@speech-sdk/core 0.7.0 → 0.8.1-alpha

package/dist/stt-providers/openai/index.js

@@ -1,184 +0,0 @@
-import { parseMediaTypeParam, wrapPcm16Mono } from "../../audio-utils.js";
-import { handleErrorResponse, resolveApiKey, SDK_USER_AGENT, } from "../../provider-utils.js";
-// OpenAI Whisper is advertised as 50+ languages; we list the ISO-639-1 codes
-// the API's `language` parameter accepts. Matches the TTS provider's list.
-const OPENAI_STT_LANGUAGES = [
-    "af",
-    "ar",
-    "az",
-    "be",
-    "bg",
-    "bn",
-    "bs",
-    "ca",
-    "cs",
-    "cy",
-    "da",
-    "de",
-    "el",
-    "en",
-    "es",
-    "et",
-    "fa",
-    "fi",
-    "fr",
-    "gl",
-    "he",
-    "hi",
-    "hr",
-    "hu",
-    "hy",
-    "id",
-    "is",
-    "it",
-    "ja",
-    "kk",
-    "kn",
-    "ko",
-    "lt",
-    "lv",
-    "mi",
-    "mk",
-    "mr",
-    "ms",
-    "ne",
-    "nl",
-    "no",
-    "pl",
-    "pt",
-    "ro",
-    "ru",
-    "sk",
-    "sl",
-    "sr",
-    "sv",
-    "sw",
-    "ta",
-    "th",
-    "tl",
-    "tr",
-    "uk",
-    "ur",
-    "vi",
-    "zh",
-];
-/**
- * OpenAI Whisper / gpt-4o-transcribe adapter for the SDK's derived-timestamps
- * path. Uses `/v1/audio/transcriptions` with `timestamp_granularities: ["word"]`
- * and `response_format: "verbose_json"`.
- *
- * Note: `gpt-4o-transcribe-diarize` is intentionally not listed — that
- * variant does not support `timestamp_granularities`.
- */
-export class OpenAISpeechToTextProvider {
-    id = "openai";
-    defaultModel = "whisper-1";
-    // Only whisper-1 supports word-level timestamps. The newer
-    // gpt-4o-transcribe / gpt-4o-mini-transcribe models accept `json` /
-    // `text` response formats only and don't expose `timestamp_granularities`,
-    // so they can't satisfy this provider's contract.
-    models = [
-        {
-            id: "whisper-1",
-            releaseDate: "2023-03-01",
-            languages: OPENAI_STT_LANGUAGES,
-        },
-    ];
-    apiKey;
-    baseURL;
-    fetchFn;
-    constructor(config = {}) {
-        this.apiKey = config.apiKey;
-        this.baseURL = config.baseURL ?? "https://api.openai.com/v1";
-        this.fetchFn = config.fetch ?? globalThis.fetch.bind(globalThis);
-    }
-    async transcribe(options) {
-        const { audio, mediaType } = await normalizeAudioForOpenAI(options.audio, options.mediaType);
-        const form = new FormData();
-        const filename = `audio.${mediaTypeToExtension(mediaType)}`;
-        // Cast via BlobPart: TS narrowing of Uint8Array<ArrayBufferLike> vs
-        // Blob's required ArrayBuffer-backed view is stricter than runtime.
-        form.append("file", new Blob([audio], { type: mediaType }), filename);
-        form.append("model", options.modelId);
-        form.append("response_format", "verbose_json");
-        form.append("timestamp_granularities[]", "word");
-        if (options.language) {
-            form.append("language", options.language);
-        }
-        const response = await this.fetchFn(`${this.baseURL}/audio/transcriptions`, {
-            method: "POST",
-            headers: {
-                Authorization: `Bearer ${resolveApiKey(this.apiKey, "OPENAI_API_KEY", "OpenAI")}`,
-                "X-User-Agent": SDK_USER_AGENT,
-                ...options.headers,
-            },
-            body: form,
-            signal: options.abortSignal,
-        });
-        await handleErrorResponse(response, `openai/${options.modelId}`);
-        const data = (await response.json());
-        const timestamps = (data.words ?? []).map((w) => ({
-            text: w.word,
-            start: w.start,
-            end: w.end,
-        }));
-        return {
-            timestamps,
-            text: data.text,
-        };
-    }
-}
-export function createOpenAISTT(config = {}) {
-    const provider = new OpenAISpeechToTextProvider(config);
-    return function openaiSTT(modelId) {
-        return {
-            provider,
-            modelId: modelId ?? provider.defaultModel,
-        };
-    };
-}
-// OpenAI transcription accepts mp3/mp4/mpeg/mpga/m4a/wav/webm/flac/ogg/opus
-// but rejects raw PCM. When a TTS provider hands us raw little-endian PCM
-// (stitch mode), we wrap it with a WAV header so the STT endpoint will
-// parse it. `audio/l16` is intentionally NOT handled: RFC 2586 defines it
-// as big-endian and `wrapPcm16Mono` writes little-endian — silently mis-
-// wrapping would corrupt audio. No current provider emits L16; add an
-// explicit byte-swap branch here if one does.
-async function normalizeAudioForOpenAI(audio, mediaType) {
-    if (mediaTypeBase(mediaType) === "audio/pcm") {
-        const sampleRate = parseMediaTypeParam(mediaType, "rate") ?? 24_000;
-        return {
-            audio: await wrapPcm16Mono(audio, sampleRate),
-            mediaType: "audio/wav",
-        };
-    }
-    return { audio, mediaType };
-}
-function mediaTypeBase(mediaType) {
-    return mediaType.split(";")[0]?.trim().toLowerCase() ?? "";
-}
-function mediaTypeToExtension(mediaType) {
-    switch (mediaTypeBase(mediaType)) {
-        case "audio/mpeg":
-        case "audio/mp3":
-            return "mp3";
-        case "audio/wav":
-        case "audio/x-wav":
-            return "wav";
-        case "audio/ogg":
-            return "ogg";
-        case "audio/opus":
-            return "opus";
-        case "audio/flac":
-            return "flac";
-        case "audio/webm":
-            return "webm";
-        case "audio/mp4":
-        case "audio/m4a":
-        case "audio/x-m4a":
-            return "m4a";
-        default:
-            return "mp3";
-    }
-}
-//# sourceMappingURL=index.js.map

package/dist/stt-providers/openai/index.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/stt-providers/openai/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,mBAAmB,EAAE,aAAa,EAAE,MAAM,sBAAsB,CAAC;AAC1E,OAAO,EACL,mBAAmB,EACnB,aAAa,EACb,cAAc,GACf,MAAM,yBAAyB,CAAC;AAajC,6EAA6E;AAC7E,2EAA2E;AAC3E,MAAM,oBAAoB,GAAG;IAC3B,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;CACI,CAAC;AAEX;;;;;;;GAOG;AACH,MAAM,OAAO,0BAA0B;IAC5B,EAAE,GAAG,QAAQ,CAAC;IACd,YAAY,GAAG,WAAW,CAAC;IAEpC,2DAA2D;IAC3D,oEAAoE;IACpE,2EAA2E;IAC3E,kDAAkD;IACzC,MAAM,GAAG;QAChB;YACE,EAAE,EAAE,WAAW;YACf,WAAW,EAAE,YAAY;YACzB,SAAS,EAAE,oBAAoB;SAChC;KACO,CAAC;IAEM,MAAM,CAAqB;IAC3B,OAAO,CAAS;IAChB,OAAO,CAA0B;IAElD,YAAY,SAA2C,EAAE;QACvD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,2BAA2B,CAAC;QAC7D,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,IAAI,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;IACnE,CAAC;IAED,KAAK,CAAC,UAAU,CAAC,OAOhB;QAKC,MAAM,EAAE,KAAK,EAAE,SAAS,EAAE,GAAG,MAAM,uBAAuB,CACxD,OAAO,CAAC,KAAK,EACb,OAAO,CAAC,SAAS,CAClB,CAAC;QAEF,MAAM,IAAI,GAAG,IAAI,QAAQ,EAAE,CAAC;QAC5B,MAAM,QAAQ,GAAG,SAAS,oBAAoB,CAAC,SAAS,CAAC,EAAE,CAAC;QAC5D,oEAAoE;QACpE,oEAAoE;QACpE,IAAI,CAAC,MAAM,CACT,MAAM,EACN,IAAI,IAAI,CAAC,CAAC,KAAiB,CAAC,EAAE,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC,EAClD,QAAQ,CACT,CAAC;QACF,IAAI,CAAC,MAAM,CAAC,OAAO,EAAE,OAAO,CAAC,OAAO,CAAC,CAAC;QACtC,IAAI,CAAC,MAAM,CAAC,iBAAiB,EAAE,cAAc,CAAC,CAAC;QAC/C,IAAI,CAAC,MAAM,CAAC,2BAA2B,EAAE,MAAM,CAAC,CAAC;QACjD,IAAI,OAAO,CAAC,QAAQ,EAAE,CAAC;YACrB,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,OAAO,CAAC,QAAQ,CAAC,CAAC;QAC5C,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CACjC,GAAG,IAAI,CAAC,OAAO,uBAAuB,EACtC;YACE,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,aAAa,EAAE,UAAU,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,gBAAgB,EAAE,QAAQ,CAAC,EAAE;gBACjF,cAAc,EAAE,cAAc;gBAC9B,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI;YACV,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CACF,CAAC;QAEF,MAAM,mBAAmB,CAAC,QAAQ,EAAE,UAAU,OAAO,CAAC,OAAO,EAAE,CAAC,CAAC;QAEjE,MAAM,IAAI,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAGlC,CAAC;QAEF,MAAM,UAAU,GAAoB,CAAC,IAAI,CAAC,KAAK,IAAI,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;YACjE,IAAI,EAAE,CAAC,CAAC,IAAI;YACZ,KAAK,EAAE,CAAC,CAAC,KAAK;YACd,GAAG,EAAE,CAAC,CAAC,GAAG;SACX,CAAC,CAAC,CAAC;QAEJ,OAAO;YACL,UAAU;YACV,IAAI,EAAE,IAAI,CAAC,IAAI;SAChB,CAAC;IACJ,CAAC;CACF;AAED,MAAM,UAAU,eAAe,CAAC,SAA2C,EAAE;IAC3E,MAAM,QAAQ,GAAG,IAAI,0BAA0B,CAAC,MAAM,CAAC,CAAC;IAExD,OAAO,SAAS,SAAS,CAAC,OAAgB;QACxC,OAAO;YACL,QAAQ;YACR,OAAO,EAAE,OAAO,IAAI,QAAQ,CAAC,YAAY;SAC1C,CAAC;IACJ,CAAC,CAAC;AACJ,CAAC;AAED,4EAA4E;AAC5E,0EAA0E;AAC1E,uEAAuE;AACvE,0EAA0E;AAC1E,yEAAyE;AACzE,sEAAsE;AACtE,8CAA8C;AAC9C,KAAK,UAAU,uBAAuB,CACpC,KAAiB,EACjB,SAAiB;IAEjB,IAAI,aAAa,CAAC,SAAS,CAAC,KAAK,WAAW,EAAE,CAAC;QAC7C,MAAM,UAAU,GAAG,mBAAmB,CAAC,SAAS,EAAE,MAAM,CAAC,IAAI,MAAM,CAAC;QACpE,OAAO;YACL,KAAK,EAAE,MAAM,aAAa,CAAC,KAAK,EAAE,UAAU,CAAC;YAC7C,SAAS,EAAE,WAAW;SACvB,CAAC;IACJ,CAAC;IACD,OAAO,EAAE,KAAK,EAAE,SAAS,EAAE,CAAC;AAC9B,CAAC;AAED,SAAS,aAAa,CAAC,SAAiB;IACtC,OAAO,SAAS,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,CAAC,WAAW,EAAE,IAAI,EAAE,CAAC;AAC7D,CAAC;AAED,SAAS,oBAAoB,CAAC,SAAiB;IAC7C,QAAQ,aAAa,CAAC,SAAS,CAAC,EAAE,CAAC;QACjC,KAAK,YAAY,CAAC;QAClB,KAAK,WAAW;YACd,OAAO,KAAK,CAAC;QACf,KAAK,WAAW,CAAC;QACjB,KAAK,aAAa;YAChB,OAAO,KAAK,CAAC;QACf,KAAK,WAAW;YACd,OAAO,KAAK,CAAC;QACf,KAAK,YAAY;YACf,OAAO,MAAM,CAAC;QAChB,KAAK,YAAY;YACf,OAAO,MAAM,CAAC;QAChB,KAAK,YAAY;YACf,OAAO,MAAM,CAAC;QAChB,KAAK,WAAW,CAAC;QACjB,KAAK,WAAW,CAAC;QACjB,KAAK,aAAa;YAChB,OAAO,KAAK,CAAC;QACf;YACE,OAAO,KAAK,CAAC;IACjB,CAAC;AACH,CAAC"}