Can't stop promise chain javascript

小能豆

Can't stop promise chain javascript

javascript

I have a stream of text that I send to a function to turn into speech. I send the text sentence by sentence for better latency. The issue im running into is handling interruptions. Right now, I send each sentence to a function containing a promise. So I end up with a chain of awaits that pile on fairly quickly, so when I try to interrupt the function, althought it properly breaks out, the await chain continues to run. Here is a snippet of the code (send_voice returns a promise)

const stream = await openai.chat.completions.create({
      model: "gpt-3.5-turbo",
      messages: conversation_history,
      stream: true,
    });

    let sentence = ""; 

    for await (const chunk of stream) {

      const content = chunk.choices[0]?.delta?.content || "";
      sentence += content; 

      if (interruptionState.value){
        console.log("breaking");
        break; //break correctly called but send_voice chain continues
      }

      if (sentence.endsWith('.') || sentence.endsWith('!') || sentence.endsWith('?')) {

        console.log("FINAL: ", sentence);

        conversation_history.push({ "role": "system", "content": sentence });

        await send_voice(sentence, ws, elleven_key, voice_id,interruptionState); 
        sentence = ""; 
      }
    }

I know that I can’t stop the await chain, but is there any way around this? I essentially need to call send_voice sequentially and be able to stop it quickly on interruption. I’ve been stuck on this for a while so any help would be much appreciated!

I’ve been stuck on this for a while so any help would be much appreciated!

阅读 75

2023-12-21

共1个答案

小能豆

The issue you’re facing is likely due to the fact that await inside a loop creates a chain of promises, and even if you break out of the loop, the existing promises in the chain will continue to resolve. One way to handle this is by using a flag to indicate whether you should continue processing or stop.

Here’s an example of how you could modify your code to use a flag for interruption:

const stream = await openai.chat.completions.create({
  model: "gpt-3.5-turbo",
  messages: conversation_history,
  stream: true,
});

let sentence = "";
let continueProcessing = true;

const processChunk = async (chunk) => {
  const content = chunk.choices[0]?.delta?.content || "";
  sentence += content;

  if (!continueProcessing) {
    console.log("Breaking");
    return;
  }

  if (sentence.endsWith('.') || sentence.endsWith('!') || sentence.endsWith('?')) {
    console.log("FINAL: ", sentence);

    conversation_history.push({ "role": "system", "content": sentence });

    await send_voice(sentence, ws, elleven_key, voice_id, interruptionState);
    sentence = "";
  }
};

for await (const chunk of stream) {
  await processChunk(chunk);
}

// If you need to stop processing externally (e.g., on interruption)
// you can set continueProcessing to false.
continueProcessing = false;

By setting continueProcessing to false, you can stop processing at the next iteration of the loop. The processChunk function is then responsible for handling the logic within each chunk.

2023-12-21