sarvam-ai-sdk 0.0.4 → 0.0.5

package/README.md

@@ -69,6 +69,25 @@ const { text } = await transcribe({
 console.log(text); // പാചകം തുടരും സുഹൃത്തുക്കളെ
 ```
 
+## Translation
+
+> Only transliterates `prompt` and `role:user` messages, not `system` not `assistant`.
+
+```ts
+import { sarvam } from "sarvam-ai-sdk";
+import { generateText } from "ai";
+
+const result = await generateText({
+    model: sarvam.translation({
+        "to": "en-IN",
+        "from": "ml-IN"
+    }),
+    prompt: "ഇതൊക്കെ ശ്രദ്ധിക്കണ്ടേ അംബാനെ?",
+});
+
+console.log(result.text); // Shouldn't we be careful about this, Ambane?
+```
+
 ## Transliterate
 
 > Only transliterates `prompt` and `role:user` messages, not `system` not `assistant`.

package/dist/index.cjs

@@ -55,7 +55,7 @@ __export(index_exports, {
 module.exports = __toCommonJS(index_exports);
 
 // src/sarvam-provider.ts
-var import_provider_utils8 = require("@ai-sdk/provider-utils");
+var import_provider_utils9 = require("@ai-sdk/provider-utils");
 
 // src/sarvam-chat-language-model.ts
 var import_provider3 = require("@ai-sdk/provider");
@@ -758,140 +758,12 @@ var sarvamChatChunkSchema = import_zod2.z.union([
   sarvamErrorDataSchema
 ]);
 
-// src/sarvam-transcription-model.ts
-var import_provider_utils5 = require("@ai-sdk/provider-utils");
-var import_zod4 = require("zod");
-
-// src/sarvam-transcription-settings.ts
-var import_zod3 = require("zod");
-var SarvamProviderOptionsSchema = import_zod3.z.object({
-  with_timestamps: import_zod3.z.boolean().nullish().default(false),
-  with_diarization: import_zod3.z.boolean().nullish().default(false),
-  num_speakers: import_zod3.z.number().int().nullish()
-});
-
-// src/sarvam-transcription-model.ts
-var SarvamTranscriptionModel = class {
-  constructor(modelId, languageCode, config) {
-    this.modelId = modelId;
-    this.languageCode = languageCode;
-    this.config = config;
-    this.specificationVersion = "v1";
-  }
-  get provider() {
-    return this.config.provider;
-  }
-  getArgs({
-    audio,
-    mediaType,
-    providerOptions
-  }) {
-    const warnings = [];
-    if (this.modelId === "saarika:v1" && this.languageCode === "unknown")
-      throw new Error(
-        "Language code unknown is not supported for model saarika:v1"
-      );
-    const sarvamOptions = (0, import_provider_utils5.parseProviderOptions)({
-      provider: "sarvam",
-      providerOptions: {
-        sarvam: __spreadValues(__spreadValues({}, providerOptions == null ? void 0 : providerOptions.sarvam), this.config.transcription)
-      },
-      schema: SarvamProviderOptionsSchema
-    });
-    const formData = new FormData();
-    const blob = audio instanceof Blob ? audio : new Blob([audio], { type: mediaType });
-    formData.append("file", blob);
-    formData.append("model", this.modelId);
-    if (sarvamOptions) {
-      formData.append("language_code", this.languageCode);
-      formData.append(
-        "with_timestamps",
-        sarvamOptions.with_timestamps ? "true" : "false"
-      );
-      formData.append(
-        "with_diarization",
-        sarvamOptions.with_diarization ? "true" : "false"
-      );
-      if (sarvamOptions.num_speakers !== null && sarvamOptions.num_speakers !== void 0) {
-        formData.append(
-          "num_speakers",
-          sarvamOptions.num_speakers.toString()
-        );
-      }
-    }
-    return {
-      formData,
-      warnings
-    };
-  }
-  async doGenerate(options) {
-    var _a, _b, _c, _d, _e;
-    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_utils5.postFormDataToApi)({
-      url: this.config.url({
-        path: "/speech-to-text",
-        modelId: this.modelId
-      }),
-      headers: (0, import_provider_utils5.combineHeaders)(this.config.headers(), options.headers),
-      formData,
-      failedResponseHandler: sarvamFailedResponseHandler,
-      successfulResponseHandler: (0, import_provider_utils5.createJsonResponseHandler)(
-        sarvamTranscriptionResponseSchema
-      ),
-      abortSignal: options.abortSignal,
-      fetch: this.config.fetch
-    });
-    return {
-      text: response.transcript,
-      segments: response.timestamps ? response.timestamps.words.map((word, index) => ({
-        text: word,
-        startSecond: response.timestamps.start_time_seconds[index],
-        endSecond: response.timestamps.end_time_seconds[index]
-      })) : [],
-      language: response.language_code ? response.language_code : void 0,
-      durationInSeconds: (_e = (_d = response.timestamps) == null ? void 0 : _d.end_time_seconds[response.timestamps.end_time_seconds.length - 1]) != null ? _e : void 0,
-      warnings,
-      response: {
-        timestamp: currentDate,
-        modelId: this.modelId,
-        headers: responseHeaders,
-        body: rawResponse
-      }
-    };
-  }
-};
-var sarvamTranscriptionResponseSchema = import_zod4.z.object({
-  request_id: import_zod4.z.string().nullable(),
-  transcript: import_zod4.z.string(),
-  language_code: import_zod4.z.string().nullable(),
-  timestamps: import_zod4.z.object({
-    end_time_seconds: import_zod4.z.array(import_zod4.z.number()),
-    start_time_seconds: import_zod4.z.array(import_zod4.z.number()),
-    words: import_zod4.z.array(import_zod4.z.string())
-  }).optional(),
-  diarized_transcript: import_zod4.z.object({
-    entries: import_zod4.z.array(
-      import_zod4.z.object({
-        end_time_seconds: import_zod4.z.array(import_zod4.z.number()),
-        start_time_seconds: import_zod4.z.array(import_zod4.z.number()),
-        transcript: import_zod4.z.string(),
-        speaker_id: import_zod4.z.string()
-      })
-    )
-  }).optional()
-});
-
 // src/sarvam-speech-model.ts
-var import_provider_utils6 = require("@ai-sdk/provider-utils");
+var import_provider_utils5 = require("@ai-sdk/provider-utils");
 
 // src/sarvam-speech-settings.ts
-var import_zod5 = require("zod");
-var SpeakerSchema = import_zod5.z.enum([
+var import_zod3 = require("zod");
+var SpeakerSchema = import_zod3.z.enum([
   "meera",
   "pavithra",
   "maitreyi",
@@ -912,22 +784,22 @@ var SpeakerSchema = import_zod5.z.enum([
   "karun",
   "hitesh"
 ]).default("meera");
-var SarvamProviderOptionsSchema2 = import_zod5.z.object({
+var SarvamProviderOptionsSchema = import_zod3.z.object({
   speaker: SpeakerSchema,
-  pitch: import_zod5.z.number().min(-0.75).max(0.75).default(0),
-  pace: import_zod5.z.number().min(0.5).max(2).default(1),
-  loudness: import_zod5.z.number().min(0.3).max(3).default(1),
-  speech_sample_rate: import_zod5.z.union([
-    import_zod5.z.literal(8e3),
-    import_zod5.z.literal(16e3),
-    import_zod5.z.literal(22050),
-    import_zod5.z.literal(24e3)
+  pitch: import_zod3.z.number().min(-0.75).max(0.75).default(0),
+  pace: import_zod3.z.number().min(0.5).max(2).default(1),
+  loudness: import_zod3.z.number().min(0.3).max(3).default(1),
+  speech_sample_rate: import_zod3.z.union([
+    import_zod3.z.literal(8e3),
+    import_zod3.z.literal(16e3),
+    import_zod3.z.literal(22050),
+    import_zod3.z.literal(24e3)
   ]).default(22050),
-  enable_preprocessing: import_zod5.z.boolean().default(false)
+  enable_preprocessing: import_zod3.z.boolean().default(false)
 }).partial();
 
 // src/sarvam-speech-model.ts
-var import_zod6 = require("zod");
+var import_zod4 = require("zod");
 var SarvamSpeechModel = class {
   constructor(modelId, languageCode, config) {
     this.modelId = modelId;
@@ -947,12 +819,12 @@ var SarvamSpeechModel = class {
     providerOptions
   }) {
     const warnings = [];
-    const sarvamOptions = (0, import_provider_utils6.parseProviderOptions)({
+    const sarvamOptions = (0, import_provider_utils5.parseProviderOptions)({
       provider: "sarvam",
       providerOptions: {
         sarvam: __spreadValues(__spreadValues({}, providerOptions == null ? void 0 : providerOptions.sarvam), this.config.speech)
       },
-      schema: SarvamProviderOptionsSchema2
+      schema: SarvamProviderOptionsSchema
     });
     const getSpeaker = () => {
       if (sarvamOptions == null ? void 0 : sarvamOptions.speaker) return sarvamOptions.speaker;
@@ -1011,18 +883,18 @@ var SarvamSpeechModel = class {
       value,
       responseHeaders,
       rawValue: rawResponse
-    } = await (0, import_provider_utils6.postJsonToApi)({
+    } = await (0, import_provider_utils5.postJsonToApi)({
       url: this.config.url({
         path: "/text-to-speech",
         modelId: this.modelId
       }),
-      headers: (0, import_provider_utils6.combineHeaders)(this.config.headers(), options.headers),
+      headers: (0, import_provider_utils5.combineHeaders)(this.config.headers(), options.headers),
       body: requestBody,
       failedResponseHandler: sarvamFailedResponseHandler,
-      successfulResponseHandler: (0, import_provider_utils6.createJsonResponseHandler)(
-        import_zod6.z.object({
-          request_id: import_zod6.z.string(),
-          audios: import_zod6.z.array(import_zod6.z.string())
+      successfulResponseHandler: (0, import_provider_utils5.createJsonResponseHandler)(
+        import_zod4.z.object({
+          request_id: import_zod4.z.string(),
+          audios: import_zod4.z.array(import_zod4.z.string())
         })
       ),
       abortSignal: options.abortSignal,
@@ -1045,7 +917,135 @@ var SarvamSpeechModel = class {
   }
 };
 
-// src/sarvam-transliterate-model.ts
+// src/sarvam-transcription-model.ts
+var import_provider_utils6 = require("@ai-sdk/provider-utils");
+var import_zod6 = require("zod");
+
+// src/sarvam-transcription-settings.ts
+var import_zod5 = require("zod");
+var SarvamProviderOptionsSchema2 = import_zod5.z.object({
+  with_timestamps: import_zod5.z.boolean().nullish().default(false),
+  with_diarization: import_zod5.z.boolean().nullish().default(false),
+  num_speakers: import_zod5.z.number().int().nullish()
+});
+
+// src/sarvam-transcription-model.ts
+var SarvamTranscriptionModel = class {
+  constructor(modelId, languageCode, config) {
+    this.modelId = modelId;
+    this.languageCode = languageCode;
+    this.config = config;
+    this.specificationVersion = "v1";
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  getArgs({
+    audio,
+    mediaType,
+    providerOptions
+  }) {
+    const warnings = [];
+    if (this.modelId === "saarika:v1" && this.languageCode === "unknown")
+      throw new Error(
+        "Language code unknown is not supported for model saarika:v1"
+      );
+    const sarvamOptions = (0, import_provider_utils6.parseProviderOptions)({
+      provider: "sarvam",
+      providerOptions: {
+        sarvam: __spreadValues(__spreadValues({}, providerOptions == null ? void 0 : providerOptions.sarvam), this.config.transcription)
+      },
+      schema: SarvamProviderOptionsSchema2
+    });
+    const formData = new FormData();
+    const blob = audio instanceof Blob ? audio : new Blob([audio], { type: mediaType });
+    formData.append("file", blob);
+    formData.append("model", this.modelId);
+    if (sarvamOptions) {
+      formData.append("language_code", this.languageCode);
+      formData.append(
+        "with_timestamps",
+        sarvamOptions.with_timestamps ? "true" : "false"
+      );
+      formData.append(
+        "with_diarization",
+        sarvamOptions.with_diarization ? "true" : "false"
+      );
+      if (sarvamOptions.num_speakers !== null && sarvamOptions.num_speakers !== void 0) {
+        formData.append(
+          "num_speakers",
+          sarvamOptions.num_speakers.toString()
+        );
+      }
+    }
+    return {
+      formData,
+      warnings
+    };
+  }
+  async doGenerate(options) {
+    var _a, _b, _c, _d, _e;
+    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_utils6.postFormDataToApi)({
+      url: this.config.url({
+        path: "/speech-to-text",
+        modelId: this.modelId
+      }),
+      headers: (0, import_provider_utils6.combineHeaders)(this.config.headers(), options.headers),
+      formData,
+      failedResponseHandler: sarvamFailedResponseHandler,
+      successfulResponseHandler: (0, import_provider_utils6.createJsonResponseHandler)(
+        sarvamTranscriptionResponseSchema
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    return {
+      text: response.transcript,
+      segments: response.timestamps ? response.timestamps.words.map((word, index) => ({
+        text: word,
+        startSecond: response.timestamps.start_time_seconds[index],
+        endSecond: response.timestamps.end_time_seconds[index]
+      })) : [],
+      language: response.language_code ? response.language_code : void 0,
+      durationInSeconds: (_e = (_d = response.timestamps) == null ? void 0 : _d.end_time_seconds[response.timestamps.end_time_seconds.length - 1]) != null ? _e : void 0,
+      warnings,
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: rawResponse
+      }
+    };
+  }
+};
+var sarvamTranscriptionResponseSchema = import_zod6.z.object({
+  request_id: import_zod6.z.string().nullable(),
+  transcript: import_zod6.z.string(),
+  language_code: import_zod6.z.string().nullable(),
+  timestamps: import_zod6.z.object({
+    end_time_seconds: import_zod6.z.array(import_zod6.z.number()),
+    start_time_seconds: import_zod6.z.array(import_zod6.z.number()),
+    words: import_zod6.z.array(import_zod6.z.string())
+  }).optional(),
+  diarized_transcript: import_zod6.z.object({
+    entries: import_zod6.z.array(
+      import_zod6.z.object({
+        end_time_seconds: import_zod6.z.array(import_zod6.z.number()),
+        start_time_seconds: import_zod6.z.array(import_zod6.z.number()),
+        transcript: import_zod6.z.string(),
+        speaker_id: import_zod6.z.string()
+      })
+    )
+  }).optional()
+});
+
+// src/sarvam-translation-model.ts
 var import_provider_utils7 = require("@ai-sdk/provider-utils");
 var import_zod8 = require("zod");
 
@@ -1065,7 +1065,110 @@ var SarvamLanguageCodeSchema = import_zod7.z.enum([
   "gu-IN"
 ]);
 
+// src/sarvam-translation-model.ts
+var SarvamTranslationModel = class {
+  constructor(settings, config) {
+    this.specificationVersion = "v1";
+    this.supportsStructuredOutputs = false;
+    this.defaultObjectGenerationMode = "json";
+    this.modelId = "unknown";
+    this.settings = settings;
+    this.config = config;
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  get supportsImageUrls() {
+    return false;
+  }
+  getArgs({
+    mode,
+    prompt
+  }) {
+    var _a, _b, _c, _d, _e, _f;
+    const type = mode.type;
+    const warnings = [];
+    if (this.settings.from !== "auto") {
+      if (this.settings.to !== "en-IN" && this.settings.from !== "en-IN")
+        throw new Error(
+          "Sarvam doesn't support Indic-Indic Transliteration yet"
+        );
+    }
+    if (type !== "regular") {
+      const _exhaustiveCheck = type;
+      throw new Error(`Unsupported type: ${_exhaustiveCheck}`);
+    }
+    const messages = convertToSarvamChatMessages(prompt);
+    return {
+      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",
+        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",
+      },
+      warnings
+    };
+  }
+  async doGenerate(options) {
+    var _b;
+    const { args, warnings, messages } = this.getArgs(__spreadProps(__spreadValues({}, options), {
+      stream: false
+    }));
+    const body = JSON.stringify(args);
+    const {
+      responseHeaders,
+      value: response,
+      rawValue: rawResponse
+    } = await (0, import_provider_utils7.postJsonToApi)({
+      url: this.config.url({
+        path: "/translate"
+      }),
+      headers: (0, import_provider_utils7.combineHeaders)(this.config.headers(), options.headers),
+      body: args,
+      failedResponseHandler: sarvamFailedResponseHandler,
+      successfulResponseHandler: (0, import_provider_utils7.createJsonResponseHandler)(
+        sarvamTranslationResponseSchema
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    const _a = args, { input: rawPrompt } = _a, rawSettings = __objRest(_a, ["input"]);
+    const text = (_b = response.translated_text) != null ? _b : void 0;
+    return {
+      text,
+      toolCalls: void 0,
+      reasoning: void 0,
+      finishReason: "unknown",
+      usage: {
+        promptTokens: NaN,
+        completionTokens: NaN
+      },
+      rawCall: { rawPrompt, rawSettings },
+      rawResponse: { headers: responseHeaders, body: rawResponse },
+      response: void 0,
+      warnings,
+      request: { body }
+    };
+  }
+  async doStream(options) {
+    throw new Error("Translation feature doesn't streaming yet");
+  }
+};
+var sarvamTranslationResponseSchema = import_zod8.z.object({
+  translated_text: import_zod8.z.string().nullish(),
+  source_language_code: SarvamLanguageCodeSchema.nullable(),
+  request_id: import_zod8.z.string().nullish()
+});
+
 // src/sarvam-transliterate-model.ts
+var import_provider_utils8 = require("@ai-sdk/provider-utils");
+var import_zod9 = require("zod");
 var SarvamTransliterateModel = class {
   constructor(settings, config) {
     this.specificationVersion = "v1";
@@ -1123,14 +1226,14 @@ var SarvamTransliterateModel = class {
       responseHeaders,
       value: response,
       rawValue: rawResponse
-    } = await (0, import_provider_utils7.postJsonToApi)({
+    } = await (0, import_provider_utils8.postJsonToApi)({
       url: this.config.url({
         path: "/transliterate"
       }),
-      headers: (0, import_provider_utils7.combineHeaders)(this.config.headers(), options.headers),
+      headers: (0, import_provider_utils8.combineHeaders)(this.config.headers(), options.headers),
       body: args,
       failedResponseHandler: sarvamFailedResponseHandler,
-      successfulResponseHandler: (0, import_provider_utils7.createJsonResponseHandler)(
+      successfulResponseHandler: (0, import_provider_utils8.createJsonResponseHandler)(
         sarvamTransliterateResponseSchema
       ),
       abortSignal: options.abortSignal,
@@ -1158,17 +1261,17 @@ var SarvamTransliterateModel = class {
     throw new Error("Transliterate feature doesn't streaming yet");
   }
 };
-var sarvamTransliterateResponseSchema = import_zod8.z.object({
-  transliterated_text: import_zod8.z.string().nullish(),
+var sarvamTransliterateResponseSchema = import_zod9.z.object({
+  transliterated_text: import_zod9.z.string().nullish(),
   source_language_code: SarvamLanguageCodeSchema.nullable(),
-  request_id: import_zod8.z.string().nullish()
+  request_id: import_zod9.z.string().nullish()
 });
 
 // src/sarvam-provider.ts
 function createSarvam(options = {}) {
   var _a;
-  const baseURL = (_a = (0, import_provider_utils8.withoutTrailingSlash)(options.baseURL)) != null ? _a : "https://api.sarvam.ai";
-  const ApiKey = (0, import_provider_utils8.loadApiKey)({
+  const baseURL = (_a = (0, import_provider_utils9.withoutTrailingSlash)(options.baseURL)) != null ? _a : "https://api.sarvam.ai";
+  const ApiKey = (0, import_provider_utils9.loadApiKey)({
     apiKey: options.apiKey,
     environmentVariableName: "SARVAM_API_KEY",
     description: "Sarvam"
@@ -1201,7 +1304,7 @@ function createSarvam(options = {}) {
     });
   };
   const createSpeechModel = (modelId, languageCode, settings) => new SarvamSpeechModel(modelId, languageCode, {
-    provider: `sarvam.speech`,
+    provider: "sarvam.speech",
     url: ({ path }) => `${baseURL}${path}`,
     headers: getHeaders,
     fetch: options.fetch,
@@ -1219,12 +1322,25 @@ function createSarvam(options = {}) {
       fetch: options.fetch
     }
   );
+  const createTranslationModel = (settings) => new SarvamTranslationModel(
+    {
+      to: settings.to,
+      from: settings.from
+    },
+    {
+      provider: "sarvam.translation",
+      url: ({ path }) => `${baseURL}${path}`,
+      headers: getHeaders,
+      fetch: options.fetch
+    }
+  );
   const provider = (modelId, settings) => createLanguageModel(modelId, settings);
   provider.languageModel = createLanguageModel;
   provider.chat = createChatModel;
   provider.transcription = createTranscriptionModel;
   provider.speech = createSpeechModel;
   provider.transliterate = createTransliterateModel;
+  provider.translation = createTranslationModel;
   return provider;
 }
 var sarvam = createSarvam();