@vanillaskyai/video 0.10.23 → 0.11.1

package/dist/scene-validation-SGLLLFY5.js

@@ -0,0 +1,9 @@
+import {
+  validateBuiltinScene
+} from "./chunk-HIQM4J3U.js";
+import "./chunk-ISWQ6T5X.js";
+import "./chunk-AUPC6MDK.js";
+import "./chunk-7NZOMTAL.js";
+export {
+  validateBuiltinScene
+};

package/dist/server.d.ts

@@ -1,9 +1,8 @@
-import { k as VideoGenerationContext, l as VideoPlanner, m as VideoRequest, a as VideoCapabilities, j as VideoInput, d as VideoAudio, n as VideoSceneValidator, g as VideoTemplatePacing, o as VideoSnapshotRetention, p as VideoResumeCursor, q as VideoSceneValidationContext, e as VideoScene, b as VideoOrientation } from './types-DdZw4GRQ.js';
-import { a as VideoFinishReason, V as VideoWarning, b as VideoEvent } from './events-B4YCc4vc.js';
-export { c as VideoWarningCategory } from './events-B4YCc4vc.js';
-import { S as SceneTemplateMetadata } from './catalog-types-WTbLP6Jh.js';
-import { V as VideoChatMode, f as VideoChatWelcomeOptions } from './types-BqB8zC9u.js';
-export { d as VideoChatCapabilities, g as VideoChatConversationTurn, h as VideoChatWelcomePrompt } from './types-BqB8zC9u.js';
+import { k as VideoPlanner, l as VideoRequest, c as VideoCapabilities, j as VideoInput, d as VideoAudio, m as VideoSceneValidator, g as VideoTemplatePacing, n as VideoSnapshotRetention, o as VideoResumeCursor, p as VideoGenerationContext, V as VideoOrientation, e as VideoScene } from './types-ht-Zw3Wv.js';
+import { a as VideoFinishReason, V as VideoWarning, b as VideoEvent } from './events-BZAfl0Dm.js';
+export { c as VideoWarningCategory } from './events-BZAfl0Dm.js';
+import { V as VideoChatConversationTurn, a as VideoChatMode, b as VideoChatWelcomeOptions } from './types-CG-kPI81.js';
+export { c as VideoChatCapabilities, d as VideoChatWelcomePrompt } from './types-CG-kPI81.js';
 
 interface VideoProviderUsage {
     inputTokens?: number;
@@ -31,39 +30,6 @@ interface VideoGenerationSummary {
     warnings: VideoWarning[];
 }
 
-interface TextDeltaVideoPlannerOptions {
-    /** Capture provider credentials in this server-only closure. */
-    streamText: (context: VideoGenerationContext) => AsyncIterable<string> | TextDeltaVideoSource;
-    /** Retain bounded provider-native usage and metadata on the server summary. */
-    includeRawProviderData?: boolean;
-}
-type Awaitable<T> = T | PromiseLike<T>;
-type ProviderFinishReason = string | {
-    unified?: string;
-    raw?: string;
-};
-interface TextDeltaVideoSource {
-    textStream: AsyncIterable<string>;
-    /** The provider SDK's normalized finish reason, when available. */
-    finishReason?: Awaitable<ProviderFinishReason | undefined>;
-    /** The provider's original finish reason, when available. */
-    rawFinishReason?: Awaitable<string | undefined>;
-    /** Vercel AI SDK usage. Both current and prior structural shapes are accepted. */
-    usage?: Awaitable<unknown>;
-    totalUsage?: Awaitable<unknown>;
-    providerMetadata?: Awaitable<unknown>;
-    warnings?: Awaitable<unknown>;
-    response?: Awaitable<unknown>;
-    finalStep?: Awaitable<unknown>;
-    /** Compatibility with older structural adapters. Prefer finalStep. */
-    lastStep?: Awaitable<unknown>;
-    steps?: Awaitable<unknown>;
-    /** Optional provider-neutral model hints for native adapters. */
-    requestedModelId?: Awaitable<string | undefined>;
-    resolvedModelId?: Awaitable<string | undefined>;
-    modelId?: Awaitable<string | undefined>;
-}
-
 interface VideoStreamHandlerOptions {
     generate: VideoPlanner;
     systemPrompt?: string | ((context: {
@@ -105,86 +71,48 @@ interface VideoStreamHandlerOptions {
     }) => AsyncIterable<VideoEvent>;
 }
 
-interface ServerTemplateRegistry {
-    readonly capabilities: VideoCapabilities;
-    getTemplateMetadata(id: string): SceneTemplateMetadata | undefined;
-    listTemplateMetadata(): SceneTemplateMetadata[];
+interface TextDeltaVideoPlannerOptions {
+    /** Capture provider credentials in this server-only closure. */
+    streamText: (context: VideoGenerationContext) => AsyncIterable<string> | TextDeltaVideoSource;
+    /** Retain bounded provider-native usage and metadata on the server summary. */
+    includeRawProviderData?: boolean;
+}
+type Awaitable<T> = T | PromiseLike<T>;
+type ProviderFinishReason = string | {
+    unified?: string;
+    raw?: string;
+};
+interface TextDeltaVideoSource {
+    textStream: AsyncIterable<string>;
+    /** The provider SDK's normalized finish reason, when available. */
+    finishReason?: Awaitable<ProviderFinishReason | undefined>;
+    /** The provider's original finish reason, when available. */
+    rawFinishReason?: Awaitable<string | undefined>;
+    /** Vercel AI SDK usage. Both current and prior structural shapes are accepted. */
+    usage?: Awaitable<unknown>;
+    totalUsage?: Awaitable<unknown>;
+    providerMetadata?: Awaitable<unknown>;
+    warnings?: Awaitable<unknown>;
+    response?: Awaitable<unknown>;
+    finalStep?: Awaitable<unknown>;
+    /** Compatibility with older structural adapters. Prefer finalStep. */
+    lastStep?: Awaitable<unknown>;
+    steps?: Awaitable<unknown>;
+    /** Optional provider-neutral model hints for native adapters. */
+    requestedModelId?: Awaitable<string | undefined>;
+    resolvedModelId?: Awaitable<string | undefined>;
+    modelId?: Awaitable<string | undefined>;
 }
-declare function createServerTemplateRegistry(options: {
-    templates: readonly SceneTemplateMetadata[];
-}): ServerTemplateRegistry;
-
-declare function createTemplateSceneValidator(options: {
-    kit: ServerTemplateRegistry;
-    /** Authorize an app-approved URL in addition to URLs supplied with the request. */
-    allowMediaUrl?: (url: string, context: VideoSceneValidationContext & {
-        scene: VideoScene;
-        variable: string;
-    }) => boolean;
-}): VideoSceneValidator;
 
 interface ResolvedMedia {
     url: string;
     type: "image" | "video";
     posterUrl?: string;
+    /** Actual footage duration when the provider or delivery pipeline knows it. */
+    durationSec?: number;
 }
-interface MediaResolverContext {
-    input: VideoInput;
-    requestId: string;
-    scene: Readonly<VideoScene>;
-    templateId: string;
-    preferredType: "image" | "video" | "any";
-    /**
-     * The visual language generated media must match, when the application set
-     * one. Append it to a provider prompt; a shot that ignores it is the
-     * mismatch this exists to prevent.
-     */
-    generatedLook?: string;
-    signal: AbortSignal;
-}
-type MediaResolver = (query: string, context: MediaResolverContext) => ResolvedMedia | null | Promise<ResolvedMedia | null>;
 
-interface VideoHandlerOptions extends Omit<VideoStreamHandlerOptions, "generate" | "systemPrompt" | "supportedCapabilities" | "validateScene" | "getTemplatePacing"> {
-    /** Customer-owned templates that replace matching built-ins and add new IDs. */
-    templates?: ServerTemplateRegistry;
-    /** App-owned provider adapter. Choose any model per request and keep provider clients and credentials in this closure. */
-    streamText: TextDeltaVideoPlannerOptions["streamText"];
-    /** Opt in to bounded provider-native usage and metadata in onComplete. */
-    includeRawProviderData?: boolean;
-    /** Additional video-direction rules prepended to the generated trusted-template catalog. */
-    basePrompt?: string;
-    /** Authorize an app-approved media URL in addition to URLs supplied in the request. */
-    allowMediaUrl?: Parameters<typeof createTemplateSceneValidator>[0]["allowMediaUrl"];
-    /** Resolve bounded semantic media intent through an application-owned provider. */
-    resolveMedia?: MediaResolver;
-    /**
-     * How many scenes may resolve media at once. Defaults to one.
-     *
-     * Raise it when resolution is slow enough to be felt - generated video rather
-     * than a stock search. Scenes are still emitted in the order they were
-     * planned; only the waiting overlaps.
-     */
-    mediaConcurrency?: number;
-    /**
-     * How many scenes in one request may resolve media at all.
-     *
-     * Unbounded by default, which is right when media is searched for and wrong
-     * when it is generated: the planner decides the scene count, and every scene
-     * is then a paid clip. Past the ceiling a scene uses its grounded text fallback, and a `media_budget_reached` warning says it happened.
-     */
-    maxResolvedMedia?: number;
-    /**
-     * Ask the planner to write each scene's spoken line, on the scene itself.
-     *
-     * A narrated video otherwise costs a round trip per scene - handing a model
-     * the scene that was just planned and asking what to say over it - and those
-     * calls have to be chained, because a line is written knowing the ones
-     * before it. The planner already knows the scene and the ones around it.
-     */
-    narrate?: boolean;
-}
-
-type VideoChatTextTask = "narration" | "suggestions";
+type VideoChatTextTask = "narration" | "narration-rewrite" | "suggestions";
 interface VideoChatTextContext {
     task: VideoChatTextTask;
     systemPrompt: string;
@@ -221,7 +149,27 @@ interface VideoChatMediaContext {
     fallbackQuery?: string;
 }
 type VideoChatMediaResolver = (query: string, context: VideoChatMediaContext) => ResolvedMedia | null | Promise<ResolvedMedia | null>;
-interface VideoChatHandlerOptions extends Pick<VideoHandlerOptions, "templates" | "streamText" | "includeRawProviderData" | "mediaConcurrency" | "allowedOrigins" | "authorize" | "maxBodyBytes" | "heartbeatMs" | "onError" | "onWarning" | "onComplete" | "invalidPartBehavior" | "requireCloser" | "allowCredentials"> {
+interface VideoChatVideoContext extends VideoChatMediaContext {
+    purpose: "response";
+    requestedDurationSec: number;
+    shotDirection: string;
+    /** Absolute epoch-millisecond deadline; never automatically resubmit a paid job. */
+    deadlineAt: number;
+}
+type VideoChatVideoGenerator = (query: string, context: VideoChatVideoContext) => ResolvedMedia | null | Promise<ResolvedMedia | null>;
+interface VideoChatHandlerOptions extends Pick<VideoStreamHandlerOptions, "allowedOrigins" | "authorize" | "maxBodyBytes" | "heartbeatMs" | "onError" | "onWarning" | "onComplete" | "invalidPartBehavior" | "requireCloser" | "allowCredentials"> {
+    /** Use an existing assistant's completed answer as the sole factual source for the video. */
+    resolveAnswer?: (context: {
+        prompt: string;
+        conversation: readonly VideoChatConversationTurn[];
+        signal: AbortSignal;
+    }) => string | Promise<string>;
+    /** Application-owned text stream; accepts a native async iterable or an AI SDK-shaped result. */
+    streamText: TextDeltaVideoPlannerOptions["streamText"];
+    /** Opt in to bounded provider metadata in the server-only completion callback. */
+    includeRawProviderData?: boolean;
+    /** Concurrent media jobs, bounded to 1–5. Results play in narrative order. */
+    mediaConcurrency?: number;
     /** Generate the small non-streaming text tasks around the visual plan. */
     generateText: VideoChatTextGenerator;
     /** Optional generated speech. Browsers can speak locally when absent. */
@@ -231,10 +179,10 @@ interface VideoChatHandlerOptions extends Pick<VideoHandlerOptions, "templates"
     /** Optional stock or application-owned media search. */
     searchMedia?: VideoChatMediaResolver;
     /** Optional generated-video provider. Its presence enables paid visual modes. */
-    generateVideo?: VideoChatMediaResolver;
+    generateVideo?: VideoChatVideoGenerator;
     /** Maximum generated-video attempts per response, including failures. Defaults to 5. */
     maxGeneratedVideos?: number;
-    /** Generated media deadline in milliseconds, 1–120000. Defaults to 15000. Host providers must honor cancellation. */
+    /** Generated media deadline in milliseconds, 1–600000. Defaults to 15000. Host providers must honor cancellation. */
     generateVideoTimeoutMs?: number;
     /** Provider-supported clip duration in seconds, 2–20. Defaults to 5; does not change provider billing configuration. */
     generatedClipDurationSec?: number;
@@ -242,11 +190,13 @@ interface VideoChatHandlerOptions extends Pick<VideoHandlerOptions, "templates"
     onDiagnostic?: (event: {
         requestId: string;
         mode: VideoChatMode;
-        phase: "request-accepted" | "opening-authored" | "shot-authored" | "media-start" | "media-end" | "media-skipped";
+        phase: "request-accepted" | "opening-authored" | "shot-authored" | "media-start" | "media-end" | "media-skipped" | "narration-fit" | "narration-rewrite";
         elapsedMs: number;
         sceneId?: string;
         durationMs?: number;
-        reason?: "ready" | "empty" | "provider-error" | "timeout" | "cancelled" | "allowance" | "deadline" | "not-configured";
+        estimatedSpeechSec?: number;
+        clipDurationSec?: number;
+        reason?: "ready" | "empty" | "provider-error" | "timeout" | "cancelled" | "allowance" | "deadline" | "not-configured" | "fit" | "rewritten" | "oversized";
     }) => unknown;
     /** Trusted application guidance appended to the general-purpose response brief. */
     instructions?: string;
@@ -266,4 +216,4 @@ type VideoChatHandler = (request: Request) => Promise<Response>;
  */
 declare function createVideoChatHandler(options: VideoChatHandlerOptions): VideoChatHandler;
 
-export { SceneTemplateMetadata as ServerTemplateMetadata, type ServerTemplateRegistry, type VideoChatHandlerOptions, VideoChatMode, VideoChatWelcomeOptions, VideoFinishReason, type VideoGenerationSummary, type VideoProviderUsage, VideoWarning, createServerTemplateRegistry, createVideoChatHandler };
+export { VideoChatConversationTurn, type VideoChatHandlerOptions, VideoChatMode, VideoChatWelcomeOptions, VideoFinishReason, type VideoGenerationSummary, type VideoProviderUsage, VideoWarning, createVideoChatHandler };