@forgeframework/ai-voice

Voice processing for the Forge Framework.

Part of the Forge Framework — a TypeScript framework for backend microservices and web applications.

Installation

npm install @forgeframework/ai-voice

Usage

import {
  VoiceProcessor,
  SttProvider,
  AudioFormat,
  VoiceConfig,
} from '@forgeframework/ai-voice';
import { AIProviderFactory, AIFramework } from '@forgeframework/ai-llm';
import { readFile } from 'node:fs/promises';

const llmProvider = AIProviderFactory.createInstance({
  framework: AIFramework.NATIVE,
  model: 'gpt-4o',
});

const voiceConfig: VoiceConfig = {
  sttProvider: SttProvider.WHISPER,
  language: 'en',
  diarization: { enabled: true, maxSpeakers: 4 },
  punctuation: true,
  transcription: {
    model: 'whisper-1',
    language: 'en',
    chunkDuration: 15,
    overlapDuration: 0.2,
    maxFileSize: 307200,
    maxConcurrency: 4,
    timeout: 30000,
    retryAttempts: 2,
  },
  summarization: {
    model: 'gpt-4o',
    maxTokens: 1024,
    systemPrompt: 'Summarize the following meeting transcript.',
    extractActionItems: true,
    extractKeyPoints: true,
  },
};

const processor = new VoiceProcessor(voiceConfig, llmProvider);

const audio = await readFile('/recordings/meeting-2026-04-27.wav');

const result = await processor.transcribeAndSummarize(
  new Uint8Array(audio),
  'Summarize this meeting and extract action items.',
  { format: AudioFormat.WAV, diarization: true },
);

console.log(`Transcript: ${result.transcript}`);
console.log(`Summary: ${result.summary}`);
console.log(`Key Points: ${result.keyPoints.join(', ')}`);
console.log(`Action Items: ${result.actionItems.length}`);
for (const item of result.actionItems) {
  console.log(`  - ${item.description} (${item.assignee ?? 'unassigned'})`);
}
console.log(`Speakers: ${result.speakers?.join(', ')}`);

Documentation

Full API reference and guides: https://carbonforge.io/framework/docs/ai/ai-voice

(Documentation site launching soon.)

License

MIT © Carbon Forge