any tips on debugging

[foldl]

I am writing an OpenAI API compatible server for twinny. After enabled debugging, I can only see information about the request in Developer Tool.

The issue I encounter is that streamed FIM response is not recognized by twinny. Any tips on how to debug this?

[rjmacarthy]

Hello, the best way to debug it is to build it from source.

1. Clone the repository,
2. Run npm i inside the root dir
3. Open the dir it inside vscode
4. Press f5 to start debugging
5. Open stream.ts set some breakpoints on the response/request

Many thanks.

[foldl]

Thanks for your suggestions. I am scared of using npm, 😄 Fortunately, chatllm.cpp is OK when setting provider to litellm.

From utils.ts, I see that there are subtle differences in different providers. Could twinny add a generic OpenAI Compatible provider? I think litellm (and my chatllm.cpp) belongs to this category.