@agentfield/sdk 0.1.89 → 0.1.90-rc.1

package/dist/index.d.ts

@@ -1344,6 +1344,32 @@ declare class Audio {
      */
     static fromBase64(base64Data: string, format?: 'wav' | 'mp3' | 'flac' | 'ogg'): Promise<Audio>;
 }
+/**
+ * Represents video content in a multimodal prompt.
+ */
+declare class Video {
+    readonly type: 'video_url';
+    readonly videoUrl: {
+        url: string;
+    };
+    private constructor();
+    /**
+     * Create Video from a local file by converting to a base64 data URL.
+     */
+    static fromFile(filePath: string): Promise<Video>;
+    /**
+     * Create Video from a URL.
+     */
+    static fromUrl(url: string): Video;
+    /**
+     * Create Video from a buffer.
+     */
+    static fromBuffer(buffer: Buffer | Uint8Array, mimeType?: string): Promise<Video>;
+    /**
+     * Create Video from a base64 string.
+     */
+    static fromBase64(base64Data: string, mimeType?: string): Promise<Video>;
+}
 /**
  * Represents a generic file content in a multimodal prompt.
  */
@@ -1407,6 +1433,22 @@ declare function audioFromBuffer(buffer: Buffer | Uint8Array, format?: 'wav' | '
  * Create audio content from a base64 string.
  */
 declare function audioFromBase64(base64Data: string, format?: 'wav' | 'mp3' | 'flac' | 'ogg'): Promise<Audio>;
+/**
+ * Create video content from a local file.
+ */
+declare function videoFromFile(filePath: string): Promise<Video>;
+/**
+ * Create video content from a URL.
+ */
+declare function videoFromUrl(url: string): Video;
+/**
+ * Create video content from a buffer.
+ */
+declare function videoFromBuffer(buffer: Buffer | Uint8Array, mimeType?: string): Promise<Video>;
+/**
+ * Create video content from a base64 string.
+ */
+declare function videoFromBase64(base64Data: string, mimeType?: string): Promise<Video>;
 /**
  * Create file content from a local file.
  */
@@ -1423,7 +1465,7 @@ declare function fileFromBuffer(buffer: Buffer | Uint8Array, mimeType: string):
  * Create file content from a base64 string.
  */
 declare function fileFromBase64(base64Data: string, mimeType: string): Promise<File>;
-type MultimodalContent = Text | Image | Audio | File;
+type MultimodalContent = Text | Image | Audio | Video | File;
 
 /**
  * Multimodal response classes for handling LLM multimodal outputs.
@@ -1879,4 +1921,4 @@ declare class ApprovalClient {
     waitForApproval(executionId: string, opts?: WaitForApprovalOptions): Promise<ApprovalStatusResponse>;
 }
 
-export { ACTIVE_STATUSES, AIClient, type AIConfig, type AIEmbeddingOptions, type AIRequestOptions, type AIStream, type AIToolRequestOptions, Agent, type AgentCapability, type AgentConfig, type AgentHandler, AgentRouter, type AgentRouterOptions, type AgentState, ApprovalClient, type ApprovalRequestResponse, type ApprovalStatusResponse, Audio, type AudioOutput, type AudioRequest, type AuditTrailExport, type AuditTrailFilters, type Awaitable, CANONICAL_STATUSES, type CompactCapability, type CompactDiscoveryResponse, DIDAuthenticator, type DIDIdentity, type DIDIdentityPackage, type DIDRegistrationRequest, type DIDRegistrationResponse, type DeploymentType, DidClient, DidInterface, DidManager, type DiscoveryFormat, type DiscoveryOptions, type DiscoveryPagination, type DiscoveryResponse, type DiscoveryResult, ExecutionContext, type ExecutionCredential, type ExecutionLogAttributes, type ExecutionLogBatchPayload, type ExecutionLogContext, type ExecutionLogEmitOptions, type ExecutionLogEntry, type ExecutionLogLevel, type ExecutionLogTransport, type ExecutionLogTransportPayload, type ExecutionLogWireEntry, ExecutionLogger, type ExecutionLoggerOptions, type ExecutionMetadata, ExecutionStatus, type ExecutionStatusValue, File, type FileOutput, type GenerateCredentialOptions, type GenerateCredentialParams, HEADER_CALLER_DID, HEADER_DID_NONCE, HEADER_DID_SIGNATURE, HEADER_DID_TIMESTAMP, type HarnessConfig, type HarnessOptions, type HarnessProvider, type HarnessResult, HarnessRunner, type HealthStatus, Image, type ImageOutput, type ImageRequest, type MediaProvider, MediaProviderError, type MediaResponse, MediaRouter, type MemoryChangeEvent, MemoryClient, MemoryClientBase, type MemoryConfig, MemoryEventClient, type MemoryEventHandler, type MemoryEventHistoryOptions, MemoryInterface, type MemoryRequestMetadata, type MemoryRequestOptions, type MemoryScope, type MemoryWatchHandler, type Metrics, type MultimodalContent, MultimodalResponse, OpenRouterMediaProvider, type OpenRouterMediaProviderOptions, RateLimitError, type RateLimiterOptions, type RawExecutionContext, type RawResult, type ReasonerCapability, ReasonerContext, type ReasonerDefinition, type ReasonerHandler, type ReasonerOptions, type RequestApprovalPayload, SUPPORTED_PROVIDERS, type ServerlessAdapter, type ServerlessEvent, type ServerlessResponse, type SkillCapability, SkillContext, type SkillDefinition, type SkillHandler, type SkillOptions, StatelessRateLimiter, TERMINAL_STATUSES, Text, type ToolCallConfig, type ToolCallRecord, type ToolCallTrace, type ToolsOption, type VectorSearchOptions, type VectorSearchResult, type VideoFrameImage, type VideoInputReference, type VideoRequest, type WaitForApprovalOptions, type WorkflowCredential, type WorkflowMetadata, type WorkflowProgressOptions, WorkflowReporter, type ZodSchema, audioFromBase64, audioFromBuffer, audioFromFile, audioFromUrl, buildProvider, buildToolConfig, capabilitiesToTools, capabilityToMetadataTool, capabilityToTool, createExecutionLogger, createHarnessResult, createMetrics, createMultimodalResponse, createRawResult, executeToolCallLoop, fileFromBase64, fileFromBuffer, fileFromPath, fileFromUrl, getCurrentContext, getCurrentSkillContext, imageFromBase64, imageFromBuffer, imageFromFile, imageFromUrl, isActive, isExecutionLogBatchPayload, isTerminal, normalizeExecutionLogEntry, normalizeStatus, serializeExecutionLogEntry, text };
+export { ACTIVE_STATUSES, AIClient, type AIConfig, type AIEmbeddingOptions, type AIRequestOptions, type AIStream, type AIToolRequestOptions, Agent, type AgentCapability, type AgentConfig, type AgentHandler, AgentRouter, type AgentRouterOptions, type AgentState, ApprovalClient, type ApprovalRequestResponse, type ApprovalStatusResponse, Audio, type AudioOutput, type AudioRequest, type AuditTrailExport, type AuditTrailFilters, type Awaitable, CANONICAL_STATUSES, type CompactCapability, type CompactDiscoveryResponse, DIDAuthenticator, type DIDIdentity, type DIDIdentityPackage, type DIDRegistrationRequest, type DIDRegistrationResponse, type DeploymentType, DidClient, DidInterface, DidManager, type DiscoveryFormat, type DiscoveryOptions, type DiscoveryPagination, type DiscoveryResponse, type DiscoveryResult, ExecutionContext, type ExecutionCredential, type ExecutionLogAttributes, type ExecutionLogBatchPayload, type ExecutionLogContext, type ExecutionLogEmitOptions, type ExecutionLogEntry, type ExecutionLogLevel, type ExecutionLogTransport, type ExecutionLogTransportPayload, type ExecutionLogWireEntry, ExecutionLogger, type ExecutionLoggerOptions, type ExecutionMetadata, ExecutionStatus, type ExecutionStatusValue, File, type FileOutput, type GenerateCredentialOptions, type GenerateCredentialParams, HEADER_CALLER_DID, HEADER_DID_NONCE, HEADER_DID_SIGNATURE, HEADER_DID_TIMESTAMP, type HarnessConfig, type HarnessOptions, type HarnessProvider, type HarnessResult, HarnessRunner, type HealthStatus, Image, type ImageOutput, type ImageRequest, type MediaProvider, MediaProviderError, type MediaResponse, MediaRouter, type MemoryChangeEvent, MemoryClient, MemoryClientBase, type MemoryConfig, MemoryEventClient, type MemoryEventHandler, type MemoryEventHistoryOptions, MemoryInterface, type MemoryRequestMetadata, type MemoryRequestOptions, type MemoryScope, type MemoryWatchHandler, type Metrics, type MultimodalContent, MultimodalResponse, OpenRouterMediaProvider, type OpenRouterMediaProviderOptions, RateLimitError, type RateLimiterOptions, type RawExecutionContext, type RawResult, type ReasonerCapability, ReasonerContext, type ReasonerDefinition, type ReasonerHandler, type ReasonerOptions, type RequestApprovalPayload, SUPPORTED_PROVIDERS, type ServerlessAdapter, type ServerlessEvent, type ServerlessResponse, type SkillCapability, SkillContext, type SkillDefinition, type SkillHandler, type SkillOptions, StatelessRateLimiter, TERMINAL_STATUSES, Text, type ToolCallConfig, type ToolCallRecord, type ToolCallTrace, type ToolsOption, type VectorSearchOptions, type VectorSearchResult, Video, type VideoFrameImage, type VideoInputReference, type VideoRequest, type WaitForApprovalOptions, type WorkflowCredential, type WorkflowMetadata, type WorkflowProgressOptions, WorkflowReporter, type ZodSchema, audioFromBase64, audioFromBuffer, audioFromFile, audioFromUrl, buildProvider, buildToolConfig, capabilitiesToTools, capabilityToMetadataTool, capabilityToTool, createExecutionLogger, createHarnessResult, createMetrics, createMultimodalResponse, createRawResult, executeToolCallLoop, fileFromBase64, fileFromBuffer, fileFromPath, fileFromUrl, getCurrentContext, getCurrentSkillContext, imageFromBase64, imageFromBuffer, imageFromFile, imageFromUrl, isActive, isExecutionLogBatchPayload, isTerminal, normalizeExecutionLogEntry, normalizeStatus, serializeExecutionLogEntry, text, videoFromBase64, videoFromBuffer, videoFromFile, videoFromUrl };

package/dist/index.js

@@ -4982,6 +4982,13 @@ var AUDIO_MIME_TYPES = {
   ".flac": "audio/flac",
   ".ogg": "audio/ogg"
 };
+var VIDEO_MIME_TYPES = {
+  ".mp4": "video/mp4",
+  ".mpeg": "video/mpeg",
+  ".mpg": "video/mpeg",
+  ".mov": "video/quicktime",
+  ".webm": "video/webm"
+};
 var Text = class {
   type = "text";
   text;
@@ -5079,6 +5086,43 @@ var Audio = class _Audio {
     return new _Audio({ data: base64Data, format });
   }
 };
+var Video = class _Video {
+  type = "video_url";
+  videoUrl;
+  constructor(videoUrl) {
+    this.videoUrl = videoUrl;
+  }
+  /**
+   * Create Video from a local file by converting to a base64 data URL.
+   */
+  static async fromFile(filePath) {
+    const absolutePath = resolve(filePath);
+    const buffer = await readFile(absolutePath);
+    const base64Data = buffer.toString("base64");
+    const ext = getExtension(absolutePath).toLowerCase();
+    const mimeType = VIDEO_MIME_TYPES[ext] || "video/mp4";
+    return new _Video({ url: `data:${mimeType};base64,${base64Data}` });
+  }
+  /**
+   * Create Video from a URL.
+   */
+  static fromUrl(url) {
+    return new _Video({ url });
+  }
+  /**
+   * Create Video from a buffer.
+   */
+  static async fromBuffer(buffer, mimeType = "video/mp4") {
+    const base64Data = Buffer.from(buffer).toString("base64");
+    return new _Video({ url: `data:${mimeType};base64,${base64Data}` });
+  }
+  /**
+   * Create Video from a base64 string.
+   */
+  static async fromBase64(base64Data, mimeType = "video/mp4") {
+    return new _Video({ url: `data:${mimeType};base64,${base64Data}` });
+  }
+};
 var File = class _File {
   type = "file";
   file;
@@ -5133,6 +5177,9 @@ function guessMimeType(filePath) {
   if (ext in AUDIO_MIME_TYPES) {
     return AUDIO_MIME_TYPES[ext];
   }
+  if (ext in VIDEO_MIME_TYPES) {
+    return VIDEO_MIME_TYPES[ext];
+  }
   const documentMimeTypes = {
     ".pdf": "application/pdf",
     ".doc": "application/msword",
@@ -5175,6 +5222,18 @@ async function audioFromBuffer(buffer, format = "wav") {
 async function audioFromBase64(base64Data, format = "wav") {
   return Audio.fromBase64(base64Data, format);
 }
+async function videoFromFile(filePath) {
+  return Video.fromFile(filePath);
+}
+function videoFromUrl(url) {
+  return Video.fromUrl(url);
+}
+async function videoFromBuffer(buffer, mimeType = "video/mp4") {
+  return Video.fromBuffer(buffer, mimeType);
+}
+async function videoFromBase64(base64Data, mimeType = "video/mp4") {
+  return Video.fromBase64(base64Data, mimeType);
+}
 async function fileFromPath(filePath, mimeType) {
   return File.fromFile(filePath, mimeType);
 }
@@ -5591,6 +5650,8 @@ var DEFAULT_TIMEOUT = 6e5;
 var API_TIMEOUT = 3e4;
 var DOWNLOAD_TIMEOUT = 12e4;
 var MAX_CONSECUTIVE_PARSE_ERRORS = 50;
+var DEFAULT_IMAGE_MODEL = "google/gemini-3.1-flash-image-preview";
+var DEFAULT_TTS_MODEL = "hexgrad/kokoro-82m";
 var apiKeyStore = /* @__PURE__ */ new WeakMap();
 var modelMetaStore = /* @__PURE__ */ new WeakMap();
 function emptyMediaResponse(raw) {
@@ -5599,6 +5660,9 @@ function emptyMediaResponse(raw) {
 function stripPrefix(model) {
   return model.startsWith("openrouter/") ? model.slice("openrouter/".length) : model;
 }
+function defaultVoiceForModel(model) {
+  return stripPrefix(model) === "hexgrad/kokoro-82m" ? "af_alloy" : "alloy";
+}
 function wrapPcm16AsWav(pcm, sampleRate = 24e3) {
   const channels = 1;
   const bitsPerSample = 16;
@@ -5857,7 +5921,7 @@ var OpenRouterMediaProvider = class {
   }
   // ── Image ──────────────────────────────────────────────────────────
   async generateImage(request) {
-    const model = stripPrefix(request.model ?? "openai/gpt-image-1");
+    const model = stripPrefix(request.model ?? DEFAULT_IMAGE_MODEL);
     let userContent = request.prompt;
     if (request.imageUrls && request.imageUrls.length > 0) {
       userContent = [
@@ -5946,7 +6010,8 @@ var OpenRouterMediaProvider = class {
   }
   // ── Audio ──────────────────────────────────────────────────────────
   async generateAudio(request) {
-    const model = stripPrefix(request.model ?? "openai/gpt-4o-mini-tts");
+    const model = stripPrefix(request.model ?? DEFAULT_TTS_MODEL);
+    const voice = request.voice ?? defaultVoiceForModel(model);
     const requestedFormat = request.format ?? "wav";
     const meta = await this.fetchModelMeta(model);
     const outMods = meta.outputModalities;
@@ -5955,7 +6020,7 @@ var OpenRouterMediaProvider = class {
       return this.generateAudioViaSpeechEndpoint(
         model,
         request.text,
-        request.voice ?? "alloy",
+        voice,
         requestedFormat,
         request
       );
@@ -5968,7 +6033,7 @@ var OpenRouterMediaProvider = class {
       modalities: ["text", "audio"],
       stream: true,
       audio: {
-        voice: request.voice ?? "alloy",
+        voice,
         format: wireFormat
       }
     };
@@ -6313,6 +6378,6 @@ function sleep2(ms) {
   return new Promise((resolve2) => setTimeout(resolve2, ms));
 }
 
-export { ACTIVE_STATUSES, AIClient, Agent, AgentRouter, ApprovalClient, Audio, CANONICAL_STATUSES, DIDAuthenticator, DidClient, DidInterface, DidManager, ExecutionContext, ExecutionLogger, ExecutionStatus, File, HEADER_CALLER_DID, HEADER_DID_NONCE, HEADER_DID_SIGNATURE, HEADER_DID_TIMESTAMP, HarnessRunner, Image, MediaProviderError, MediaRouter, MemoryClient, MemoryClientBase, MemoryEventClient, MemoryInterface, MultimodalResponse, OpenRouterMediaProvider, RateLimitError, ReasonerContext, SUPPORTED_PROVIDERS, SkillContext, StatelessRateLimiter, TERMINAL_STATUSES, Text, WorkflowReporter, audioFromBase64, audioFromBuffer, audioFromFile, audioFromUrl, buildProvider, buildToolConfig, capabilitiesToTools, capabilityToMetadataTool, capabilityToTool, createExecutionLogger, createHarnessResult, createMetrics, createMultimodalResponse, createRawResult, executeToolCallLoop, fileFromBase64, fileFromBuffer, fileFromPath, fileFromUrl, getCurrentContext, getCurrentSkillContext, imageFromBase64, imageFromBuffer, imageFromFile, imageFromUrl, isActive, isExecutionLogBatchPayload, isTerminal, normalizeExecutionLogEntry, normalizeStatus, serializeExecutionLogEntry, text };
+export { ACTIVE_STATUSES, AIClient, Agent, AgentRouter, ApprovalClient, Audio, CANONICAL_STATUSES, DIDAuthenticator, DidClient, DidInterface, DidManager, ExecutionContext, ExecutionLogger, ExecutionStatus, File, HEADER_CALLER_DID, HEADER_DID_NONCE, HEADER_DID_SIGNATURE, HEADER_DID_TIMESTAMP, HarnessRunner, Image, MediaProviderError, MediaRouter, MemoryClient, MemoryClientBase, MemoryEventClient, MemoryInterface, MultimodalResponse, OpenRouterMediaProvider, RateLimitError, ReasonerContext, SUPPORTED_PROVIDERS, SkillContext, StatelessRateLimiter, TERMINAL_STATUSES, Text, Video, WorkflowReporter, audioFromBase64, audioFromBuffer, audioFromFile, audioFromUrl, buildProvider, buildToolConfig, capabilitiesToTools, capabilityToMetadataTool, capabilityToTool, createExecutionLogger, createHarnessResult, createMetrics, createMultimodalResponse, createRawResult, executeToolCallLoop, fileFromBase64, fileFromBuffer, fileFromPath, fileFromUrl, getCurrentContext, getCurrentSkillContext, imageFromBase64, imageFromBuffer, imageFromFile, imageFromUrl, isActive, isExecutionLogBatchPayload, isTerminal, normalizeExecutionLogEntry, normalizeStatus, serializeExecutionLogEntry, text, videoFromBase64, videoFromBuffer, videoFromFile, videoFromUrl };
 //# sourceMappingURL=index.js.map
 //# sourceMappingURL=index.js.map