search

chatchat-space / Langchain-Chatchat

Langchain-Chatchat（原Langchain-ChatGLM）基于 Langchain 与 ChatGLM, Qwen 与 Llama 等语言模型的 RAG 与 Agent 应用 | Langchain-Chatchat (formerly langchain-ChatGLM), local knowledge based LLM (like ChatGLM, Qwen and Llama) RAG and Agent app with langchain
Apache License 2.0
31.43k stars 5.48k forks source link

initialize knowledge base failed #4362

Closed umbraclet16 closed 2 months ago

umbraclet16 commented 3 months ago

I registered an embedding model 'autodl-tmp-bge-large-zh' in xinference and configured it in /root/chatchat-data/model_providers.yaml to use custom platform loading.

Running command chatchat-kb -r failed.

relevant log:

model_providers.bootstrap_web.openai_bootstrap_web 2700 WARNING Warning: model not found. Using cl100k_base encoding.

Internet is unavailable so the default model cannot be downloaded.

I wonder if model_providers only supports builtin embedding models? Or did I miss any configuration?

Thx.

liunux4odoo commented 2 months ago

0.3.1 版已经发布，优化了配置方式，修改配置项无需重启服务器，可以更新尝试。

umbraclet16 commented 2 months ago

定位问题在0.3.0版本的AutoDL镜像中model_providers部分代码，在此仅用于记录：

Langchain-Chatchat/libs/model-providers/model_providers/bootstrap_web/openai_bootstrap_web.py

create_embeddings()函数中if判断条件有问题，错误地进入了if分支。

简单解决：注释掉if else逻辑，改为：

input_embeddings_request.input
if isinstance(input, list):
    input = input[0]

另: list_models()存在几处变量错误。