.internal:11434/ ，报同样的错

baseurl should add prefix of http://

[hanwsf]

vectorize_model: &vectorize_model
api_key: empty
base_url: http://127.0.0.1:11434/v1
model: bge-m3:latest
type: openai
vector_dimensions: 1024
vectorizer: *vectorize_model

以上配置在开发模式配置中可以创建知识库。
但是在产品模式中总是提示unknown error
<class 'RuntimeError'>: invalid vectorizer config: Connection error.

[2314254971]

host.docker.internal

我在网站上配置时使用的就是http://host.docker.internal:11434/ ，报同样的错

baseurl should add prefix of http://

您好，我按照您的说法，加上了前缀

但是问答还是有概率报错，图片如下

[hanwsf]

https://openspg.yuque.com/ndx6g9/docs/iu5cok24efl1z2nc
已经解决。

docker exec -it ...进入容器，查得网关地址为172.20.0.1. curl 172.20.0.1:11434/v1，回复是Ollama is running.
不要使用ollama类型本地模型。
在maas类型下增加模型。
http://172.20.0.1:11434/v1

---end---