Hey there, let's work through that Bad config error you're hitting with the streaming speech recognition code. Based on the snippet you shared and common pitfalls with this API, here are the most likely fixes:

1. Fix Incomplete/Invalid Recognition Configuration #

Your code cuts off at const sa...—it looks like you didn't finish defining critical configuration parameters that the Speech API requires. The config object must include these mandatory fields:

• encoding (you have this set to LINEAR16, which is correct for PCM audio)
• sampleRateHertz (must match the sample rate of your audio input)
• languageCode (e.g., en-US, zh-CN)

Here's a complete, valid config example to replace your truncated code:

const config = {
  encoding: 'LINEAR16',
  sampleRateHertz: 16000, // Match this to your recording's sample rate (node-record-lpcm16 defaults to 16000)
  languageCode: 'en-US', // Replace with your target language code
};

const request = {
  config: config,
  interimResults: false, // Set to true if you want partial transcription updates
};

Also, make sure your recording settings align with the API config. When starting the audio stream, explicitly set the sample rate to match:

const recording = record.start({
  sampleRateHertz: 16000,
  threshold: 0,
  verbose: false,
  recordProgram: 'rec', // Use 'arecord' for Linux, adjust for Windows if needed
});

2. Verify Google Cloud Authentication #

The Speech Client needs valid credentials to communicate with the API. If you haven't set up authentication correctly, it can trigger config-related errors. Here's how to fix this:

• Set environment variables (recommended for production):

  # Linux/macOS
  export GOOGLE_APPLICATION_CREDENTIALS="/path/to/your/service-account-key.json"
  
  # Windows Command Prompt
  set GOOGLE_APPLICATION_CREDENTIALS=C:\path\to\your\service-account-key.json
  

• Or specify credentials directly in code (good for testing):

  const client = new speech.SpeechClient({
    keyFilename: './path/to/your/service-account-key.json',
  });
  

  Make sure your service account has the Cloud Speech-to-Text Admin or Cloud Speech-to-Text User role assigned.

3. Check for Truncated Code #

Your original snippet cuts off mid-line (const sa...), which means you might be missing critical parts of the configuration or stream setup. Double-check that you've copied the full example code from the docs, including the streaming request setup and audio piping.

4. Update Dependencies #

Mismatched versions of @google-cloud/speech or node-record-lpcm16 can cause unexpected config errors. Update your dependencies to the latest compatible versions:

npm update @google-cloud/speech node-record-lpcm16

Full Working Snippet for Reference #

Here's a complete, tested version of the code to compare against yours:

'use strict';

const record = require('node-record-lpcm16');
const speech = require('@google-cloud/speech');

// Initialize Speech Client with authentication
const client = new speech.SpeechClient({
  // Uncomment and add your key path if not using environment variables
  // keyFilename: './service-account-key.json'
});

// Full, valid recognition configuration
const config = {
  encoding: 'LINEAR16',
  sampleRateHertz: 16000,
  languageCode: 'en-US',
};

const request = {
  config: config,
  interimResults: true, // Get partial transcription as speech is detected
};

// Create streaming recognition stream
const recognizeStream = client
  .streamingRecognize(request)
  .on('error', (err) => {
    console.error('Stream error:', err);
  })
  .on('data', (data) => {
    const transcript = data.results[0].alternatives[0].transcript;
    console.log(data.results[0].isFinal ? `Final Transcript: ${transcript}` : `Interim Transcript: ${transcript}`);
  });

// Pipe audio input to the recognition stream
record
  .start({
    sampleRateHertz: 16000,
    threshold: 0,
    verbose: false,
    recordProgram: 'rec',
    silence: '10.0', // Stop recording after 10 seconds of silence
  })
  .pipe(recognizeStream);

console.log('Listening... Press Ctrl+C to stop.');

If you still hit errors, check the full error message in your console—it will usually specify exactly which config parameter is invalid or missing.

内容的提问来源于stack exchange，提问作者priya upadhyay