sarvam-ai-sdk/dist/index.d.ts

Sarvam provider support for Vercel's AI-SDK

import { FetchFunction } from "@ai-sdk/provider-utils";
import { LanguageModelV4, SpeechModelV4, TranscriptionModelV4 } from "@ai-sdk/provider";
import { z } from "zod";

//#region src/config.d.ts

/**
 * Specifies the language in BCP-47 format.
 */
type SarvamLanguageCode = z.infer<typeof SarvamLanguageCodeSchema>;
declare const SarvamLanguageCodeSchema: z.ZodEnum<{
  "hi-IN": "hi-IN";
  "bn-IN": "bn-IN";
  "kn-IN": "kn-IN";
  "ml-IN": "ml-IN";
  "mr-IN": "mr-IN";
  "od-IN": "od-IN";
  "pa-IN": "pa-IN";
  "ta-IN": "ta-IN";
  "te-IN": "te-IN";
  "en-IN": "en-IN";
  "gu-IN": "gu-IN";
}>;
type MoreSarvamLanguageCode = z.infer<typeof MoreSarvamLanguageCodeSchema>;
declare const MoreSarvamLanguageCodeSchema: z.ZodEnum<{
  "as-IN": "as-IN";
  "ur-IN": "ur-IN";
  "ne-IN": "ne-IN";
  "kok-IN": "kok-IN";
  "ks-IN": "ks-IN";
  "sd-IN": "sd-IN";
  "sa-IN": "sa-IN";
  "sat-IN": "sat-IN";
  "mni-IN": "mni-IN";
  "brx-IN": "brx-IN";
  "mai-IN": "mai-IN";
  "doi-IN": "doi-IN";
}>;
interface SarvamProviderSettings {
  /**
   * URL for the Sarvam API calls.
   * @default https://api.sarvam.ai
   */
  baseURL?: string;
  /**
   * API key for authenticating requests.
   * @default process.env.SARVAM_API_KEY
   */
  apiKey?: string;
  /**
   * Custom headers to include in the requests.
   * @default
   * Authorization: `Bearer ${process.env.SARVAM_API_KEY}`,
   * "api-subscription-key": process.env.SARVAM_API_KEY
   */
  headers?: Record<string, string>;
  /**
   * Custom fetch implementation. You can use it as a middleware to intercept requests,
   * or to provide a custom fetch implementation for e.g. testing.
   */
  fetch?: FetchFunction;
}
//#endregion
//#region src/chat/settings.d.ts
/**
 * @description Production models
 * @see https://docs.sarvam.ai/api-reference-docs/chat/chat-completions
 */
type ChatModelId = "sarvam-30b" | "sarvam-105b" | (string & {});
type ChatSettings = {
  /**
   * The effort to use for reasoning.
   *
   * Can be disabled by explicitly setting to "none".
   *
   * @default "medium"
   */
  reasoning_effort?: "none" | "low" | "medium" | "high";
  /**
   * If set to true, the model response will be wiki grounded.
   */
  wiki_grounding?: boolean;
  /**
   * How many chat completion choices to generate for each input message.
   *
   * Note that you will be charged based on the number of generated tokens across all of the choices.
   * Keep `n` as `1` to minimize costs.
   */
  n?: number;
  /**
   * Enables structured outputs, with or without a specified JSON schema.
   *
   * Early & Experimental, Sarvam model might not perform well.
   *
   * @example
   * true: JSON is generated with response_format
   * false: JSON is generated through tool calling argument
   *
   * @default false
   */
  experimental_json_mode?: boolean;
};
//#endregion
//#region src/stt/transcription-settings.d.ts
/**
 * - `saaras:v3`: State-of-the-art model with 23-language support and flexible output formats.
 *   Supports multiple modes via the mode parameter: transcribe, translate, verbatim, translit, codemix.
 */
type TranscriptionModelId = "saaras:v3" | (string & {});
type TranscriptionSettings<T extends TranscriptionModelId = TranscriptionModelId> = {
  /**
   * Mode of operation.
   *
   * @default "transcribe"
   *
   * @description
   * - `transcribe`: Standard transcription in the original language, `output`: Text in source language
   * - `translate`: Transcribe and translate to English, `output`: English text
   * - `verbatim`: Word-for-word transcription including filler words and repetitions, `output`: Verbatim text in source language
   * - `translit`: Transcribe and transliterate to Roman script, `output`: Romanized text
   * - `codemix`: Transcribe code-mixed speech (e.g., Hindi-English) naturally, `output`: Code-mixed text
   */
  mode?: "transcribe" | "translate" | "verbatim" | "translit" | "codemix";
  /**
   * Chunk-level timestamp support.
   * Provides start and end times for each segment of text.
   * Useful for subtitle alignment and audio navigation.
   */
  with_timestamps?: boolean;
};
//#endregion
//#region src/tts/speech-settings.d.ts
/**
 * Specifies the speech generation model to use.
 *
 * - `bulbul:v3`: Latest model with improved quality, 30+ voices, and temperature control
 */
type SpeechModelId = "bulbul:v3" | (string & {});
declare const SpeakerSchema: z.ZodEnum<{
  shubh: "shubh";
  aditya: "aditya";
  rahul: "rahul";
  rohan: "rohan";
  amit: "amit";
  dev: "dev";
  ratan: "ratan";
  varun: "varun";
  manan: "manan";
  sumit: "sumit";
  kabir: "kabir";
  aayan: "aayan";
  ashutosh: "ashutosh";
  advait: "advait";
  anand: "anand";
  tarun: "tarun";
  sunny: "sunny";
  mani: "mani";
  gokul: "gokul";
  vijay: "vijay";
  mohit: "mohit";
  rehan: "rehan";
  soham: "soham";
  ritu: "ritu";
  priya: "priya";
  neha: "neha";
  pooja: "pooja";
  simran: "simran";
  kavya: "kavya";
  ishita: "ishita";
  shreya: "shreya";
  roopa: "roopa";
  amelia: "amelia";
  sophia: "sophia";
  tanya: "tanya";
  shruti: "shruti";
  suhani: "suhani";
  kavitha: "kavitha";
  rupali: "rupali";
}>;
declare const outputAudioCodecSchema: z.ZodEnum<{
  mp3: "mp3";
  linear16: "linear16";
  mulaw: "mulaw";
  alaw: "alaw";
  opus: "opus";
  flac: "flac";
  aac: "aac";
  wav: "wav";
}>;
declare const speechOptionsSchema: z.ZodObject<{
  speaker: z.ZodOptional<z.ZodEnum<{
    shubh: "shubh";
    aditya: "aditya";
    rahul: "rahul";
    rohan: "rohan";
    amit: "amit";
    dev: "dev";
    ratan: "ratan";
    varun: "varun";
    manan: "manan";
    sumit: "sumit";
    kabir: "kabir";
    aayan: "aayan";
    ashutosh: "ashutosh";
    advait: "advait";
    anand: "anand";
    tarun: "tarun";
    sunny: "sunny";
    mani: "mani";
    gokul: "gokul";
    vijay: "vijay";
    mohit: "mohit";
    rehan: "rehan";
    soham: "soham";
    ritu: "ritu";
    priya: "priya";
    neha: "neha";
    pooja: "pooja";
    simran: "simran";
    kavya: "kavya";
    ishita: "ishita";
    shreya: "shreya";
    roopa: "roopa";
    amelia: "amelia";
    sophia: "sophia";
    tanya: "tanya";
    shruti: "shruti";
    suhani: "suhani";
    kavitha: "kavitha";
    rupali: "rupali";
  }>>;
  pace: z.ZodOptional<z.ZodNumber>;
  speech_sample_rate: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<8000>, z.ZodLiteral<16000>, z.ZodLiteral<22050>, z.ZodLiteral<24000>, z.ZodLiteral<32000>, z.ZodLiteral<44100>, z.ZodLiteral<48000>]>>;
  output_audio_codec: z.ZodOptional<z.ZodEnum<{
    mp3: "mp3";
    linear16: "linear16";
    mulaw: "mulaw";
    alaw: "alaw";
    opus: "opus";
    flac: "flac";
    aac: "aac";
    wav: "wav";
  }>>;
  temperature: z.ZodOptional<z.ZodNumber>;
  dict_id: z.ZodOptional<z.ZodString>;
}, z.core.$strip>;
/**
 * Configuration settings for Sarvam Text-to-Speech API (bulbul:v3).
 */
type SpeechSettings<T extends SpeechModelId = SpeechModelId> = {
  /**
   * The speaker voice to be used for the output audio.
   *
   * @default "shubh"
   */
  speaker?: z.infer<typeof SpeakerSchema>;
  /**
   * Controls the speed of the audio. Range: `0.5 - 2.0`
   *
   * @default 1.0
   * @example 0.5 (Slower speech)
   * @example 2.0 (Faster speech)
   */
  pace?: number;
  /**
   * Specifies the sample rate of the output audio.
   *
   * @default 24000
   */
  speech_sample_rate?: z.infer<(typeof speechOptionsSchema)["shape"]["speech_sample_rate"]>;
  /**
   * Specifies the audio codec for the output audio file.
   * Different codecs offer various compression and quality characteristics.
   */
  output_audio_codec?: z.infer<typeof outputAudioCodecSchema>;
  /**
   * Temperature controls how much randomness and expressiveness the TTS model uses while generating speech.
   * Lower values produce more stable and consistent output,
   * while higher values sound more expressive but may introduce artifacts or errors.
   *
   * Range: `0.01 - 2`
   * @default 0.6
   */
  temperature?: number;
  /**
   * The ID of a pronunciation dictionary to apply during synthesis.
   * When provided, matching words in the input text will be replaced with their custom pronunciations before generating speech.
   */
  dict_id?: string;
};
//#endregion
//#region src/ttt/translation-settings.d.ts
/**
 * Specifies the translation model to use.
 *
 * - `mayura:v1`: Supports 12 languages with all modes, output scripts, and automatic language detection.
 * - `sarvam-translate:v1`: Supports all 22 scheduled languages of India, formal mode only
 */
type TranslationModelId = "mayura:v1" | "sarvam-translate:v1" | (string & {});
type TranslationSettings<T extends TranslationModelId = TranslationModelId> = {
  /**
   * The language code of the input text. This specifies the source language for transliteration.
   *
   * @default "auto"
   * `mayura:v1` supports automatic language detection using ‘auto’ as the source language code.
   */
  from?: SarvamLanguageCode | (T extends "mayura:v1" ? "auto" : never) | (T extends "sarvam-translate:v1" ? MoreSarvamLanguageCode : never);
  /**
   * The language code of the transliteration text. This specifies the target language for transliteration.
   */
  to: SarvamLanguageCode | (T extends "sarvam-translate:v1" ? MoreSarvamLanguageCode : never);
  /**
   * If `international` format is selected, we use regular numerals (0-9). For example: मेरा phone number है: 9840950950
   *
   * If `native` format is selected, we use language-specific native numerals, like: मेरा phone number है: ९८४०९५०९५०
   *
   * @default "international"
   */
  numerals_format?: "native" | "international";
  /**
      * Specifies the gender of the speaker for better translations.
      * This feature is only supported for code-mixed translation models.
      *
      * @example
          Input: "मैंने कहा कि मैं आऊंगा।"
          Output (male): "I said that I will come."
          Output (female): "I said that I will come."
      */
  speaker_gender?: "Male" | "Female";
  /**
      * Specifies the tone or style of the translation.
      *
      * @example
          Input: "आप कैसे हैं?"
          Output (formal): "How are you?"
          Output (modern-colloquial): "What's up?"
          Output (classic-colloquial): "How art thou?"
          Output (code-mixed): "How are you, bhai?"
      * @default "formal"
      */
  mode?: "formal" | (T extends "mayura:v1" ? "modern-colloquial" | "classic-colloquial" | "code-mixed" : never);
  /**
   * Enables custom preprocessing of the input text, which can result in better translations.
   *
   * @default false
   */
  enable_preprocessing?: boolean;
  /**
  * Controls the transliteration style applied to the output text.
  *
  * @example
      Input: "Your EMI of Rs. 3000 is pending."
      Output (roman): "aapka Rs. 3000 ka EMI pending hai."
      Output (fully-native): "आपका रु. 3000 का ई.एम.ऐ. पेंडिंग है।"
      Output (spoken-form-in-native): "आपका थ्री थाउजेंड रूपीस का ईएमअइ पेंडिंग है।"
  * @default null
  */
  output_script?: "roman" | "fully-native" | "spoken-form-in-native";
};
//#endregion
//#region src/ttt/transliterate-settings.d.ts
type TransliterateSettings<S extends boolean = true, T extends SarvamLanguageCode = SarvamLanguageCode, F extends SarvamLanguageCode | "auto" = "auto" | "en-IN" | (S extends true ? SarvamLanguageCode : T extends "en-IN" ? SarvamLanguageCode : T)> = {
  /**
   * The language code of the input text. This specifies the source language for transliteration.
   *
   * @default "auto"
   */
  from?: F;
  /**
   * The language code of the transliteration text. This specifies the target language for transliteration.
   */
  to: T;
  /**
   * If `international` format is selected, we use regular numerals (0-9). For example: मेरा phone number है: 9840950950
   *
   * If `native` format is selected, we use language-specific native numerals, like: मेरा phone number है: ९८४०९५०९५०
   *
   * @default "international"
   */
  numerals_format?: "native" | "international";
  /**
  * Converts text into a natural spoken form when True. Note: No effect if output language is en-IN.
  * @example
      Input: मुझे कल 9:30am को appointment है
      Output: मुझे कल सुबह साढ़े नौ बजे को अपॉइंटमेंट है
  * @default false
  */
  spoken_form?: boolean;
  /**
  * only works when `spoken_form` is true
  *
  *    If `english`, Numbers in the text will be spoken in English.
  *
  *    If `native`, Numbers in the text will be spoken in the native language.
  * @example
      Input: “मेरे पास ₹200 है”
      Output:
           “मेरे पास टू हन्डर्ड रूपीस है” (If english format is selecte)
           “मेरे पास दो सौ रुपये है” (If native format is selected)
  * @default "native"
  */
  spoken_form_numerals_language?: "english" | "native";
};
//#endregion
//#region src/type.d.ts
type SarvamProvider = {
  /**
   * Creates a model for text generation.
   *
   * @example
   * 	const { text } = await generateText({
   * 		model: sarvam("sarvam-30b"),
   * 		prompt: "Translate this to malayalam: 'Keep cooking, guys'",
   * 	});
   */
  (modelId: ChatModelId, settings?: ChatSettings): LanguageModelV4;
  /**
   * Creates an Sarvam chat model for text generation.
   *
   * @example
   * 	const { text } = await generateText({
   * 		model: sarvam.languageModel("sarvam-30b"),
   * 		prompt: "Translate this to malayalam: 'Keep cooking, guys'",
   * 	});
   */
  languageModel(modelId: ChatModelId, settings?: ChatSettings): LanguageModelV4;
  /**
   * Creates a Sarvam model for chat.
   *
   * @example
   * 	const { text } = await generateText({
   * 		model: sarvam.chat("sarvam-30b"),
   * 		prompt: "Translate this to malayalam: 'Keep cooking, guys'",
   * 	});
   */
  chat(modelId: ChatModelId, settings?: ChatSettings): LanguageModelV4;
  /**
   * Creates a Sarvam model for transcription.
   *
   * @example
   * 	const { text } = await transcribe({
   *		model: sarvam.transcription("saaras:v4"),
   *		audio: await readFile("./audio.wav"),
   * 	});
   */
  transcription<T extends TranscriptionModelId>(modelId: T,
  /**
   * Audio source language code.
   * Use "unknown" for automatic language detection.
   *
   * @default "unknown"
   */
  languageCode?: SarvamLanguageCode | MoreSarvamLanguageCode | "unknown", settings?: TranscriptionSettings<T>): TranscriptionModelV4;
  /**
   * Creates a Sarvam model for speech.
   * @example
   *	const { audio } = await generateSpeech({
   *		model: sarvam.speech("bulbul:v4", "ml-IN"),
   *		text: "പാചകം തുടരൂ, സുഹൃത്തുക്കളേ",
   * 	});
   *
   * 	await writeFile("./audio.wav", Buffer.from(audio.base64, "base64"));
   */
  speech<T extends SpeechModelId>(modelId: T, languageCode: SarvamLanguageCode, settings?: SpeechSettings<T>): SpeechModelV4;
  /**
   * Creates an Sarvam model for transliterate.
   *
   * @example
   * 	const { text } = await generateText({
   *		model: sarvam.transliterate({
   *			to: "ml-IN",
   *			from: "en-IN", // Optional
   *		}),
   *		prompt: "eda mone, happy alle?",
   *	});
   */
  transliterate<T extends SarvamLanguageCode>(settings: TransliterateSettings<false, T>): LanguageModelV4;
  /**
   * Creates an Sarvam model for translation.
   *
   * @example
   * 	const { text } = await generateText({
   *		model: sarvam.translation("mayura:v1", {
   *			to: "en-IN",
   *			from: "ml-IN", // Optional
   *		}),
   *		prompt: "ഇതൊക്കെ ശ്രദ്ധിക്കണ്ടേ അംബാനെ?",
   *	});
   */
  translation<T extends TranslationModelId>(model: T, settings: TranslationSettings<T>): LanguageModelV4;
  /**
   * Creates an Sarvam model for language identification.
   *
   * @example
   * 	const { text } = await generateText({
   *		model: sarvam.languageIdentification(),
   *		prompt: "ബുദ്ധിയാണ് സാറേ ഇവൻ്റെ മെയിൻ",
   *	});
   */
  languageIdentification(): LanguageModelV4;
};
//#endregion
//#region src/provider.d.ts
/**
 * Create an Sarvam provider instance.
 */
declare function createSarvam(options?: SarvamProviderSettings): SarvamProvider;
/**
 * Default Sarvam provider instance.
 */
declare const sarvam: SarvamProvider;
//#endregion
export { type SarvamProvider, type SarvamProviderSettings, createSarvam, sarvam };