Any ideas on how I can debug why my message stream gets cut off abruptly?

[dosstx]

I am using the AI SDK to generate short stories. I am trying to debug why the messages stream gets cut off sometimes mid-sentence. For example:

I am using the OpenAI API SDK with AI SDK in a Firebase Cloud Function.

Anyone got any advice on how to debug this? Here are my openAI chat completions settings:

    try {
      const { messages } = request.body
      const config = {
        // model: 'gpt-3.5-turbo',
        model: 'gpt-4-1106-preview',
        stream: true,
        messages: [
          {
            role: 'system',
            content:
              "<My custom system message here.>"
          },
          {
            role: messages[0].role,
            content: messages[0].content
          }
        ],
        temperature: 1,
        max_tokens: 3000,
        top_p: 0.82,
        frequency_penalty: 0,
        presence_penalty: 0
      }
      const openaiResponse = await openai.chat.completions.create(config)

      // Transform the OpenAI response into a readable stream
      const stream = OpenAIStream(openaiResponse)
      return streamToResponse(stream, response)
      ```

The total characters in the generated stories is about 3000.

Any thoughts?

[KesleyDavid]

@dosstx same problem with me, did you manage to solve it?