Copilot chat participant with tool calls: how to debug error 400 responses?

[shouples]

While working on integrating tool calls to our extension's chat participant, I was trying to figure out how to get LanguageModelToolResults back into the conversation flow*, and I kept seeing this:

And looking at our extension's logging, all I would see is:

Error: Request Failed: 400 Bad Request\n
        at v2._provideLanguageModelResponse (~/.vscode/extensions/github.copilot-chat-0.27.0/dist/extension.js:478:7171)
        at processTicksAndRejections (node:internal/process/task_queues:95:5)

I didn't see any extra information in the dev tools, and it seemed to come from the for await (const fragment of response.stream) after calling the .sendRequest() method with a specific arrangement of messages. Looking at the GitHub Copilot Chat output channel also didn't help much either:

Is there a recommended way for extension developers to get details out of error 400 responses?

Note

*I ended up handling this by passing the LanguageModelToolResult's LanguageModelTextPart items through to a LanguageModelToolResultPart, which was then wrapped in a User message. I then managed to get past the error 400 responses by adding an Assistant message wrapping the LanguageModelToolCallPart, so it ended up looking like:

const messages = [
  ...pastMessages,
  LanguageModelChatMessage.Assistant([LanguageModelToolCallPart], PARTICIPANT_ID),
  LanguageModelChatMessage.User([LanguageModelToolResultPart], "tool:<tool name>"),
];
const response: LanguageModelChatResponse = await model.sendRequest(
  messages,
  requestOptions,
  token,
);