sarvam-ai-sdk 0.0.6 → 0.0.8

package/README.md

@@ -69,6 +69,21 @@ const { text } = await transcribe({
 console.log(text); // പാചകം തുടരും സുഹൃത്തുക്കളെ
 ```
 
+## Speech-to-Text-Translate
+
+```ts
+import { sarvam } from "sarvam-ai-sdk";
+import { experimental_transcribe as transcribe } from "ai";
+import { readFile } from "fs/promises";
+
+const result = await transcribe({
+    model: sarvam.speechTranslation("saaras:v2"),
+    audio: await readFile("./src/transcript-test.wav"),
+});
+
+console.log(result.text); // Cooking continues, my friends
+```
+
 ## Translation
 
 > NB: Only transliterates `prompt` and `role:user` messages, not `system` not `assistant`.
@@ -79,8 +94,8 @@ import { generateText } from "ai";
 
 const result = await generateText({
     model: sarvam.translation({
+        "from": "ml-IN",
         "to": "en-IN",
-        "from": "ml-IN"
     }),
     prompt: "ഇതൊക്കെ ശ്രദ്ധിക്കണ്ടേ അംബാനെ?",
 });
@@ -98,7 +113,7 @@ import { generateText } from "ai";
 
 const result = await generateText({
   model: sarvam.transliterate({
-      from: "en-IN"
+      from: "en-IN",
       to: "ml-IN",
   }),
   prompt: "eda mone, happy alle?",

package/dist/index.cjs

@@ -55,7 +55,7 @@ __export(index_exports, {
 module.exports = __toCommonJS(index_exports);
 
 // src/sarvam-provider.ts
-var import_provider_utils10 = require("@ai-sdk/provider-utils");
+var import_provider_utils11 = require("@ai-sdk/provider-utils");
 
 // src/sarvam-chat-language-model.ts
 var import_provider3 = require("@ai-sdk/provider");
@@ -1112,7 +1112,8 @@ var SarvamTranslationModel = class {
     this.specificationVersion = "v1";
     this.supportsStructuredOutputs = false;
     this.defaultObjectGenerationMode = "json";
-    this.modelId = "unknown";
+    var _a;
+    this.modelId = (_a = settings.model) != null ? _a : "mayura:v1";
     this.settings = settings;
     this.config = config;
   }
@@ -1126,13 +1127,22 @@ var SarvamTranslationModel = class {
     mode,
     prompt
   }) {
-    var _a, _b, _c, _d, _e, _f;
+    var _a, _b, _c, _d, _e, _f, _g, _h;
     const type = mode.type;
     const warnings = [];
-    if (this.settings.from !== "auto") {
-      if (this.settings.to !== "en-IN" && this.settings.from !== "en-IN")
+    if (this.settings.from === this.settings.to) {
+      throw new Error(
+        "Source and target languages code must be different."
+      );
+    }
+    if (this.modelId === "sarvam-translate:v1") {
+      if (((_a = this.settings.mode) != null ? _a : "formal") !== "formal")
         throw new Error(
-          "Sarvam doesn't support Indic-Indic Transliteration yet"
+          "Sarvam 'sarvam-translate:v1' only support mode formal."
+        );
+      if (((_b = this.settings.from) != null ? _b : "auto") === "auto")
+        throw new Error(
+          "Sarvam 'sarvam-translate:v1' requires source language code."
         );
     }
     if (type !== "regular") {
@@ -1144,14 +1154,14 @@ var SarvamTranslationModel = class {
       messages,
       args: {
         input: messages.filter((m) => m.role === "user").map((m) => m.content).join("\n"),
-        source_language_code: (_a = this.settings.from) != null ? _a : "auto",
+        source_language_code: (_c = this.settings.from) != null ? _c : "auto",
         target_language_code: this.settings.to,
-        numerals_format: (_b = this.settings.numerals_format) != null ? _b : "international",
-        enable_preprocessing: (_c = this.settings.enable_preprocessing) != null ? _c : false,
-        output_script: (_d = this.settings.output_script) != null ? _d : null,
-        speaker_gender: (_e = this.settings.speaker_gender) != null ? _e : "Male",
-        mode: (_f = this.settings.mode) != null ? _f : "formal"
-        // model: this.settings.model ?? "male",
+        numerals_format: (_d = this.settings.numerals_format) != null ? _d : "international",
+        enable_preprocessing: (_e = this.settings.enable_preprocessing) != null ? _e : false,
+        output_script: (_f = this.settings.output_script) != null ? _f : null,
+        speaker_gender: (_g = this.settings.speaker_gender) != null ? _g : "Male",
+        mode: (_h = this.settings.mode) != null ? _h : "formal",
+        model: this.modelId
       },
       warnings
     };
@@ -1198,7 +1208,7 @@ var SarvamTranslationModel = class {
     };
   }
   async doStream(options) {
-    throw new Error("Translation feature doesn't streaming yet");
+    throw new Error("Translation feature doesn't support streaming yet");
   }
 };
 var sarvamTranslationResponseSchema = import_zod8.z.object({
@@ -1395,11 +1405,98 @@ var sarvamLidResponseSchema = import_zod10.z.object({
   request_id: import_zod10.z.string().nullish()
 });
 
+// src/sarvam-speech-translation-model.ts
+var import_provider_utils10 = require("@ai-sdk/provider-utils");
+var import_zod11 = require("zod");
+var SarvamSpeechTranslationModel = class {
+  constructor(modelId, config) {
+    this.modelId = modelId;
+    this.config = config;
+    this.specificationVersion = "v1";
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  getArgs({
+    audio,
+    mediaType,
+    providerOptions
+  }) {
+    const warnings = [];
+    const formData = new FormData();
+    const blob = audio instanceof Blob ? audio : new Blob([audio], { type: mediaType });
+    formData.append("file", blob);
+    formData.append("model", this.modelId);
+    return {
+      formData,
+      warnings
+    };
+  }
+  async doGenerate(options) {
+    var _a, _b, _c;
+    const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
+    const { formData, warnings } = this.getArgs(options);
+    const {
+      value: response,
+      responseHeaders,
+      rawValue: rawResponse
+    } = await (0, import_provider_utils10.postFormDataToApi)({
+      url: this.config.url({
+        path: "/speech-to-text-translate",
+        modelId: this.modelId
+      }),
+      headers: (0, import_provider_utils10.combineHeaders)(this.config.headers(), options.headers),
+      formData,
+      failedResponseHandler: sarvamFailedResponseHandler,
+      successfulResponseHandler: (0, import_provider_utils10.createJsonResponseHandler)(
+        sarvamTranscriptionResponseSchema2
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    return {
+      text: response.transcript,
+      segments: [],
+      language: response.language_code ? response.language_code : void 0,
+      durationInSeconds: void 0,
+      warnings,
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: rawResponse
+      }
+    };
+  }
+};
+var sarvamTranscriptionResponseSchema2 = import_zod11.z.object({
+  request_id: import_zod11.z.string().nullable(),
+  transcript: import_zod11.z.string(),
+  language_code: import_zod11.z.string().nullable(),
+  // timestamps: z
+  //   .object({
+  //     end_time_seconds: z.array(z.number()),
+  //     start_time_seconds: z.array(z.number()),
+  //     words: z.array(z.string()),
+  //   })
+  //   .optional(),
+  diarized_transcript: import_zod11.z.object({
+    entries: import_zod11.z.array(
+      import_zod11.z.object({
+        end_time_seconds: import_zod11.z.array(import_zod11.z.number()),
+        start_time_seconds: import_zod11.z.array(import_zod11.z.number()),
+        transcript: import_zod11.z.string(),
+        speaker_id: import_zod11.z.string()
+      })
+    )
+  }).nullable().optional()
+});
+
 // src/sarvam-provider.ts
 function createSarvam(options = {}) {
   var _a;
-  const baseURL = (_a = (0, import_provider_utils10.withoutTrailingSlash)(options.baseURL)) != null ? _a : "https://api.sarvam.ai";
-  const ApiKey = (0, import_provider_utils10.loadApiKey)({
+  const baseURL = (_a = (0, import_provider_utils11.withoutTrailingSlash)(options.baseURL)) != null ? _a : "https://api.sarvam.ai";
+  const ApiKey = (0, import_provider_utils11.loadApiKey)({
     apiKey: options.apiKey,
     environmentVariableName: "SARVAM_API_KEY",
     description: "Sarvam"
@@ -1422,15 +1519,19 @@ function createSarvam(options = {}) {
     }
     return createChatModel(modelId, settings);
   };
-  const createTranscriptionModel = (modelId, languageCode = "unknown", settings) => {
-    return new SarvamTranscriptionModel(modelId, languageCode, {
-      provider: "sarvam.transcription",
-      url: ({ path }) => `${baseURL}${path}`,
-      headers: getHeaders,
-      fetch: options.fetch,
-      transcription: settings
-    });
-  };
+  const createTranscriptionModel = (modelId, languageCode = "unknown", settings) => new SarvamTranscriptionModel(modelId, languageCode, {
+    provider: "sarvam.transcription",
+    url: ({ path }) => `${baseURL}${path}`,
+    headers: getHeaders,
+    fetch: options.fetch,
+    transcription: settings
+  });
+  const createSpeechTranslation = (modelId) => new SarvamSpeechTranslationModel(modelId, {
+    provider: "sarvam.transcription",
+    url: ({ path }) => `${baseURL}${path}`,
+    headers: getHeaders,
+    fetch: options.fetch
+  });
   const createSpeechModel = (modelId, languageCode, settings) => new SarvamSpeechModel(modelId, languageCode, {
     provider: "sarvam.speech",
     url: ({ path }) => `${baseURL}${path}`,
@@ -1439,10 +1540,7 @@ function createSarvam(options = {}) {
     speech: settings
   });
   const createTransliterateModel = (settings) => new SarvamTransliterateModel(
-    {
-      to: settings.to,
-      from: settings.from
-    },
+    settings,
     {
       provider: "sarvam.transliterate",
       url: ({ path }) => `${baseURL}${path}`,
@@ -1451,10 +1549,7 @@ function createSarvam(options = {}) {
     }
   );
   const createTranslationModel = (settings) => new SarvamTranslationModel(
-    {
-      to: settings.to,
-      from: settings.from
-    },
+    settings,
     {
       provider: "sarvam.translation",
       url: ({ path }) => `${baseURL}${path}`,
@@ -1474,6 +1569,7 @@ function createSarvam(options = {}) {
   provider.languageModel = createLanguageModel;
   provider.chat = createChatModel;
   provider.transcription = createTranscriptionModel;
+  provider.speechTranslation = createSpeechTranslation;
   provider.speech = createSpeechModel;
   provider.transliterate = createTransliterateModel;
   provider.translation = createTranslationModel;