@ai-sdk/groq 2.0.0-canary.1 → 2.0.0-canary.11

package/dist/index.mjs

@@ -20,13 +20,12 @@ import {
   parseProviderOptions,
   postJsonToApi
 } from "@ai-sdk/provider-utils";
-import { z as z2 } from "zod";
+import { z as z3 } from "zod";
 
 // src/convert-to-groq-chat-messages.ts
 import {
   UnsupportedFunctionalityError
 } from "@ai-sdk/provider";
-import { convertUint8ArrayToBase64 } from "@ai-sdk/provider-utils";
 function convertToGroqChatMessages(prompt) {
   const messages = [];
   for (const { role, content } of prompt) {
@@ -43,24 +42,24 @@ function convertToGroqChatMessages(prompt) {
         messages.push({
           role: "user",
           content: content.map((part) => {
-            var _a;
             switch (part.type) {
               case "text": {
                 return { type: "text", text: part.text };
               }
-              case "image": {
+              case "file": {
+                if (!part.mediaType.startsWith("image/")) {
+                  throw new UnsupportedFunctionalityError({
+                    functionality: "Non-image file content parts"
+                  });
+                }
+                const mediaType = part.mediaType === "image/*" ? "image/jpeg" : part.mediaType;
                 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)}`
+                    url: part.data instanceof URL ? part.data.toString() : `data:${mediaType};base64,${part.data}`
                   }
                 };
               }
-              case "file": {
-                throw new UnsupportedFunctionalityError({
-                  functionality: "File content parts in user messages"
-                });
-              }
             }
           })
         });
@@ -127,13 +126,28 @@ function getResponseMetadata({
   };
 }
 
-// src/groq-error.ts
+// src/groq-chat-options.ts
 import { z } from "zod";
+var groqProviderOptions = z.object({
+  reasoningFormat: z.enum(["parsed", "raw", "hidden"]).nullish(),
+  /**
+   * Whether to enable parallel function calling during tool use. Default to true.
+   */
+  parallelToolCalls: z.boolean().nullish(),
+  /**
+   * A unique identifier representing your end-user, which can help OpenAI to
+   * monitor and detect abuse. Learn more.
+   */
+  user: z.string().nullish()
+});
+
+// src/groq-error.ts
+import { z as z2 } from "zod";
 import { createJsonErrorResponseHandler } from "@ai-sdk/provider-utils";
-var groqErrorDataSchema = z.object({
-  error: z.object({
-    message: z.string(),
-    type: z.string()
+var groqErrorDataSchema = z2.object({
+  error: z2.object({
+    message: z2.string(),
+    type: z2.string()
   })
 });
 var groqFailedResponseHandler = createJsonErrorResponseHandler({
@@ -146,15 +160,14 @@ import {
   UnsupportedFunctionalityError as UnsupportedFunctionalityError2
 } from "@ai-sdk/provider";
 function prepareTools({
-  mode
+  tools,
+  toolChoice
 }) {
-  var _a;
-  const tools = ((_a = mode.tools) == null ? void 0 : _a.length) ? mode.tools : void 0;
+  tools = (tools == null ? void 0 : tools.length) ? tools : void 0;
   const toolWarnings = [];
   if (tools == null) {
-    return { tools: void 0, tool_choice: void 0, toolWarnings };
+    return { tools: void 0, toolChoice: void 0, toolWarnings };
   }
-  const toolChoice = mode.toolChoice;
   const groqTools = [];
   for (const tool of tools) {
     if (tool.type === "provider-defined") {
@@ -171,18 +184,18 @@ function prepareTools({
     }
   }
   if (toolChoice == null) {
-    return { tools: groqTools, tool_choice: void 0, toolWarnings };
+    return { tools: groqTools, toolChoice: void 0, toolWarnings };
   }
   const type = toolChoice.type;
   switch (type) {
     case "auto":
     case "none":
     case "required":
-      return { tools: groqTools, tool_choice: type, toolWarnings };
+      return { tools: groqTools, toolChoice: type, toolWarnings };
     case "tool":
       return {
         tools: groqTools,
-        tool_choice: {
+        toolChoice: {
           type: "function",
           function: {
             name: toolChoice.toolName
@@ -193,7 +206,7 @@ function prepareTools({
     default: {
       const _exhaustiveCheck = type;
       throw new UnsupportedFunctionalityError2({
-        functionality: `Unsupported tool choice type: ${_exhaustiveCheck}`
+        functionality: `tool choice type: ${_exhaustiveCheck}`
       });
     }
   }
@@ -218,24 +231,22 @@ function mapGroqFinishReason(finishReason) {
 
 // src/groq-chat-language-model.ts
 var GroqChatLanguageModel = class {
-  constructor(modelId, settings, config) {
+  constructor(modelId, config) {
     this.specificationVersion = "v2";
-    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;
+  async getSupportedUrls() {
+    return {
+      "image/*": [/^https:\/\/.*$/]
+    };
   }
   getArgs({
-    mode,
     prompt,
-    maxTokens,
+    maxOutputTokens,
     temperature,
     topP,
     topK,
@@ -245,9 +256,10 @@ var GroqChatLanguageModel = class {
     responseFormat,
     seed,
     stream,
-    providerMetadata
+    tools,
+    toolChoice,
+    providerOptions
   }) {
-    const type = mode.type;
     const warnings = [];
     if (topK != null) {
       warnings.push({
@@ -264,89 +276,47 @@ var GroqChatLanguageModel = class {
     }
     const groqOptions = parseProviderOptions({
       provider: "groq",
-      providerOptions: providerMetadata,
-      schema: z2.object({
-        reasoningFormat: z2.enum(["parsed", "raw", "hidden"]).nullish()
-      })
+      providerOptions,
+      schema: groqProviderOptions
     });
-    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: groqOptions == null ? void 0 : groqOptions.reasoningFormat,
-      // messages:
-      messages: convertToGroqChatMessages(prompt)
+    const {
+      tools: groqTools,
+      toolChoice: groqToolChoice,
+      toolWarnings
+    } = prepareTools({ tools, toolChoice });
+    return {
+      args: {
+        // model id:
+        model: this.modelId,
+        // model specific settings:
+        user: groqOptions == null ? void 0 : groqOptions.user,
+        parallel_tool_calls: groqOptions == null ? void 0 : groqOptions.parallelToolCalls,
+        // standardized settings:
+        max_tokens: maxOutputTokens,
+        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: groqOptions == null ? void 0 : groqOptions.reasoningFormat,
+        // messages:
+        messages: convertToGroqChatMessages(prompt),
+        // tools:
+        tools: groqTools,
+        tool_choice: groqToolChoice
+      },
+      warnings: [...warnings, ...toolWarnings]
     };
-    switch (type) {
-      case "regular": {
-        const { tools, tool_choice, toolWarnings } = prepareTools({ mode });
-        return {
-          args: {
-            ...baseArgs,
-            tools,
-            tool_choice
-          },
-          warnings: [...warnings, ...toolWarnings]
-        };
-      }
-      case "object-json": {
-        return {
-          args: {
-            ...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: {
-            ...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 _a, _b, _c, _d, _e, _f, _g;
+    var _a, _b, _c, _d, _e;
     const { args, warnings } = this.getArgs({ ...options, stream: false });
     const body = JSON.stringify(args);
     const {
@@ -367,28 +337,43 @@ var GroqChatLanguageModel = class {
       abortSignal: options.abortSignal,
       fetch: this.config.fetch
     });
-    const { messages: rawPrompt, ...rawSettings } = args;
     const choice = response.choices[0];
-    return {
-      text: (_a = choice.message.content) != null ? _a : void 0,
-      reasoning: (_b = choice.message.reasoning) != null ? _b : void 0,
-      toolCalls: (_c = choice.message.tool_calls) == null ? void 0 : _c.map((toolCall) => {
-        var _a2;
-        return {
+    const content = [];
+    const text = choice.message.content;
+    if (text != null && text.length > 0) {
+      content.push({ type: "text", text });
+    }
+    const reasoning = choice.message.reasoning;
+    if (reasoning != null && reasoning.length > 0) {
+      content.push({
+        type: "reasoning",
+        reasoningType: "text",
+        text: reasoning
+      });
+    }
+    if (choice.message.tool_calls != null) {
+      for (const toolCall of choice.message.tool_calls) {
+        content.push({
+          type: "tool-call",
           toolCallType: "function",
-          toolCallId: (_a2 = toolCall.id) != null ? _a2 : generateId(),
+          toolCallId: (_a = toolCall.id) != null ? _a : generateId(),
           toolName: toolCall.function.name,
           args: toolCall.function.arguments
-        };
-      }),
+        });
+      }
+    }
+    return {
+      content,
       finishReason: mapGroqFinishReason(choice.finish_reason),
       usage: {
-        promptTokens: (_e = (_d = response.usage) == null ? void 0 : _d.prompt_tokens) != null ? _e : NaN,
-        completionTokens: (_g = (_f = response.usage) == null ? void 0 : _f.completion_tokens) != null ? _g : NaN
+        inputTokens: (_c = (_b = response.usage) == null ? void 0 : _b.prompt_tokens) != null ? _c : void 0,
+        outputTokens: (_e = (_d = response.usage) == null ? void 0 : _d.completion_tokens) != null ? _e : void 0
+      },
+      response: {
+        ...getResponseMetadata(response),
+        headers: responseHeaders,
+        body: rawResponse
       },
-      rawCall: { rawPrompt, rawSettings },
-      rawResponse: { headers: responseHeaders, body: rawResponse },
-      response: getResponseMetadata(response),
       warnings,
       request: { body }
     };
@@ -411,18 +396,20 @@ var GroqChatLanguageModel = class {
       abortSignal: options.abortSignal,
       fetch: this.config.fetch
     });
-    const { messages: rawPrompt, ...rawSettings } = args;
     const toolCalls = [];
     let finishReason = "unknown";
-    let usage = {
-      promptTokens: void 0,
-      completionTokens: void 0
+    const usage = {
+      inputTokens: void 0,
+      outputTokens: void 0
     };
     let isFirstChunk = true;
     let providerMetadata;
     return {
       stream: response.pipeThrough(
         new TransformStream({
+          start(controller) {
+            controller.enqueue({ type: "stream-start", warnings });
+          },
           transform(chunk, controller) {
             var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o;
             if (!chunk.success) {
@@ -444,10 +431,8 @@ var GroqChatLanguageModel = class {
               });
             }
             if (((_a = value.x_groq) == null ? void 0 : _a.usage) != null) {
-              usage = {
-                promptTokens: (_b = value.x_groq.usage.prompt_tokens) != null ? _b : void 0,
-                completionTokens: (_c = value.x_groq.usage.completion_tokens) != null ? _c : void 0
-              };
+              usage.inputTokens = (_b = value.x_groq.usage.prompt_tokens) != null ? _b : void 0;
+              usage.outputTokens = (_c = value.x_groq.usage.completion_tokens) != null ? _c : void 0;
             }
             const choice = value.choices[0];
             if ((choice == null ? void 0 : choice.finish_reason) != null) {
@@ -460,13 +445,14 @@ var GroqChatLanguageModel = class {
             if (delta.reasoning != null && delta.reasoning.length > 0) {
               controller.enqueue({
                 type: "reasoning",
-                textDelta: delta.reasoning
+                reasoningType: "text",
+                text: delta.reasoning
               });
             }
             if (delta.content != null && delta.content.length > 0) {
               controller.enqueue({
-                type: "text-delta",
-                textDelta: delta.content
+                type: "text",
+                text: delta.content
               });
             }
             if (delta.tool_calls != null) {
@@ -552,91 +538,211 @@ var GroqChatLanguageModel = class {
             }
           },
           flush(controller) {
-            var _a, _b;
             controller.enqueue({
               type: "finish",
               finishReason,
-              usage: {
-                promptTokens: (_a = usage.promptTokens) != null ? _a : NaN,
-                completionTokens: (_b = usage.completionTokens) != null ? _b : NaN
-              },
+              usage,
               ...providerMetadata != null ? { providerMetadata } : {}
             });
           }
         })
       ),
-      rawCall: { rawPrompt, rawSettings },
-      rawResponse: { headers: responseHeaders },
-      warnings,
-      request: { body }
+      request: { body },
+      response: { headers: responseHeaders }
     };
   }
 };
-var groqChatResponseSchema = 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()
+var groqChatResponseSchema = z3.object({
+  id: z3.string().nullish(),
+  created: z3.number().nullish(),
+  model: z3.string().nullish(),
+  choices: z3.array(
+    z3.object({
+      message: z3.object({
+        content: z3.string().nullish(),
+        reasoning: z3.string().nullish(),
+        tool_calls: z3.array(
+          z3.object({
+            id: z3.string().nullish(),
+            type: z3.literal("function"),
+            function: z3.object({
+              name: z3.string(),
+              arguments: z3.string()
             })
           })
         ).nullish()
       }),
-      index: z2.number(),
-      finish_reason: z2.string().nullish()
+      index: z3.number(),
+      finish_reason: z3.string().nullish()
     })
   ),
-  usage: z2.object({
-    prompt_tokens: z2.number().nullish(),
-    completion_tokens: z2.number().nullish()
+  usage: z3.object({
+    prompt_tokens: z3.number().nullish(),
+    completion_tokens: z3.number().nullish()
   }).nullish()
 });
-var groqChatChunkSchema = 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()
+var groqChatChunkSchema = z3.union([
+  z3.object({
+    id: z3.string().nullish(),
+    created: z3.number().nullish(),
+    model: z3.string().nullish(),
+    choices: z3.array(
+      z3.object({
+        delta: z3.object({
+          content: z3.string().nullish(),
+          reasoning: z3.string().nullish(),
+          tool_calls: z3.array(
+            z3.object({
+              index: z3.number(),
+              id: z3.string().nullish(),
+              type: z3.literal("function").optional(),
+              function: z3.object({
+                name: z3.string().nullish(),
+                arguments: z3.string().nullish()
               })
             })
           ).nullish()
         }).nullish(),
-        finish_reason: z2.string().nullable().optional(),
-        index: z2.number()
+        finish_reason: z3.string().nullable().optional(),
+        index: z3.number()
       })
     ),
-    x_groq: z2.object({
-      usage: z2.object({
-        prompt_tokens: z2.number().nullish(),
-        completion_tokens: z2.number().nullish()
+    x_groq: z3.object({
+      usage: z3.object({
+        prompt_tokens: z3.number().nullish(),
+        completion_tokens: z3.number().nullish()
       }).nullish()
     }).nullish()
   }),
   groqErrorDataSchema
 ]);
 
+// src/groq-transcription-model.ts
+import {
+  combineHeaders as combineHeaders2,
+  convertBase64ToUint8Array,
+  createJsonResponseHandler as createJsonResponseHandler2,
+  parseProviderOptions as parseProviderOptions2,
+  postFormDataToApi
+} from "@ai-sdk/provider-utils";
+import { z as z4 } from "zod";
+var groqProviderOptionsSchema = z4.object({
+  language: z4.string().nullish(),
+  prompt: z4.string().nullish(),
+  responseFormat: z4.string().nullish(),
+  temperature: z4.number().min(0).max(1).nullish(),
+  timestampGranularities: z4.array(z4.string()).nullish()
+});
+var GroqTranscriptionModel = class {
+  constructor(modelId, config) {
+    this.modelId = modelId;
+    this.config = config;
+    this.specificationVersion = "v1";
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  getArgs({
+    audio,
+    mediaType,
+    providerOptions
+  }) {
+    var _a, _b, _c, _d, _e;
+    const warnings = [];
+    const groqOptions = parseProviderOptions2({
+      provider: "groq",
+      providerOptions,
+      schema: groqProviderOptionsSchema
+    });
+    const formData = new FormData();
+    const blob = audio instanceof Uint8Array ? new Blob([audio]) : new Blob([convertBase64ToUint8Array(audio)]);
+    formData.append("model", this.modelId);
+    formData.append("file", new File([blob], "audio", { type: mediaType }));
+    if (groqOptions) {
+      const transcriptionModelOptions = {
+        language: (_a = groqOptions.language) != null ? _a : void 0,
+        prompt: (_b = groqOptions.prompt) != null ? _b : void 0,
+        response_format: (_c = groqOptions.responseFormat) != null ? _c : void 0,
+        temperature: (_d = groqOptions.temperature) != null ? _d : void 0,
+        timestamp_granularities: (_e = groqOptions.timestampGranularities) != null ? _e : void 0
+      };
+      for (const key in transcriptionModelOptions) {
+        const value = transcriptionModelOptions[key];
+        if (value !== void 0) {
+          formData.append(key, String(value));
+        }
+      }
+    }
+    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: "/audio/transcriptions",
+        modelId: this.modelId
+      }),
+      headers: combineHeaders2(this.config.headers(), options.headers),
+      formData,
+      failedResponseHandler: groqFailedResponseHandler,
+      successfulResponseHandler: createJsonResponseHandler2(
+        groqTranscriptionResponseSchema
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    return {
+      text: response.text,
+      segments: (_e = (_d = response.segments) == null ? void 0 : _d.map((segment) => ({
+        text: segment.text,
+        startSecond: segment.start,
+        endSecond: segment.end
+      }))) != null ? _e : [],
+      language: response.language,
+      durationInSeconds: response.duration,
+      warnings,
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: rawResponse
+      }
+    };
+  }
+};
+var groqTranscriptionResponseSchema = z4.object({
+  task: z4.string(),
+  language: z4.string(),
+  duration: z4.number(),
+  text: z4.string(),
+  segments: z4.array(
+    z4.object({
+      id: z4.number(),
+      seek: z4.number(),
+      start: z4.number(),
+      end: z4.number(),
+      text: z4.string(),
+      tokens: z4.array(z4.number()),
+      temperature: z4.number(),
+      avg_logprob: z4.number(),
+      compression_ratio: z4.number(),
+      no_speech_prob: z4.number()
+    })
+  ),
+  x_groq: z4.object({
+    id: z4.string()
+  })
+});
+
 // src/groq-provider.ts
 function createGroq(options = {}) {
   var _a;
@@ -649,22 +755,30 @@ function createGroq(options = {}) {
     })}`,
     ...options.headers
   });
-  const createChatModel = (modelId, settings = {}) => new GroqChatLanguageModel(modelId, settings, {
+  const createChatModel = (modelId) => new GroqChatLanguageModel(modelId, {
     provider: "groq.chat",
     url: ({ path }) => `${baseURL}${path}`,
     headers: getHeaders,
     fetch: options.fetch
   });
-  const createLanguageModel = (modelId, settings) => {
+  const createLanguageModel = (modelId) => {
     if (new.target) {
       throw new Error(
         "The Groq model function cannot be called with the new keyword."
       );
     }
-    return createChatModel(modelId, settings);
+    return createChatModel(modelId);
+  };
+  const createTranscriptionModel = (modelId) => {
+    return new GroqTranscriptionModel(modelId, {
+      provider: "groq.transcription",
+      url: ({ path }) => `${baseURL}${path}`,
+      headers: getHeaders,
+      fetch: options.fetch
+    });
   };
-  const provider = function(modelId, settings) {
-    return createLanguageModel(modelId, settings);
+  const provider = function(modelId) {
+    return createLanguageModel(modelId);
   };
   provider.languageModel = createLanguageModel;
   provider.chat = createChatModel;
@@ -674,6 +788,7 @@ function createGroq(options = {}) {
   provider.imageModel = (modelId) => {
     throw new NoSuchModelError({ modelId, modelType: "imageModel" });
   };
+  provider.transcription = createTranscriptionModel;
   return provider;
 }
 var groq = createGroq();