sarvam-ai-sdk 0.0.1

package/dist/index.js

@@ -0,0 +1,1037 @@
+var __defProp = Object.defineProperty;
+var __defProps = Object.defineProperties;
+var __getOwnPropDescs = Object.getOwnPropertyDescriptors;
+var __getOwnPropSymbols = Object.getOwnPropertySymbols;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __propIsEnum = Object.prototype.propertyIsEnumerable;
+var __defNormalProp = (obj, key, value) => key in obj ? __defProp(obj, key, { enumerable: true, configurable: true, writable: true, value }) : obj[key] = value;
+var __spreadValues = (a, b) => {
+  for (var prop in b || (b = {}))
+    if (__hasOwnProp.call(b, prop))
+      __defNormalProp(a, prop, b[prop]);
+  if (__getOwnPropSymbols)
+    for (var prop of __getOwnPropSymbols(b)) {
+      if (__propIsEnum.call(b, prop))
+        __defNormalProp(a, prop, b[prop]);
+    }
+  return a;
+};
+var __spreadProps = (a, b) => __defProps(a, __getOwnPropDescs(b));
+var __objRest = (source, exclude) => {
+  var target = {};
+  for (var prop in source)
+    if (__hasOwnProp.call(source, prop) && exclude.indexOf(prop) < 0)
+      target[prop] = source[prop];
+  if (source != null && __getOwnPropSymbols)
+    for (var prop of __getOwnPropSymbols(source)) {
+      if (exclude.indexOf(prop) < 0 && __propIsEnum.call(source, prop))
+        target[prop] = source[prop];
+    }
+  return target;
+};
+
+// src/sarvam-provider.ts
+import {
+  loadApiKey,
+  withoutTrailingSlash
+} from "@ai-sdk/provider-utils";
+
+// src/sarvam-chat-language-model.ts
+import {
+  InvalidResponseDataError
+} from "@ai-sdk/provider";
+import {
+  combineHeaders,
+  createEventSourceResponseHandler,
+  createJsonResponseHandler,
+  generateId,
+  isParsableJson,
+  parseProviderOptions,
+  postJsonToApi
+} from "@ai-sdk/provider-utils";
+import { z as z2 } from "zod";
+
+// src/convert-to-sarvam-chat-messages.ts
+import {
+  UnsupportedFunctionalityError
+} from "@ai-sdk/provider";
+import { convertUint8ArrayToBase64 } from "@ai-sdk/provider-utils";
+function convertToSarvamChatMessages(prompt) {
+  const messages = [];
+  for (const { role, content } of prompt) {
+    switch (role) {
+      case "system": {
+        messages.push({ role: "system", content });
+        break;
+      }
+      case "user": {
+        if (content.length === 1 && content[0].type === "text") {
+          messages.push({ role: "user", content: content[0].text });
+          break;
+        }
+        messages.push({
+          role: "user",
+          content: content.map((part) => {
+            var _a;
+            switch (part.type) {
+              case "text": {
+                return { type: "text", text: part.text };
+              }
+              case "image": {
+                return {
+                  type: "image_url",
+                  image_url: {
+                    url: part.image instanceof URL ? part.image.toString() : `data:${(_a = part.mimeType) != null ? _a : "image/jpeg"};base64,${convertUint8ArrayToBase64(part.image)}`
+                  }
+                };
+              }
+              case "file": {
+                throw new UnsupportedFunctionalityError({
+                  functionality: "File content parts in user messages"
+                });
+              }
+            }
+          })
+        });
+        break;
+      }
+      case "assistant": {
+        let text = "";
+        const toolCalls = [];
+        for (const part of content) {
+          switch (part.type) {
+            case "text": {
+              text += part.text;
+              break;
+            }
+            case "tool-call": {
+              toolCalls.push({
+                id: part.toolCallId,
+                type: "function",
+                function: {
+                  name: part.toolName,
+                  arguments: JSON.stringify(part.args)
+                }
+              });
+              break;
+            }
+          }
+        }
+        messages.push({
+          role: "assistant",
+          content: text,
+          tool_calls: toolCalls.length > 0 ? toolCalls : void 0
+        });
+        break;
+      }
+      case "tool": {
+        for (const toolResponse of content) {
+          messages.push({
+            role: "tool",
+            tool_call_id: toolResponse.toolCallId,
+            content: JSON.stringify(toolResponse.result)
+          });
+        }
+        break;
+      }
+      default: {
+        const _exhaustiveCheck = role;
+        throw new Error(`Unsupported role: ${_exhaustiveCheck}`);
+      }
+    }
+  }
+  return messages;
+}
+
+// src/get-response-metadata.ts
+function getResponseMetadata({
+  id,
+  model,
+  created
+}) {
+  return {
+    id: id != null ? id : void 0,
+    modelId: model != null ? model : void 0,
+    timestamp: created != null ? new Date(created * 1e3) : void 0
+  };
+}
+
+// src/sarvam-error.ts
+import { z } from "zod";
+import { createJsonErrorResponseHandler } from "@ai-sdk/provider-utils";
+var sarvamErrorDataSchema = z.object({
+  error: z.object({
+    message: z.string(),
+    type: z.string()
+  })
+});
+var sarvamFailedResponseHandler = createJsonErrorResponseHandler({
+  errorSchema: sarvamErrorDataSchema,
+  errorToMessage: (data) => data.error.message
+});
+
+// src/sarvam-prepare-tools.ts
+import {
+  UnsupportedFunctionalityError as UnsupportedFunctionalityError2
+} from "@ai-sdk/provider";
+function prepareTools({
+  mode
+}) {
+  var _a;
+  const tools = ((_a = mode.tools) == null ? void 0 : _a.length) ? mode.tools : void 0;
+  const toolWarnings = [];
+  if (tools == null) {
+    return { tools: void 0, tool_choice: void 0, toolWarnings };
+  }
+  const toolChoice = mode.toolChoice;
+  const sarvamTools = [];
+  for (const tool of tools) {
+    if (tool.type === "provider-defined") {
+      toolWarnings.push({ type: "unsupported-tool", tool });
+    } else {
+      sarvamTools.push({
+        type: "function",
+        function: {
+          name: tool.name,
+          description: tool.description,
+          parameters: tool.parameters
+        }
+      });
+    }
+  }
+  if (toolChoice == null) {
+    return { tools: sarvamTools, tool_choice: void 0, toolWarnings };
+  }
+  const type = toolChoice.type;
+  switch (type) {
+    case "auto":
+    case "none":
+    case "required":
+      return { tools: sarvamTools, tool_choice: type, toolWarnings };
+    case "tool":
+      return {
+        tools: sarvamTools,
+        tool_choice: {
+          type: "function",
+          function: {
+            name: toolChoice.toolName
+          }
+        },
+        toolWarnings
+      };
+    default: {
+      const _exhaustiveCheck = type;
+      throw new UnsupportedFunctionalityError2({
+        functionality: `Unsupported tool choice type: ${_exhaustiveCheck}`
+      });
+    }
+  }
+}
+
+// src/map-sarvam-finish-reason.ts
+function mapSarvamFinishReason(finishReason) {
+  switch (finishReason) {
+    case "stop":
+      return "stop";
+    case "length":
+      return "length";
+    case "content_filter":
+      return "content-filter";
+    case "function_call":
+    case "tool_calls":
+      return "tool-calls";
+    default:
+      return "unknown";
+  }
+}
+
+// src/sarvam-chat-language-model.ts
+var SarvamChatLanguageModel = class {
+  constructor(modelId, settings, config) {
+    this.specificationVersion = "v1";
+    this.supportsStructuredOutputs = false;
+    this.defaultObjectGenerationMode = "json";
+    this.modelId = modelId;
+    this.settings = settings;
+    this.config = config;
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  get supportsImageUrls() {
+    return !this.settings.downloadImages;
+  }
+  getArgs({
+    mode,
+    prompt,
+    maxTokens,
+    temperature,
+    topP,
+    topK,
+    frequencyPenalty,
+    presencePenalty,
+    stopSequences,
+    responseFormat,
+    seed,
+    stream,
+    providerMetadata
+  }) {
+    const type = mode.type;
+    const warnings = [];
+    if (topK != null) {
+      warnings.push({
+        type: "unsupported-setting",
+        setting: "topK"
+      });
+    }
+    if (responseFormat != null && responseFormat.type === "json" && responseFormat.schema != null) {
+      warnings.push({
+        type: "unsupported-setting",
+        setting: "responseFormat",
+        details: "JSON response format schema is not supported"
+      });
+    }
+    const sarvamOptions = parseProviderOptions({
+      provider: "sarvam",
+      providerOptions: providerMetadata,
+      schema: z2.object({
+        reasoningFormat: z2.enum(["parsed", "raw", "hidden"]).nullish()
+      })
+    });
+    const baseArgs = {
+      // model id:
+      model: this.modelId,
+      // model specific settings:
+      user: this.settings.user,
+      parallel_tool_calls: this.settings.parallelToolCalls,
+      // standardized settings:
+      max_tokens: maxTokens,
+      temperature,
+      top_p: topP,
+      frequency_penalty: frequencyPenalty,
+      presence_penalty: presencePenalty,
+      stop: stopSequences,
+      seed,
+      // response format:
+      response_format: (
+        // json object response format is not supported for streaming:
+        stream === false && (responseFormat == null ? void 0 : responseFormat.type) === "json" ? { type: "json_object" } : void 0
+      ),
+      // provider options:
+      reasoning_format: sarvamOptions == null ? void 0 : sarvamOptions.reasoningFormat,
+      // messages:
+      messages: convertToSarvamChatMessages(prompt)
+    };
+    switch (type) {
+      case "regular": {
+        const { tools, tool_choice, toolWarnings } = prepareTools({
+          mode
+        });
+        return {
+          args: __spreadProps(__spreadValues({}, baseArgs), {
+            tools,
+            tool_choice
+          }),
+          warnings: [...warnings, ...toolWarnings]
+        };
+      }
+      case "object-json": {
+        return {
+          args: __spreadProps(__spreadValues({}, baseArgs), {
+            response_format: (
+              // json object response format is not supported for streaming:
+              stream === false ? { type: "json_object" } : void 0
+            )
+          }),
+          warnings
+        };
+      }
+      case "object-tool": {
+        return {
+          args: __spreadProps(__spreadValues({}, baseArgs), {
+            tool_choice: {
+              type: "function",
+              function: { name: mode.tool.name }
+            },
+            tools: [
+              {
+                type: "function",
+                function: {
+                  name: mode.tool.name,
+                  description: mode.tool.description,
+                  parameters: mode.tool.parameters
+                }
+              }
+            ]
+          }),
+          warnings
+        };
+      }
+      default: {
+        const _exhaustiveCheck = type;
+        throw new Error(`Unsupported type: ${_exhaustiveCheck}`);
+      }
+    }
+  }
+  async doGenerate(options) {
+    var _b, _c, _d, _e, _f, _g, _h;
+    const { args, warnings } = this.getArgs(__spreadProps(__spreadValues({}, options), { stream: false }));
+    const body = JSON.stringify(args);
+    const {
+      responseHeaders,
+      value: response,
+      rawValue: rawResponse
+    } = await postJsonToApi({
+      url: this.config.url({
+        path: "/chat/completions",
+        modelId: this.modelId
+      }),
+      headers: combineHeaders(this.config.headers(), options.headers),
+      body: args,
+      failedResponseHandler: sarvamFailedResponseHandler,
+      successfulResponseHandler: createJsonResponseHandler(
+        sarvamChatResponseSchema
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    const _a = args, { messages: rawPrompt } = _a, rawSettings = __objRest(_a, ["messages"]);
+    const choice = response.choices[0];
+    return {
+      text: (_b = choice.message.content) != null ? _b : void 0,
+      reasoning: (_c = choice.message.reasoning) != null ? _c : void 0,
+      toolCalls: (_d = choice.message.tool_calls) == null ? void 0 : _d.map((toolCall) => {
+        var _a2;
+        return {
+          toolCallType: "function",
+          toolCallId: (_a2 = toolCall.id) != null ? _a2 : generateId(),
+          toolName: toolCall.function.name,
+          args: toolCall.function.arguments
+        };
+      }),
+      finishReason: mapSarvamFinishReason(choice.finish_reason),
+      usage: {
+        promptTokens: (_f = (_e = response.usage) == null ? void 0 : _e.prompt_tokens) != null ? _f : NaN,
+        completionTokens: (_h = (_g = response.usage) == null ? void 0 : _g.completion_tokens) != null ? _h : NaN
+      },
+      rawCall: { rawPrompt, rawSettings },
+      rawResponse: { headers: responseHeaders, body: rawResponse },
+      response: getResponseMetadata(response),
+      warnings,
+      request: { body }
+    };
+  }
+  async doStream(options) {
+    const { args, warnings } = this.getArgs(__spreadProps(__spreadValues({}, options), { stream: true }));
+    const body = JSON.stringify(__spreadProps(__spreadValues({}, args), { stream: true }));
+    const { responseHeaders, value: response } = await postJsonToApi({
+      url: this.config.url({
+        path: "/chat/completions",
+        modelId: this.modelId
+      }),
+      headers: combineHeaders(this.config.headers(), options.headers),
+      body: __spreadProps(__spreadValues({}, args), {
+        stream: true
+      }),
+      failedResponseHandler: sarvamFailedResponseHandler,
+      successfulResponseHandler: createEventSourceResponseHandler(sarvamChatChunkSchema),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    const _a = args, { messages: rawPrompt } = _a, rawSettings = __objRest(_a, ["messages"]);
+    const toolCalls = [];
+    let finishReason = "unknown";
+    let usage = {
+      promptTokens: void 0,
+      completionTokens: void 0
+    };
+    let isFirstChunk = true;
+    let providerMetadata;
+    return {
+      stream: response.pipeThrough(
+        new TransformStream({
+          transform(chunk, controller) {
+            var _a2, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o;
+            if (!chunk.success) {
+              finishReason = "error";
+              controller.enqueue({
+                type: "error",
+                error: chunk.error
+              });
+              return;
+            }
+            const value = chunk.value;
+            if ("error" in value) {
+              finishReason = "error";
+              controller.enqueue({
+                type: "error",
+                error: value.error
+              });
+              return;
+            }
+            if (isFirstChunk) {
+              isFirstChunk = false;
+              controller.enqueue(__spreadValues({
+                type: "response-metadata"
+              }, getResponseMetadata(value)));
+            }
+            if (((_a2 = value.x_sarvam) == null ? void 0 : _a2.usage) != null) {
+              usage = {
+                promptTokens: (_b = value.x_sarvam.usage.prompt_tokens) != null ? _b : void 0,
+                completionTokens: (_c = value.x_sarvam.usage.completion_tokens) != null ? _c : void 0
+              };
+            }
+            const choice = value.choices[0];
+            if ((choice == null ? void 0 : choice.finish_reason) != null) {
+              finishReason = mapSarvamFinishReason(
+                choice.finish_reason
+              );
+            }
+            if ((choice == null ? void 0 : choice.delta) == null) {
+              return;
+            }
+            const delta = choice.delta;
+            if (delta.reasoning != null && delta.reasoning.length > 0) {
+              controller.enqueue({
+                type: "reasoning",
+                textDelta: delta.reasoning
+              });
+            }
+            if (delta.content != null && delta.content.length > 0) {
+              controller.enqueue({
+                type: "text-delta",
+                textDelta: delta.content
+              });
+            }
+            if (delta.tool_calls != null) {
+              for (const toolCallDelta of delta.tool_calls) {
+                const index = toolCallDelta.index;
+                if (toolCalls[index] == null) {
+                  if (toolCallDelta.type !== "function") {
+                    throw new InvalidResponseDataError({
+                      data: toolCallDelta,
+                      message: `Expected 'function' type.`
+                    });
+                  }
+                  if (toolCallDelta.id == null) {
+                    throw new InvalidResponseDataError({
+                      data: toolCallDelta,
+                      message: `Expected 'id' to be a string.`
+                    });
+                  }
+                  if (((_d = toolCallDelta.function) == null ? void 0 : _d.name) == null) {
+                    throw new InvalidResponseDataError({
+                      data: toolCallDelta,
+                      message: `Expected 'function.name' to be a string.`
+                    });
+                  }
+                  toolCalls[index] = {
+                    id: toolCallDelta.id,
+                    type: "function",
+                    function: {
+                      name: toolCallDelta.function.name,
+                      arguments: (_e = toolCallDelta.function.arguments) != null ? _e : ""
+                    },
+                    hasFinished: false
+                  };
+                  const toolCall2 = toolCalls[index];
+                  if (((_f = toolCall2.function) == null ? void 0 : _f.name) != null && ((_g = toolCall2.function) == null ? void 0 : _g.arguments) != null) {
+                    if (toolCall2.function.arguments.length > 0) {
+                      controller.enqueue({
+                        type: "tool-call-delta",
+                        toolCallType: "function",
+                        toolCallId: toolCall2.id,
+                        toolName: toolCall2.function.name,
+                        argsTextDelta: toolCall2.function.arguments
+                      });
+                    }
+                    if (isParsableJson(
+                      toolCall2.function.arguments
+                    )) {
+                      controller.enqueue({
+                        type: "tool-call",
+                        toolCallType: "function",
+                        toolCallId: (_h = toolCall2.id) != null ? _h : generateId(),
+                        toolName: toolCall2.function.name,
+                        args: toolCall2.function.arguments
+                      });
+                      toolCall2.hasFinished = true;
+                    }
+                  }
+                  continue;
+                }
+                const toolCall = toolCalls[index];
+                if (toolCall.hasFinished) {
+                  continue;
+                }
+                if (((_i = toolCallDelta.function) == null ? void 0 : _i.arguments) != null) {
+                  toolCall.function.arguments += (_k = (_j = toolCallDelta.function) == null ? void 0 : _j.arguments) != null ? _k : "";
+                }
+                controller.enqueue({
+                  type: "tool-call-delta",
+                  toolCallType: "function",
+                  toolCallId: toolCall.id,
+                  toolName: toolCall.function.name,
+                  argsTextDelta: (_l = toolCallDelta.function.arguments) != null ? _l : ""
+                });
+                if (((_m = toolCall.function) == null ? void 0 : _m.name) != null && ((_n = toolCall.function) == null ? void 0 : _n.arguments) != null && isParsableJson(toolCall.function.arguments)) {
+                  controller.enqueue({
+                    type: "tool-call",
+                    toolCallType: "function",
+                    toolCallId: (_o = toolCall.id) != null ? _o : generateId(),
+                    toolName: toolCall.function.name,
+                    args: toolCall.function.arguments
+                  });
+                  toolCall.hasFinished = true;
+                }
+              }
+            }
+          },
+          flush(controller) {
+            var _a2, _b;
+            controller.enqueue(__spreadValues({
+              type: "finish",
+              finishReason,
+              usage: {
+                promptTokens: (_a2 = usage.promptTokens) != null ? _a2 : NaN,
+                completionTokens: (_b = usage.completionTokens) != null ? _b : NaN
+              }
+            }, providerMetadata != null ? { providerMetadata } : {}));
+          }
+        })
+      ),
+      rawCall: { rawPrompt, rawSettings },
+      rawResponse: { headers: responseHeaders },
+      warnings,
+      request: { body }
+    };
+  }
+};
+var sarvamChatResponseSchema = z2.object({
+  id: z2.string().nullish(),
+  created: z2.number().nullish(),
+  model: z2.string().nullish(),
+  choices: z2.array(
+    z2.object({
+      message: z2.object({
+        content: z2.string().nullish(),
+        reasoning: z2.string().nullish(),
+        tool_calls: z2.array(
+          z2.object({
+            id: z2.string().nullish(),
+            type: z2.literal("function"),
+            function: z2.object({
+              name: z2.string(),
+              arguments: z2.string()
+            })
+          })
+        ).nullish()
+      }),
+      index: z2.number(),
+      finish_reason: z2.string().nullish()
+    })
+  ),
+  usage: z2.object({
+    prompt_tokens: z2.number().nullish(),
+    completion_tokens: z2.number().nullish()
+  }).nullish()
+});
+var sarvamChatChunkSchema = z2.union([
+  z2.object({
+    id: z2.string().nullish(),
+    created: z2.number().nullish(),
+    model: z2.string().nullish(),
+    choices: z2.array(
+      z2.object({
+        delta: z2.object({
+          content: z2.string().nullish(),
+          reasoning: z2.string().nullish(),
+          tool_calls: z2.array(
+            z2.object({
+              index: z2.number(),
+              id: z2.string().nullish(),
+              type: z2.literal("function").optional(),
+              function: z2.object({
+                name: z2.string().nullish(),
+                arguments: z2.string().nullish()
+              })
+            })
+          ).nullish()
+        }).nullish(),
+        finish_reason: z2.string().nullable().optional(),
+        index: z2.number()
+      })
+    ),
+    x_sarvam: z2.object({
+      usage: z2.object({
+        prompt_tokens: z2.number().nullish(),
+        completion_tokens: z2.number().nullish()
+      }).nullish()
+    }).nullish()
+  }),
+  sarvamErrorDataSchema
+]);
+
+// src/sarvam-transcription-model.ts
+import {
+  combineHeaders as combineHeaders2,
+  createJsonResponseHandler as createJsonResponseHandler2,
+  parseProviderOptions as parseProviderOptions2,
+  postFormDataToApi
+} from "@ai-sdk/provider-utils";
+import { z as z4 } from "zod";
+
+// src/sarvam-transcription-settings.ts
+import { z as z3 } from "zod";
+var SarvamProviderOptionsSchema = z3.object({
+  with_timestamps: z3.boolean().nullish().default(false),
+  /**
+   * Enables speaker diarization, which identifies and separates different speakers in the audio.
+   * When set to true, the API will provide speaker-specific segments in the response.
+   * Note: This parameter is currently in Beta mode.
+   */
+  with_diarization: z3.boolean().nullish().default(false),
+  /**
+   * Number of speakers to be detected in the audio.
+   * This is used when with_diarization is set to true.
+   * Can be null.
+   */
+  num_speakers: z3.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 = parseProviderOptions2({
+      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 postFormDataToApi({
+      url: this.config.url({
+        path: "/speech-to-text",
+        modelId: this.modelId
+      }),
+      headers: combineHeaders2(this.config.headers(), options.headers),
+      formData,
+      failedResponseHandler: sarvamFailedResponseHandler,
+      successfulResponseHandler: createJsonResponseHandler2(
+        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 = z4.object({
+  request_id: z4.string().nullable(),
+  transcript: z4.string(),
+  language_code: z4.string().nullable(),
+  timestamps: z4.object({
+    end_time_seconds: z4.array(z4.number()),
+    start_time_seconds: z4.array(z4.number()),
+    words: z4.array(z4.string())
+  }).optional(),
+  diarized_transcript: z4.object({
+    entries: z4.array(
+      z4.object({
+        end_time_seconds: z4.array(z4.number()),
+        start_time_seconds: z4.array(z4.number()),
+        transcript: z4.string(),
+        speaker_id: z4.string()
+      })
+    )
+  }).optional()
+});
+
+// src/sarvam-speech-model.ts
+import {
+  combineHeaders as combineHeaders3,
+  createJsonResponseHandler as createJsonResponseHandler3,
+  parseProviderOptions as parseProviderOptions3,
+  postJsonToApi as postJsonToApi2
+} from "@ai-sdk/provider-utils";
+
+// src/sarvam-speech-settings.ts
+import { z as z5 } from "zod";
+var SpeakerSchema = z5.enum([
+  "meera",
+  "pavithra",
+  "maitreyi",
+  "arvind",
+  "amol",
+  "amartya",
+  "diya",
+  "neel",
+  "misha",
+  "vian",
+  "arjun",
+  "maya",
+  "anushka",
+  "abhilash",
+  "manisha",
+  "vidya",
+  "arya",
+  "karun",
+  "hitesh"
+]).default("meera");
+var SarvamProviderOptionsSchema2 = z5.object({
+  speaker: SpeakerSchema,
+  pitch: z5.number().min(-0.75).max(0.75).default(0),
+  pace: z5.number().min(0.5).max(2).default(1),
+  loudness: z5.number().min(0.3).max(3).default(1),
+  speech_sample_rate: z5.union([
+    z5.literal(8e3),
+    z5.literal(16e3),
+    z5.literal(22050),
+    z5.literal(24e3)
+  ]).default(22050),
+  enable_preprocessing: z5.boolean().default(false)
+}).partial();
+
+// src/sarvam-speech-model.ts
+import { z as z6 } from "zod";
+var SarvamSpeechModel = class {
+  constructor(modelId, languageCode, config) {
+    this.modelId = modelId;
+    this.languageCode = languageCode;
+    this.config = config;
+    this.specificationVersion = "v1";
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  getArgs({
+    text,
+    voice,
+    outputFormat = "wav",
+    // speed,
+    // instructions,
+    providerOptions
+  }) {
+    const warnings = [];
+    const sarvamOptions = parseProviderOptions3({
+      provider: "sarvam",
+      providerOptions: {
+        sarvam: __spreadValues(__spreadValues({}, providerOptions == null ? void 0 : providerOptions.sarvam), this.config.speech)
+      },
+      schema: SarvamProviderOptionsSchema2
+    });
+    const getSpeaker = () => {
+      if (sarvamOptions == null ? void 0 : sarvamOptions.speaker) return sarvamOptions.speaker;
+      if (voice) {
+        return SpeakerSchema.parse(voice);
+      }
+      switch (this.modelId) {
+        case "bulbul:v1":
+          return "meera";
+        case "bulbul:v2":
+          return "manisha";
+      }
+      return "meera";
+    };
+    const requestBody = {
+      model: this.modelId,
+      text,
+      target_language_code: this.languageCode,
+      speaker: getSpeaker()
+      // response_format: "wav",
+      // speed,
+      // instructions,
+    };
+    if (outputFormat) {
+      if (["mp3", "opus", "aac", "flac", "wav", "pcm"].includes(
+        outputFormat
+      )) {
+        requestBody.response_format = outputFormat;
+      } else {
+        warnings.push({
+          type: "unsupported-setting",
+          setting: "outputFormat",
+          details: `Unsupported output format: ${outputFormat}. Using mp3 instead.`
+        });
+      }
+    }
+    if (sarvamOptions) {
+      const speechModelOptions = {};
+      for (const key in speechModelOptions) {
+        const value = speechModelOptions[key];
+        if (value !== void 0) {
+          requestBody[key] = value;
+        }
+      }
+    }
+    return {
+      requestBody,
+      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 { requestBody, warnings } = this.getArgs(options);
+    const {
+      value,
+      responseHeaders,
+      rawValue: rawResponse
+    } = await postJsonToApi2({
+      url: this.config.url({
+        path: "/text-to-speech",
+        modelId: this.modelId
+      }),
+      headers: combineHeaders3(this.config.headers(), options.headers),
+      body: requestBody,
+      failedResponseHandler: sarvamFailedResponseHandler,
+      successfulResponseHandler: createJsonResponseHandler3(
+        z6.object({
+          request_id: z6.string(),
+          audios: z6.array(z6.string())
+        })
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    const audio = value.audios[0];
+    return {
+      audio,
+      warnings,
+      request: {
+        body: JSON.stringify(requestBody)
+      },
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: rawResponse
+      }
+    };
+  }
+};
+
+// src/sarvam-provider.ts
+function createSarvam(options = {}) {
+  var _a;
+  const baseURL = (_a = withoutTrailingSlash(options.baseURL)) != null ? _a : "https://api.sarvam.ai";
+  const ApiKey = loadApiKey({
+    apiKey: options.apiKey,
+    environmentVariableName: "SARVAM_API_KEY",
+    description: "Sarvam"
+  });
+  const getHeaders = () => __spreadValues({
+    Authorization: `Bearer ${ApiKey}`,
+    "api-subscription-key": ApiKey
+  }, options.headers);
+  const createChatModel = (modelId, settings = {}) => new SarvamChatLanguageModel(modelId, settings, {
+    provider: "sarvam.chat",
+    url: ({ path }) => `${baseURL}/v1${path}`,
+    headers: getHeaders,
+    fetch: options.fetch
+  });
+  const createLanguageModel = (modelId, settings) => {
+    if (new.target) {
+      throw new Error(
+        "The Sarvam model function cannot be called with the new keyword."
+      );
+    }
+    return createChatModel(modelId, settings);
+  };
+  const createTranscriptionModel = (modelId, languageCode = "unknown", settings) => {
+    return new SarvamTranscriptionModel(modelId, languageCode, {
+      provider: "sarvam.transcription",
+      url: ({ path }) => `${baseURL}${path}`,
+      headers: getHeaders,
+      fetch: options.fetch,
+      transcription: settings
+    });
+  };
+  const createSpeechModel = (modelId, languageCode, settings) => new SarvamSpeechModel(modelId, languageCode, {
+    provider: `sarvam.speech`,
+    url: ({ path }) => `${baseURL}${path}`,
+    headers: getHeaders,
+    fetch: options.fetch,
+    speech: settings
+  });
+  const provider = (modelId, settings) => createLanguageModel(modelId, settings);
+  provider.languageModel = createLanguageModel;
+  provider.chat = createChatModel;
+  provider.transcription = createTranscriptionModel;
+  provider.speech = createSpeechModel;
+  return provider;
+}
+var sarvam = createSarvam();
+export {
+  createSarvam,
+  sarvam
+};
+//# sourceMappingURL=index.js.map