ai 7.0.55 → 7.0.57

package/dist/internal/index.d.ts

@@ -1399,6 +1399,8 @@ type LanguageModelCallEndEvent<TOOLS extends ToolSet = ToolSet> = ModelInfo & {
     readonly content: ReadonlyArray<ContentPart<TOOLS>>;
     /** The provider-returned response id for this model call. */
     readonly responseId: string;
+    /** Optional provider-specific metadata for this model call. */
+    readonly providerMetadata?: ProviderMetadata;
     /** Performance metrics for the model call. */
     readonly performance: {
         /** Time spent waiting for the language model response in milliseconds. */

package/dist/internal/index.js

@@ -92,7 +92,7 @@ import {
 } from "@ai-sdk/provider-utils";
 
 // src/version.ts
-var VERSION = true ? "7.0.55" : "0.0.0-test";
+var VERSION = true ? "7.0.57" : "0.0.0-test";
 
 // src/util/download/download.ts
 var download = async ({

package/docs/03-ai-sdk-core/05-generating-text.mdx

@@ -187,7 +187,7 @@ The available lifecycle callbacks are:
 - **`onStart`**: Called once when the `generateText` operation begins, before any LLM calls. Receives model info, messages, settings, and `runtimeContext`.
 - **`onStepStart`**: Called before each step (LLM call). Receives the step number, model, messages being sent, tools, and prior steps.
 - **`onLanguageModelCallStart`**: Called immediately before the provider model call begins. Useful when you want to observe the model invocation separately from later tool execution.
-- **`onLanguageModelCallEnd`**: Called after the model response has been normalized and parsed, but before any client-side tool execution begins. Receives the model-call content parts, usage, and finish reason.
+- **`onLanguageModelCallEnd`**: Called after the model response has been normalized and parsed, but before any client-side tool execution begins. Receives the model-call content parts, usage, finish reason, and provider metadata.
 - **`onToolExecutionStart`**: Called right before a tool's `execute` function runs. Receives the tool call object, messages, and `toolContext`.
 - **`onToolExecutionEnd`**: Called right after a tool's `execute` function completes or errors. Receives the tool call object, `toolExecutionMs`, and a `toolOutput` discriminated union (`type: 'tool-result'` with `output`, or `type: 'tool-error'` with `error`).
 - **`onStepEnd`**: Called after each step finishes. Includes `stepNumber` (zero-based index of the completed step).
@@ -417,7 +417,7 @@ The available lifecycle callbacks are:
 - **`onStart`**: Called once when the `streamText` operation begins, before any LLM calls. Receives model info, messages, settings, and `runtimeContext`.
 - **`onStepStart`**: Called before each step (LLM call). Receives the step number, model, messages being sent, tools, and prior steps.
 - **`onLanguageModelCallStart`**: Called immediately before the provider model call begins. Useful when you want to observe the model invocation separately from later tool execution.
-- **`onLanguageModelCallEnd`**: Called after the model response has been normalized and parsed, but before any client-side tool execution begins. Receives the model-call content parts, usage, and finish reason.
+- **`onLanguageModelCallEnd`**: Called after the model response has been normalized and parsed, but before any client-side tool execution begins. Receives the model-call content parts, usage, finish reason, and provider metadata.
 - **`onToolExecutionStart`**: Called right before a tool's `execute` function runs. Receives the tool call object, messages, and `toolContext`.
 - **`onToolExecutionEnd`**: Called right after a tool's `execute` function completes or errors. Receives the tool call object, `toolExecutionMs`, and a `toolOutput` discriminated union (`type: 'tool-result'` with `output`, or `type: 'tool-error'` with `error`).
 - **`onStepEnd`**: Called after each step finishes. Receives the finish reason, usage, and other step details.

package/docs/03-ai-sdk-core/60-telemetry.mdx

@@ -356,7 +356,7 @@ export function myIntegration(): Telemetry {
       name: 'onLanguageModelCallEnd',
       type: '(event: LanguageModelCallEndEvent) => void | PromiseLike<void>',
       description:
-        'Called after the model response has been normalized and parsed, but before any client-side tool execution begins.',
+        'Called after the model response has been normalized and parsed, but before any client-side tool execution begins. Includes provider-specific metadata when available.',
     },
     {
       name: 'onToolExecutionStart',
@@ -628,7 +628,7 @@ registerTelemetry(
 The available options are:
 
 - `usage`: detailed usage attributes that are not covered by GenAI usage attributes, such as uncached input tokens and output text/reasoning token details.
-- `providerMetadata`: `ai.response.providerMetadata`.
+- `providerMetadata`: `ai.response.providerMetadata` on operation, step, and model-call spans.
 - `embedding`: embedding inputs and outputs.
 - `reranking`: rerank input documents and ranking output.
 - `runtimeContext`: `ai.settings.context.*`.

package/docs/03-ai-sdk-core/65-lifecycle-callbacks.mdx

@@ -97,7 +97,13 @@ const result = streamText({
   model: __MODEL__,
   prompt: 'Explain partial prerendering in two paragraphs.',
 
-  onLanguageModelCallEnd({ callId, modelId, usage, performance }) {
+  onLanguageModelCallEnd({
+    callId,
+    modelId,
+    usage,
+    performance,
+    providerMetadata,
+  }) {
     metrics.histogram('ai.model.response_time_ms', performance.responseTimeMs, {
       callId,
       modelId,
@@ -108,6 +114,11 @@ const result = streamText({
       total: performance.effectiveTotalTokensPerSecond,
       tokens: usage.totalTokens,
     });
+
+    logger.info('ai.model.provider_metadata', {
+      callId,
+      providerMetadata,
+    });
   },
 });
 
@@ -675,6 +686,12 @@ Called after the provider response has been normalized and parsed, before local
       type: 'string',
       description: 'Provider-returned response ID for this model call.',
     },
+    {
+      name: 'providerMetadata',
+      type: 'ProviderMetadata | undefined',
+      description:
+        'Provider-specific metadata for this model call, when returned by the provider.',
+    },
     {
       name: 'performance',
       type: 'LanguageModelCallPerformance',

package/docs/07-reference/01-ai-sdk-core/01-generate-text.mdx

@@ -1438,6 +1438,12 @@ To see `generateText` in action, check out [these examples](#examples).
               description:
                 'The provider-returned response ID for this model call.',
             },
+            {
+              name: 'providerMetadata',
+              type: 'ProviderMetadata | undefined',
+              description:
+                'Provider-specific metadata for this model call, when returned by the provider.',
+            },
             {
               name: 'performance',
               type: '{ responseTimeMs: number; effectiveOutputTokensPerSecond: number; outputTokensPerSecond: number | undefined; inputTokensPerSecond: number | undefined; effectiveTotalTokensPerSecond: number; timeToFirstOutputMs: number | undefined; timeBetweenOutputChunksMs?: OutputChunkTimingStats }',

package/docs/07-reference/01-ai-sdk-core/02-stream-text.mdx

@@ -2422,6 +2422,12 @@ To see `streamText` in action, check out [these examples](#examples).
               description:
                 'The provider-returned response ID for this model call.',
             },
+            {
+              name: 'providerMetadata',
+              type: 'ProviderMetadata | undefined',
+              description:
+                'Provider-specific metadata for this model call, when returned by the provider.',
+            },
             {
               name: 'performance',
               type: '{ responseTimeMs: number; effectiveOutputTokensPerSecond: number; outputTokensPerSecond: number | undefined; inputTokensPerSecond: number | undefined; effectiveTotalTokensPerSecond: number; timeToFirstOutputMs: number | undefined; timeBetweenOutputChunksMs?: OutputChunkTimingStats }',

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai",
-  "version": "7.0.55",
+  "version": "7.0.57",
   "type": "module",
   "description": "AI SDK by Vercel - build apps like ChatGPT, Claude, Gemini, and more with a single interface for any model using the Vercel AI Gateway or go direct to OpenAI, Anthropic, Google, or any other model provider.",
   "license": "Apache-2.0",
@@ -42,9 +42,9 @@
     }
   },
   "dependencies": {
-    "@ai-sdk/gateway": "4.0.43",
+    "@ai-sdk/gateway": "4.0.45",
     "@ai-sdk/provider": "4.0.6",
-    "@ai-sdk/provider-utils": "5.0.23"
+    "@ai-sdk/provider-utils": "5.0.24"
   },
   "devDependencies": {
     "@edge-runtime/vm": "^5.0.0",

package/src/generate-text/generate-text.ts

@@ -1085,6 +1085,11 @@ export async function generateText<
                   usage: stepUsage,
                   content: modelCallContent,
                   responseId: currentModelResponse.response.id,
+                  ...(currentModelResponse.providerMetadata != null
+                    ? {
+                        providerMetadata: currentModelResponse.providerMetadata,
+                      }
+                    : {}),
                   performance: {
                     responseTimeMs,
                     effectiveOutputTokensPerSecond: calculateTokensPerSecond({

package/src/generate-text/language-model-events.ts

@@ -1,6 +1,7 @@
 import type { ToolSet } from '@ai-sdk/provider-utils';
 import type { Callback } from '../util/callback';
 import type { FinishReason } from '../types/language-model';
+import type { ProviderMetadata } from '../types/provider-metadata';
 import type { LanguageModelUsage } from '../types/usage';
 import type { ContentPart } from './content-part';
 import type { StandardizedPrompt } from '../prompt/standardize-prompt';
@@ -55,6 +56,9 @@ export type LanguageModelCallEndEvent<TOOLS extends ToolSet = ToolSet> =
     /** The provider-returned response id for this model call. */
     readonly responseId: string;
 
+    /** Optional provider-specific metadata for this model call. */
+    readonly providerMetadata?: ProviderMetadata;
+
     /** Performance metrics for the model call. */
     readonly performance: {
       /** Time spent waiting for the language model response in milliseconds. */

package/src/generate-text/stream-language-model-call.ts

@@ -596,6 +596,9 @@ function createLanguageModelV4StreamPartToLanguageModelStreamPartTransform<
               usage,
               content: modelCallContent,
               responseId,
+              ...(chunk.providerMetadata != null
+                ? { providerMetadata: chunk.providerMetadata }
+                : {}),
               performance,
             },
             callbacks: onLanguageModelCallEnd,

package/src/generate-video/generate-video.ts

@@ -11,6 +11,7 @@ import type {
 import {
   convertBase64ToUint8Array,
   delay as defaultDelay,
+  generateId,
   withUserAgentSuffix,
   type DataContent,
   detectMediaType,
@@ -553,13 +554,23 @@ async function executeStartStatusFlow({
     }
   }
 
-  // 2. Start the generation
-  const startResult = await retry(() =>
-    model.doStart!({
-      ...callOptions,
-      webhookUrl,
-    }),
+  // 2. Start the generation. `doStart` is billable: mint one idempotency token
+  // per logical start, outside the retry closure; a caller-supplied key wins.
+  const callerIdempotencyKey = Object.entries(callOptions.headers ?? {}).find(
+    ([key, value]) =>
+      key.toLowerCase() === 'idempotency-key' && value !== undefined,
   );
+  const startCallOptions = {
+    ...callOptions,
+    headers: {
+      ...callOptions.headers,
+      ...(callerIdempotencyKey
+        ? {}
+        : { 'idempotency-key': `aisdk_vid_${generateId()}` }),
+    },
+    webhookUrl,
+  };
+  const startResult = await retry(() => model.doStart!(startCallOptions));
 
   const allWarnings = [...earlyWarnings, ...startResult.warnings];
   let operationProviderMetadata =

package/src/ui/chat-transport.ts

@@ -64,6 +64,7 @@ export interface ChatTransport<UI_MESSAGE extends UIMessage> {
    *
    * @param options - Configuration object containing:
    * @param options.chatId - Unique identifier for the chat session to reconnect to
+   * @param options.abortSignal - Signal to abort the reconnection request if needed
    * @param options.headers - Additional HTTP headers to include in the reconnection request
    * @param options.body - Additional JSON properties to include in the request body
    * @param options.metadata - Custom metadata to attach to the request
@@ -78,6 +79,8 @@ export interface ChatTransport<UI_MESSAGE extends UIMessage> {
     options: {
       /** Unique identifier for the chat session to reconnect to */
       chatId: string;
+      /** Signal to abort the reconnection request if needed */
+      abortSignal?: AbortSignal;
     } & ChatRequestOptions,
   ) => Promise<ReadableStream<UIMessageChunk> | null>;
 }

package/src/ui/chat.ts

@@ -135,6 +135,10 @@ type ActiveResponse<UI_MESSAGE extends UIMessage> = {
   abortController: AbortController;
 };
 
+type ActiveResumeRequest = {
+  abortController: AbortController;
+};
+
 export interface ChatState<UI_MESSAGE extends UIMessage> {
   status: ChatStatus;
 
@@ -254,6 +258,7 @@ export abstract class AbstractChat<UI_MESSAGE extends UIMessage> {
   private sendAutomaticallyWhen?: ChatInit<UI_MESSAGE>['sendAutomaticallyWhen'];
 
   private activeResponse: ActiveResponse<UI_MESSAGE> | undefined = undefined;
+  private activeResumeRequest: ActiveResumeRequest | undefined = undefined;
   private jobExecutor = new SerialJobExecutor();
 
   constructor({
@@ -584,11 +589,8 @@ export abstract class AbstractChat<UI_MESSAGE extends UIMessage> {
    * Abort the current request immediately, keep the generated tokens if any.
    */
   stop = async () => {
-    if (this.status !== 'streaming' && this.status !== 'submitted') return;
-
-    if (this.activeResponse?.abortController) {
-      this.activeResponse.abortController.abort();
-    }
+    this.activeResumeRequest?.abortController.abort();
+    this.activeResponse?.abortController.abort();
   };
 
   private async shouldSendAutomatically(): Promise<boolean> {
@@ -616,6 +618,25 @@ export abstract class AbstractChat<UI_MESSAGE extends UIMessage> {
     trigger: 'submit-message' | 'resume-stream' | 'regenerate-message';
     messageId?: string;
   } & ChatRequestOptions) {
+    const abortController = new AbortController();
+    const activeResumeRequest =
+      trigger === 'resume-stream' ? { abortController } : undefined;
+
+    if (activeResumeRequest) {
+      this.activeResumeRequest?.abortController.abort();
+      this.activeResumeRequest = activeResumeRequest;
+    }
+
+    const isCurrentRequest = () =>
+      activeResumeRequest == null ||
+      this.activeResumeRequest === activeResumeRequest;
+
+    const clearActiveResumeRequest = () => {
+      if (this.activeResumeRequest === activeResumeRequest) {
+        this.activeResumeRequest = undefined;
+      }
+    };
+
     // For resume-stream, check if there's an active stream before
     // changing status. This avoids a brief flash of 'submitted' status
     // when there is no stream to resume (e.g. on page load).
@@ -624,21 +645,49 @@ export abstract class AbstractChat<UI_MESSAGE extends UIMessage> {
       try {
         const reconnect = await this.transport.reconnectToStream({
           chatId: this.id,
+          abortSignal: abortController.signal,
           metadata,
           headers,
           body,
         });
 
+        if (abortController.signal.aborted || !isCurrentRequest()) {
+          await reconnect?.cancel().catch(() => {});
+          if (isCurrentRequest()) {
+            this.setStatus({ status: 'ready' });
+          }
+          clearActiveResumeRequest();
+          return;
+        }
+
         if (reconnect == null) {
+          this.setStatus({ status: 'ready' });
+          clearActiveResumeRequest();
           return; // no active stream found, so we do not resume
         }
 
         resumeStream = reconnect;
       } catch (err) {
+        if (
+          abortController.signal.aborted ||
+          (err as { name?: string }).name === 'AbortError'
+        ) {
+          if (isCurrentRequest()) {
+            this.setStatus({ status: 'ready' });
+          }
+          clearActiveResumeRequest();
+          return;
+        }
+
+        if (!isCurrentRequest()) {
+          return;
+        }
+
         if (this.onError && err instanceof Error) {
           this.onError(err);
         }
         this.setStatus({ status: 'error', error: err as Error });
+        clearActiveResumeRequest();
         return;
       }
     }
@@ -661,7 +710,7 @@ export abstract class AbstractChat<UI_MESSAGE extends UIMessage> {
               : this.state.snapshot(lastMessage),
           messageId: this.generateId(),
         }),
-        abortController: new AbortController(),
+        abortController,
       } as ActiveResponse<UI_MESSAGE>;
 
       activeResponse = response;
@@ -696,10 +745,18 @@ export abstract class AbstractChat<UI_MESSAGE extends UIMessage> {
         }) => Promise<void>,
       ) =>
         // serialize the job execution to avoid race conditions:
-        this.jobExecutor.run(() =>
-          job({
+        this.jobExecutor.run(() => {
+          if (response.abortController.signal.aborted) {
+            return Promise.resolve();
+          }
+
+          return job({
             state: response.state,
             write: () => {
+              if (response.abortController.signal.aborted) {
+                return;
+              }
+
               // streaming is set on first write (before it should be "submitted")
               this.setStatus({ status: 'streaming' });
 
@@ -715,8 +772,8 @@ export abstract class AbstractChat<UI_MESSAGE extends UIMessage> {
                 this.state.pushMessage(response.state.message);
               }
             },
-          }),
-        );
+          });
+        });
 
       await consumeStream({
         stream: processUIMessageStream({
@@ -730,17 +787,33 @@ export abstract class AbstractChat<UI_MESSAGE extends UIMessage> {
             throw error;
           },
         }),
+        abortSignal: response.abortController.signal,
         onError: error => {
           throw error;
         },
       });
 
-      this.setStatus({ status: 'ready' });
+      if (isAbort) {
+        if (isCurrentRequest()) {
+          this.setStatus({ status: 'ready' });
+        }
+        return null;
+      }
+
+      if (isCurrentRequest()) {
+        this.setStatus({ status: 'ready' });
+      }
     } catch (err) {
       // Ignore abort errors as they are expected.
       if (isAbort || (err as any).name === 'AbortError') {
         isAbort = true;
-        this.setStatus({ status: 'ready' });
+        if (isCurrentRequest()) {
+          this.setStatus({ status: 'ready' });
+        }
+        return null;
+      }
+
+      if (!isCurrentRequest()) {
         return null;
       }
 
@@ -779,6 +852,8 @@ export abstract class AbstractChat<UI_MESSAGE extends UIMessage> {
       if (this.activeResponse === activeResponse) {
         this.activeResponse = undefined;
       }
+
+      clearActiveResumeRequest();
     }
 
     // automatically send the message if the sendAutomaticallyWhen function returns true

package/src/ui/http-chat-transport.ts

@@ -247,6 +247,7 @@ export abstract class HttpChatTransport<
       method: 'GET',
       headers,
       credentials,
+      signal: options.abortSignal,
     });
 
     // no active stream found, so we do not resume

package/src/util/consume-stream.ts

@@ -13,11 +13,23 @@
 export async function consumeStream({
   stream,
   onError,
+  abortSignal,
 }: {
   stream: ReadableStream;
   onError?: (error: unknown) => void;
+  abortSignal?: AbortSignal;
 }): Promise<void> {
   const reader = stream.getReader();
+  const cancelOnAbort = () => {
+    reader.cancel().catch(() => {});
+  };
+
+  if (abortSignal?.aborted) {
+    cancelOnAbort();
+  } else {
+    abortSignal?.addEventListener('abort', cancelOnAbort, { once: true });
+  }
+
   try {
     while (true) {
       const { done } = await reader.read();
@@ -26,6 +38,7 @@ export async function consumeStream({
   } catch (error) {
     onError?.(error);
   } finally {
+    abortSignal?.removeEventListener('abort', cancelOnAbort);
     reader.releaseLock();
   }
 }