@ai-sdk/gateway 4.0.43 → 4.0.44

package/CHANGELOG.md

@@ -1,5 +1,20 @@
 # @ai-sdk/gateway
 
+## 4.0.44
+
+### Patch Changes
+
+- 89080c8: feat (provider/gateway): support the async video operation flow (doStart/doStatus) on the v4 video model
+- 89080c8: fix (ai/gateway): make retried `doStart` calls idempotent
+
+  `generateVideo` retries `doStart`, which creates a billable generation, so a
+  retry after a lost response could start a second one. It now mints one
+  idempotency token per logical start — outside the retry closure — and forwards it
+  as an `idempotency-key` header, so a provider that deduplicates (the Vercel AI
+  Gateway does) sees the same key on every attempt. `GatewayVideoModel` simply
+  forwards the caller's headers rather than inferring retry identity from an
+  options object, which would collide across unrelated calls.
+
 ## 4.0.43
 
 ### Patch Changes

package/dist/index.js

@@ -63,7 +63,7 @@ function decodeSubprotocolValue(value) {
 // src/gateway-provider.ts
 import {
   createJsonErrorResponseHandler as createJsonErrorResponseHandler11,
-  createJsonResponseHandler as createJsonResponseHandler10,
+  createJsonResponseHandler as createJsonResponseHandler11,
   loadOptionalSetting,
   postJsonToApi as postJsonToApi8,
   withoutTrailingSlash,
@@ -1413,6 +1413,7 @@ import {
   combineHeaders as combineHeaders4,
   convertUint8ArrayToBase64 as convertUint8ArrayToBase642,
   createJsonErrorResponseHandler as createJsonErrorResponseHandler7,
+  createJsonResponseHandler as createJsonResponseHandler7,
   parseJsonEventStream,
   postJsonToApi as postJsonToApi4,
   resolve as resolve7
@@ -1429,23 +1430,9 @@ var GatewayVideoModel = class {
   get provider() {
     return this.config.provider;
   }
-  async doGenerate({
-    prompt,
-    n,
-    aspectRatio,
-    resolution,
-    duration,
-    fps,
-    seed,
-    generateAudio,
-    image,
-    frameImages,
-    inputReferences,
-    providerOptions,
-    headers,
-    abortSignal
-  }) {
-    var _a11;
+  async doGenerate(options) {
+    var _a11, _b11;
+    const { headers, abortSignal } = options;
     const resolvedHeaders = this.config.headers ? await resolve7(this.config.headers) : void 0;
     try {
       const { responseHeaders, value: responseBody } = await postJsonToApi4({
@@ -1457,29 +1444,7 @@ var GatewayVideoModel = class {
           await resolve7(this.config.o11yHeaders),
           { accept: "text/event-stream" }
         ),
-        body: {
-          prompt,
-          n,
-          ...aspectRatio && { aspectRatio },
-          ...resolution && { resolution },
-          ...duration && { duration },
-          ...fps && { fps },
-          ...seed && { seed },
-          ...generateAudio !== void 0 && { generateAudio },
-          ...providerOptions && { providerOptions },
-          ...image && { image: maybeEncodeVideoFile(image) },
-          ...frameImages && {
-            frameImages: frameImages.map((frame) => ({
-              ...frame,
-              image: maybeEncodeVideoFile(frame.image)
-            }))
-          },
-          ...inputReferences && {
-            inputReferences: inputReferences.map(
-              (reference) => maybeEncodeVideoFile(reference)
-            )
-          }
-        },
+        body: this.buildRequestBody(options),
         successfulResponseHandler: async ({
           response,
           url,
@@ -1554,7 +1519,7 @@ var GatewayVideoModel = class {
       return {
         videos: responseBody.videos,
         warnings: (_a11 = responseBody.warnings) != null ? _a11 : [],
-        providerMetadata: responseBody.providerMetadata,
+        providerMetadata: (_b11 = responseBody.providerMetadata) != null ? _b11 : void 0,
         response: {
           timestamp: /* @__PURE__ */ new Date(),
           modelId: this.modelId,
@@ -1568,9 +1533,178 @@ var GatewayVideoModel = class {
       );
     }
   }
+  // `handleWebhookOption` is intentionally NOT implemented. The Gateway's
+  // async video jobs are polling-first: it does not expose a provider->SDK
+  // webhook completion channel. If this method were present, `generateVideo`
+  // would forward a webhook URL and await a notification that never arrives.
+  // `doStart` still forwards the spec's `webhookUrl` — as the Gateway's
+  // `callbackUrl` wire field (its completion-webhook contract; unknown body
+  // keys are stripped server-side). Enabling webhooks later requires adding
+  // `handleWebhookOption` here plus handling the Gateway's HMAC-signed
+  // delivery format; the wire field is already the right one.
+  async doStart(options) {
+    var _a11, _b11;
+    const { headers, abortSignal, webhookUrl } = options;
+    const resolvedHeaders = this.config.headers ? await resolve7(this.config.headers) : void 0;
+    try {
+      const { responseHeaders, value: responseBody } = await postJsonToApi4({
+        url: this.getStartUrl(),
+        headers: combineHeaders4(
+          resolvedHeaders,
+          headers != null ? headers : {},
+          this.getModelConfigHeaders(),
+          await resolve7(this.config.o11yHeaders)
+        ),
+        body: {
+          ...this.buildRequestBody(options),
+          // The spec option is `webhookUrl`; the Gateway's wire contract for a
+          // completion webhook is `callbackUrl`.
+          ...webhookUrl && { callbackUrl: webhookUrl }
+        },
+        successfulResponseHandler: createJsonResponseHandler7(
+          gatewayVideoStartResponseSchema
+        ),
+        failedResponseHandler: createJsonErrorResponseHandler7({
+          errorSchema: z11.any(),
+          errorToMessage: (data) => data
+        }),
+        ...abortSignal && { abortSignal },
+        fetch: this.config.fetch
+      });
+      return {
+        operation: responseBody.operation,
+        warnings: (_a11 = responseBody.warnings) != null ? _a11 : [],
+        providerMetadata: (_b11 = responseBody.providerMetadata) != null ? _b11 : void 0,
+        response: {
+          timestamp: /* @__PURE__ */ new Date(),
+          modelId: this.modelId,
+          headers: responseHeaders
+        }
+      };
+    } catch (error) {
+      throw await asGatewayError(
+        error,
+        await parseAuthMethod(resolvedHeaders != null ? resolvedHeaders : {})
+      );
+    }
+  }
+  async doStatus({
+    operation,
+    abortSignal,
+    headers
+  }) {
+    var _a11, _b11, _c, _d, _e, _f;
+    const resolvedHeaders = this.config.headers ? await resolve7(this.config.headers) : void 0;
+    try {
+      const { responseHeaders, value: responseBody } = await postJsonToApi4({
+        url: this.getStatusUrl(),
+        headers: combineHeaders4(
+          resolvedHeaders,
+          headers != null ? headers : {},
+          this.getModelConfigHeaders(),
+          await resolve7(this.config.o11yHeaders)
+        ),
+        body: { operation },
+        successfulResponseHandler: createJsonResponseHandler7(
+          gatewayVideoStatusResponseSchema
+        ),
+        failedResponseHandler: createJsonErrorResponseHandler7({
+          errorSchema: z11.any(),
+          errorToMessage: (data) => data
+        }),
+        ...abortSignal && { abortSignal },
+        fetch: this.config.fetch
+      });
+      const response = {
+        timestamp: /* @__PURE__ */ new Date(),
+        modelId: this.modelId,
+        headers: responseHeaders
+      };
+      if (responseBody.status === "completed") {
+        return {
+          status: "completed",
+          videos: responseBody.videos,
+          warnings: (_a11 = responseBody.warnings) != null ? _a11 : [],
+          providerMetadata: (_b11 = responseBody.providerMetadata) != null ? _b11 : void 0,
+          response
+        };
+      }
+      if (responseBody.status === "error") {
+        return {
+          status: "error",
+          error: responseBody.error,
+          providerMetadata: (_c = responseBody.providerMetadata) != null ? _c : void 0,
+          response
+        };
+      }
+      if (responseBody.status === "cancelled") {
+        return {
+          status: "error",
+          error: "Video generation was cancelled.",
+          providerMetadata: (_d = responseBody.providerMetadata) != null ? _d : void 0,
+          response
+        };
+      }
+      return {
+        status: "pending",
+        warnings: (_e = responseBody.warnings) != null ? _e : [],
+        providerMetadata: (_f = responseBody.providerMetadata) != null ? _f : void 0,
+        response
+      };
+    } catch (error) {
+      throw await asGatewayError(
+        error,
+        await parseAuthMethod(resolvedHeaders != null ? resolvedHeaders : {})
+      );
+    }
+  }
+  buildRequestBody({
+    prompt,
+    n,
+    aspectRatio,
+    resolution,
+    duration,
+    fps,
+    seed,
+    generateAudio,
+    image,
+    frameImages,
+    inputReferences,
+    providerOptions
+  }) {
+    return {
+      prompt,
+      n,
+      ...aspectRatio && { aspectRatio },
+      ...resolution && { resolution },
+      ...duration && { duration },
+      ...fps && { fps },
+      ...seed && { seed },
+      ...generateAudio !== void 0 && { generateAudio },
+      ...providerOptions && { providerOptions },
+      ...image && { image: maybeEncodeVideoFile(image) },
+      ...frameImages && {
+        frameImages: frameImages.map((frame) => ({
+          ...frame,
+          image: maybeEncodeVideoFile(frame.image)
+        }))
+      },
+      ...inputReferences && {
+        inputReferences: inputReferences.map(
+          (reference) => maybeEncodeVideoFile(reference)
+        )
+      }
+    };
+  }
   getUrl() {
     return `${this.config.baseURL}/video-model`;
   }
+  getStartUrl() {
+    return `${this.config.baseURL}/video-model/start`;
+  }
+  getStatusUrl() {
+    return `${this.config.baseURL}/video-model/status`;
+  }
   getModelConfigHeaders() {
     return {
       "ai-video-model-specification-version": "4",
@@ -1638,12 +1772,39 @@ var gatewayVideoEventSchema = z11.discriminatedUnion("type", [
     param: z11.unknown().nullable()
   })
 ]);
+var gatewayVideoStartResponseSchema = z11.object({
+  operation: z11.unknown(),
+  warnings: z11.array(gatewayVideoWarningSchema).nullish(),
+  providerMetadata: z11.record(z11.string(), providerMetadataEntrySchema2).nullish()
+});
+var gatewayVideoStatusResponseSchema = z11.discriminatedUnion("status", [
+  z11.object({
+    status: z11.literal("pending"),
+    warnings: z11.array(gatewayVideoWarningSchema).nullish(),
+    providerMetadata: z11.record(z11.string(), providerMetadataEntrySchema2).nullish()
+  }),
+  z11.object({
+    status: z11.literal("completed"),
+    videos: z11.array(gatewayVideoDataSchema),
+    warnings: z11.array(gatewayVideoWarningSchema).nullish(),
+    providerMetadata: z11.record(z11.string(), providerMetadataEntrySchema2).nullish()
+  }),
+  z11.object({
+    status: z11.literal("error"),
+    error: z11.string(),
+    providerMetadata: z11.record(z11.string(), providerMetadataEntrySchema2).nullish()
+  }),
+  z11.object({
+    status: z11.literal("cancelled"),
+    providerMetadata: z11.record(z11.string(), providerMetadataEntrySchema2).nullish()
+  })
+]);
 
 // src/gateway-reranking-model.ts
 import {
   combineHeaders as combineHeaders5,
   createJsonErrorResponseHandler as createJsonErrorResponseHandler8,
-  createJsonResponseHandler as createJsonResponseHandler7,
+  createJsonResponseHandler as createJsonResponseHandler8,
   lazySchema as lazySchema9,
   postJsonToApi as postJsonToApi5,
   resolve as resolve8,
@@ -1688,7 +1849,7 @@ var GatewayRerankingModel = class {
           ...topN != null ? { topN } : {},
           ...providerOptions ? { providerOptions } : {}
         },
-        successfulResponseHandler: createJsonResponseHandler7(
+        successfulResponseHandler: createJsonResponseHandler8(
           gatewayRerankingResponseSchema
         ),
         failedResponseHandler: createJsonErrorResponseHandler8({
@@ -1761,7 +1922,7 @@ var gatewayRerankingResponseSchema = lazySchema9(
 import {
   combineHeaders as combineHeaders6,
   createJsonErrorResponseHandler as createJsonErrorResponseHandler9,
-  createJsonResponseHandler as createJsonResponseHandler8,
+  createJsonResponseHandler as createJsonResponseHandler9,
   postJsonToApi as postJsonToApi6,
   resolve as resolve9
 } from "@ai-sdk/provider-utils";
@@ -1810,7 +1971,7 @@ var GatewaySpeechModel = class {
           ...language && { language },
           ...providerOptions && { providerOptions }
         },
-        successfulResponseHandler: createJsonResponseHandler8(
+        successfulResponseHandler: createJsonResponseHandler9(
           gatewaySpeechResponseSchema
         ),
         failedResponseHandler: createJsonErrorResponseHandler9({
@@ -1885,7 +2046,7 @@ import {
   convertBase64ToUint8Array,
   convertUint8ArrayToBase64 as convertUint8ArrayToBase643,
   createJsonErrorResponseHandler as createJsonErrorResponseHandler10,
-  createJsonResponseHandler as createJsonResponseHandler9,
+  createJsonResponseHandler as createJsonResponseHandler10,
   connectToWebSocket,
   normalizeHeaders,
   experimental_parseTranscriptionStreamPart as parseTranscriptionStreamPart,
@@ -1932,7 +2093,7 @@ var GatewayTranscriptionModel = class {
           mediaType,
           ...providerOptions && { providerOptions }
         },
-        successfulResponseHandler: createJsonResponseHandler9(
+        successfulResponseHandler: createJsonResponseHandler10(
           gatewayTranscriptionResponseSchema
         ),
         failedResponseHandler: createJsonErrorResponseHandler10({
@@ -2631,7 +2792,7 @@ async function getVercelRequestId() {
 }
 
 // src/version.ts
-var VERSION = true ? "4.0.43" : "0.0.0-test";
+var VERSION = true ? "4.0.44" : "0.0.0-test";
 
 // src/gateway-provider.ts
 var AI_GATEWAY_PROTOCOL_VERSION = "0.0.1";
@@ -2696,7 +2857,7 @@ function createGateway(options = {}) {
             expiresIn: params.expiresAfterSeconds
           }
         },
-        successfulResponseHandler: createJsonResponseHandler10(
+        successfulResponseHandler: createJsonResponseHandler11(
           gatewayClientSecretResponseSchema
         ),
         failedResponseHandler: createJsonErrorResponseHandler11({