import { VoiceTextSDK } from 'typeless-sdk';

const sdk = new VoiceTextSDK({
  stt: {
    endpoint: 'https://api.groq.com/openai/v1/audio/transcriptions',
    model: 'whisper-large-v3-turbo',
    apiKey: process.env.GROQ_API_KEY!,
  },
  llm: {
    baseUrl: 'https://api.openai.com/v1',
    apiKey: process.env.OPENAI_API_KEY!,
    model: 'gpt-4o-mini',
  },
});

const { transcript, polishedText } = await sdk.process('meeting.m4a', {
  vocabulary: ['OpenTypeless', 'KPI', 'EBITDA'],
  language: 'zh',
  appType: 'document',
});

console.log('Raw:    ', transcript);
console.log('Cleaned:', polishedText);

// Built-in: Whisper-compatible file upload
const sdk = new VoiceTextSDK({
  stt: {
    endpoint: string,   // /audio/transcriptions URL
    model: string,
    apiKey: string,
    language?: string,
    extraFields?: Record<string, string>,
    timeoutMs?: number,
  },
  llm: LlmConfig,
});

// Custom adapter: any STT protocol
const sdk = new VoiceTextSDK({
  stt: async (audio, filename) => {
    // audio: string (file path) | Buffer | Uint8Array
    // return the raw transcript string
    return myTranscribe(audio, filename);
  },
  llm: LlmConfig,
});

const { transcript, polishedText } = await sdk.process(
  'recording.m4a',          // string path or Buffer / Uint8Array
  {
    vocabulary?: (string | VocabularyEntry)[],  // custom terms; use VocabularyEntry for soundsLike aliases
    language?: string,       // BCP-47 hint for STT ('zh', 'en', …)
    appType?: AppType,       // 'general' | 'email' | 'chat' | 'document' | 'code'
    translateEnabled?: boolean,
    targetLang?: string,     // BCP-47 translation target ('en', 'ja', …)
    onChunk?: (token: string) => void,  // enables LLM streaming mode
    polish?: boolean,        // set false to skip LLM, return raw transcript
  }
);

const raw = await sdk.transcribe('/path/to/audio.mp3');
// or from Buffer:
const raw = await sdk.transcribe(buffer, 'recording.mp3');

const cleaned = await sdk.polish('嗯那个就是说我们的方案还不错', {
  vocabulary: ['方案A'],
  appType: 'chat',
});
// → '我们的方案A还不错'

import { transcribeAudio, polishText, buildSystemPrompt } from 'typeless-sdk';

const transcript = await transcribeAudio(
  '/path/to/audio.wav',   // string | Buffer | Uint8Array
  {
    endpoint: string,      // full API endpoint URL
    model: string,         // e.g. 'whisper-1'
    apiKey: string,
    language?: string,     // BCP-47 or 'multi' for auto-detect
    extraFields?: Record<string, string>,  // provider-specific fields
    timeoutMs?: number,    // default: 60_000
  }
);

const polished = await polishText(
  rawTranscript,
  {
    baseUrl: string,       // OpenAI-compatible base URL
    apiKey: string,
    model: string,
    maxTokens?: number,    // default: 4096
    temperature?: number,  // default: 0.3
    timeoutMs?: number,    // default: 60_000
  },
  {
    appType?: AppType,
    vocabulary?: string[],
    translateEnabled?: boolean,
    targetLang?: string,
    onChunk?: (chunk: string) => void,
  }
);

const prompt = buildSystemPrompt({
  appType: 'email',
  vocabulary: ['API', 'SLA'],
  translateEnabled: true,
  targetLang: 'en',
});

const sdk = new VoiceTextSDK({
  stt: async (audio, filename = 'audio.wav') => {
    const blob = audio instanceof Blob ? audio : new Blob([audio]);
    const res = await fetch('https://your-api/chat/completions', {
      method: 'POST',
      headers: { Authorization: `Bearer ${apiKey}`, 'Content-Type': 'application/json' },
      body: JSON.stringify({
        model: 'your-asr-model',
        messages: [{ role: 'user', content: [{ type: 'input_audio', input_audio: { data: await toDataUri(blob) } }] }],
      }),
    });
    const data = await res.json();
    return data.choices[0].message.content;
  },
  llm: { ... },
});

process.stdout.write('Polishing: ');

const { polishedText } = await sdk.process('voice-memo.m4a', {
  onChunk: (token) => process.stdout.write(token),
});

console.log('\nDone:', polishedText);

const { transcript } = await sdk.process('audio.wav', { polish: false });

const { polishedText } = await sdk.process('chinese-meeting.m4a', {
  language: 'zh',
  translateEnabled: true,
  targetLang: 'en',
});
// speaks Chinese → outputs English

import { readFile } from 'fs/promises';

const audio = await readFile('recording.mp3');
const { polishedText } = await sdk.transcribe(audio, 'recording.mp3');
// or run the full pipeline via the low-level function:
// const text = await transcribeAudio(audio, sttConfig, 'recording.mp3');

const sdk = new VoiceTextSDK({
  stt: {
    endpoint: 'https://open.bigmodel.cn/api/paas/v4/audio/transcriptions',
    model: 'glm-asr-2512',
    apiKey: process.env.GLM_API_KEY!,
    extraFields: { stream: 'false' },
    language: 'zh',
  },
  llm: {
    baseUrl: 'https://open.bigmodel.cn/api/paas/v4',
    apiKey: process.env.GLM_API_KEY!,
    model: 'glm-4-flash',
  },
});

const { polishedText } = await sdk.process('standup.m4a', {
  vocabulary: ['gRPC', 'KPI', '季报'],
  appType: 'document',
});

import type { VocabularyEntry } from 'typeless-sdk';

const { polishedText } = await sdk.process('standup.m4a', {
  vocabulary: [
    { term: 'OpenTypeless', soundsLike: ['open type less', 'opentypeless'] },
    { term: 'Tauri',        soundsLike: ['towery', 'tori'] },
    { term: 'gRPC',         soundsLike: ['grpc', 'g rpc'] },
    'KPI',   // plain string — no alias needed
  ],
  appType: 'document',
});

// Default STT config is 'zh', but override to 'en' for this specific file
// (only works with built-in SttConfig; custom adapters manage their own config)
const result = await sdk.process('english-note.mp3', { language: 'en' });

import { VoiceTextSDK } from 'typeless-sdk';

const sdk = new VoiceTextSDK({
  stt: {
    endpoint: 'https://api.groq.com/openai/v1/audio/transcriptions',
    model: 'whisper-large-v3-turbo',
    apiKey: process.env.GROQ_API_KEY!,
  },
  llm: {
    baseUrl: 'https://open.bigmodel.cn/api/paas/v4',
    apiKey: process.env.GLM_API_KEY!,
    model: 'glm-4-flash',
  },
});

const { transcript, polishedText } = await sdk.process('会议录音.m4a', {
  vocabulary: ['季报', 'Q4', 'KPI', 'OpenTypeless'],  // 自定义词库
  language: 'zh',
  appType: 'document',
});

console.log('原始转录：', transcript);
console.log('润色结果：', polishedText);

const sdk = new VoiceTextSDK({
  stt: async (audio, filename) => {
    // 在这里实现任意 STT 协议，返回转录文字即可
    return myCustomTranscribe(audio, filename);
  },
  llm: { ... },
});

const { transcript, polishedText } = await sdk.process(
  'audio.m4a',              // 文件路径 或 Buffer / Uint8Array
  {
    vocabulary?: string[],       // 自定义词库，LLM 严格保持这些词的拼写
    language?: string,           // STT 语言提示（'zh'、'en' 等）
    appType?: AppType,           // 场景：'general' | 'email' | 'chat' | 'document' | 'code'
    translateEnabled?: boolean,  // 是否翻译
    targetLang?: string,         // 翻译目标语言（'en'、'ja' 等）
    onChunk?: (token) => void,   // 流式回调（提供时启用 streaming 模式）
    polish?: boolean,            // 设为 false 跳过 LLM，只返回原始转录
  }
);

const rawText = await sdk.transcribe('audio.mp3');

const cleaned = await sdk.polish('嗯那个就是说我们的项目进展不错', {
  vocabulary: ['项目X'],
  appType: 'chat',
});

const { polishedText } = await sdk.process('standup.m4a', {
  vocabulary: ['Sprint', 'P0', 'LGTM'],
  appType: 'document',
  language: 'zh',
});

const { polishedText } = await sdk.process('chinese.m4a', {
  language: 'zh',
  translateEnabled: true,
  targetLang: 'en',
});

await sdk.process('audio.wav', {
  onChunk: (token) => process.stdout.write(token),
});

const { transcript } = await sdk.process('audio.mp3', { polish: false });