Hello, and thank you for providing such a great open-source project!
I am currently using openkb with a local ollama backend.
While testing various local models, I encountered the following error:
[ERROR] litellm.UnsupportedParamsError: ollama does not support parameters: ['parallel_tool_calls'], for model=gemma4:e4b. To drop these, set 'litellm.drop_params=True' or for proxy:
'litellm_settings:
drop_params: true'.
If you want to use these params dynamically send allowed_openai_params=['parallel_tool_calls'] in your request.
Following the error message's suggestion, I tried adding the drop_params: true option in my .openkb/config.yaml, but it did not take effect.
Upon reviewing the source code, it appears that openkb currently only uses the model information and does not pass additional litellm configurations.
As a workaround, I am currently exporting it as an environment variable (e.g., export LITELLM_DROP_PARAMS=True) to bypass the issue.
Are there any future plans to support parsing and applying litellm configurations (like litellm_settings) directly from .openkb/config.yaml?
Thank you again for your time and effort!
Hello, and thank you for providing such a great open-source project!
I am currently using openkb with a local ollama backend.
While testing various local models, I encountered the following error:
Following the error message's suggestion, I tried adding the drop_params: true option in my .openkb/config.yaml, but it did not take effect.
Upon reviewing the source code, it appears that openkb currently only uses the model information and does not pass additional litellm configurations.
As a workaround, I am currently exporting it as an environment variable (e.g., export LITELLM_DROP_PARAMS=True) to bypass the issue.
Are there any future plans to support parsing and applying litellm configurations (like litellm_settings) directly from .openkb/config.yaml?
Thank you again for your time and effort!