@realtimex/sdk 1.2.5 → 1.3.1

package/dist/index.d.mts

@@ -152,6 +152,47 @@ interface TTSChunkEvent {
         error: string;
     };
 }
+interface STTListenOptions {
+    /** STT provider (e.g., 'native', 'whisper', 'groq') */
+    provider?: string;
+    /** Language code (e.g., 'en-US') */
+    language?: string;
+    /** Specific model ID (e.g., 'onnx-community/whisper-tiny.en') */
+    model: string;
+    /** Timeout in milliseconds (default: 60000) */
+    timeout?: number;
+}
+interface STTModel {
+    id: string;
+    name: string;
+    provider?: string;
+    description?: string;
+    language?: string;
+    size?: string;
+    recommended?: boolean;
+}
+interface STTProvider {
+    id: string;
+    name: string;
+    description?: string;
+    models: STTModel[];
+}
+interface STTProvidersResponse {
+    success: boolean;
+    providers: STTProvider[];
+    error?: string;
+}
+interface STTModelsResponse {
+    success: boolean;
+    models: STTModel[];
+    error?: string;
+}
+interface STTResponse {
+    success: boolean;
+    /** Transcribed text */
+    text: string;
+    error?: string;
+}
 
 /**
  * Activities Module - HTTP Proxy to RealtimeX Main App
@@ -248,7 +289,7 @@ declare class ApiModule {
     /**
      * Make an API call with automatic permission handling
      */
-    private apiCall;
+    protected apiCall<T>(method: string, endpoint: string, options?: RequestInit): Promise<T>;
     getAgents(): Promise<Agent[]>;
     getWorkspaces(): Promise<Workspace[]>;
     getThreads(workspaceSlug: string): Promise<Thread[]>;
@@ -755,6 +796,22 @@ declare class TTSModule {
     listProviders(): Promise<TTSProvider[]>;
 }
 
+declare class STTModule extends ApiModule {
+    /**
+     * Get available STT providers and their models.
+     * @returns Promise resolving to list of providers
+     */
+    listProviders(): Promise<STTProvider[]>;
+    /**
+     * Listen to microphone and transcribe speech to text.
+     * Performed on the client device (Electron) via the RealtimeX Hub.
+     *
+     * @param options Configuration options for listening
+     * @returns Promise resolving to the transcribed text
+     */
+    listen(options: STTListenOptions): Promise<STTResponse>;
+}
+
 /**
  * RealtimeX Local App SDK
  *
@@ -770,6 +827,7 @@ declare class RealtimeXSDK {
     port: PortModule;
     llm: LLMModule;
     tts: TTSModule;
+    stt: STTModule;
     readonly appId: string;
     readonly appName: string | undefined;
     readonly apiKey: string | undefined;
@@ -803,4 +861,4 @@ declare class RealtimeXSDK {
     getAppDataDir(): Promise<string>;
 }
 
-export { ActivitiesModule, type Activity, type Agent, ApiModule, type ChatMessage, type ChatOptions, type ChatResponse, type EmbedOptions, type EmbedResponse, LLMModule, LLMPermissionError, LLMProviderError, PermissionDeniedError, PermissionRequiredError, PortModule, type Provider, type ProvidersResponse, RealtimeXSDK, type SDKConfig, type StreamChunk, type TTSChunk, type TTSChunkEvent, TTSModule, type TTSOptions, type TTSProvider, type TTSProviderConfig, type TTSProvidersResponse, type Task, TaskModule, type TaskRun, type Thread, type TriggerAgentPayload, type TriggerAgentResponse, type VectorDeleteOptions, type VectorDeleteResponse, type VectorQueryOptions, type VectorQueryResponse, type VectorQueryResult, type VectorRecord, VectorStore, type VectorUpsertOptions, type VectorUpsertResponse, WebhookModule, type Workspace };
+export { ActivitiesModule, type Activity, type Agent, ApiModule, type ChatMessage, type ChatOptions, type ChatResponse, type EmbedOptions, type EmbedResponse, LLMModule, LLMPermissionError, LLMProviderError, PermissionDeniedError, PermissionRequiredError, PortModule, type Provider, type ProvidersResponse, RealtimeXSDK, type SDKConfig, type STTListenOptions, type STTModel, type STTModelsResponse, STTModule, type STTProvider, type STTProvidersResponse, type STTResponse, type StreamChunk, type TTSChunk, type TTSChunkEvent, TTSModule, type TTSOptions, type TTSProvider, type TTSProviderConfig, type TTSProvidersResponse, type Task, TaskModule, type TaskRun, type Thread, type TriggerAgentPayload, type TriggerAgentResponse, type VectorDeleteOptions, type VectorDeleteResponse, type VectorQueryOptions, type VectorQueryResponse, type VectorQueryResult, type VectorRecord, VectorStore, type VectorUpsertOptions, type VectorUpsertResponse, WebhookModule, type Workspace };

package/dist/index.d.ts

@@ -152,6 +152,47 @@ interface TTSChunkEvent {
         error: string;
     };
 }
+interface STTListenOptions {
+    /** STT provider (e.g., 'native', 'whisper', 'groq') */
+    provider?: string;
+    /** Language code (e.g., 'en-US') */
+    language?: string;
+    /** Specific model ID (e.g., 'onnx-community/whisper-tiny.en') */
+    model: string;
+    /** Timeout in milliseconds (default: 60000) */
+    timeout?: number;
+}
+interface STTModel {
+    id: string;
+    name: string;
+    provider?: string;
+    description?: string;
+    language?: string;
+    size?: string;
+    recommended?: boolean;
+}
+interface STTProvider {
+    id: string;
+    name: string;
+    description?: string;
+    models: STTModel[];
+}
+interface STTProvidersResponse {
+    success: boolean;
+    providers: STTProvider[];
+    error?: string;
+}
+interface STTModelsResponse {
+    success: boolean;
+    models: STTModel[];
+    error?: string;
+}
+interface STTResponse {
+    success: boolean;
+    /** Transcribed text */
+    text: string;
+    error?: string;
+}
 
 /**
  * Activities Module - HTTP Proxy to RealtimeX Main App
@@ -248,7 +289,7 @@ declare class ApiModule {
     /**
      * Make an API call with automatic permission handling
      */
-    private apiCall;
+    protected apiCall<T>(method: string, endpoint: string, options?: RequestInit): Promise<T>;
     getAgents(): Promise<Agent[]>;
     getWorkspaces(): Promise<Workspace[]>;
     getThreads(workspaceSlug: string): Promise<Thread[]>;
@@ -755,6 +796,22 @@ declare class TTSModule {
     listProviders(): Promise<TTSProvider[]>;
 }
 
+declare class STTModule extends ApiModule {
+    /**
+     * Get available STT providers and their models.
+     * @returns Promise resolving to list of providers
+     */
+    listProviders(): Promise<STTProvider[]>;
+    /**
+     * Listen to microphone and transcribe speech to text.
+     * Performed on the client device (Electron) via the RealtimeX Hub.
+     *
+     * @param options Configuration options for listening
+     * @returns Promise resolving to the transcribed text
+     */
+    listen(options: STTListenOptions): Promise<STTResponse>;
+}
+
 /**
  * RealtimeX Local App SDK
  *
@@ -770,6 +827,7 @@ declare class RealtimeXSDK {
     port: PortModule;
     llm: LLMModule;
     tts: TTSModule;
+    stt: STTModule;
     readonly appId: string;
     readonly appName: string | undefined;
     readonly apiKey: string | undefined;
@@ -803,4 +861,4 @@ declare class RealtimeXSDK {
     getAppDataDir(): Promise<string>;
 }
 
-export { ActivitiesModule, type Activity, type Agent, ApiModule, type ChatMessage, type ChatOptions, type ChatResponse, type EmbedOptions, type EmbedResponse, LLMModule, LLMPermissionError, LLMProviderError, PermissionDeniedError, PermissionRequiredError, PortModule, type Provider, type ProvidersResponse, RealtimeXSDK, type SDKConfig, type StreamChunk, type TTSChunk, type TTSChunkEvent, TTSModule, type TTSOptions, type TTSProvider, type TTSProviderConfig, type TTSProvidersResponse, type Task, TaskModule, type TaskRun, type Thread, type TriggerAgentPayload, type TriggerAgentResponse, type VectorDeleteOptions, type VectorDeleteResponse, type VectorQueryOptions, type VectorQueryResponse, type VectorQueryResult, type VectorRecord, VectorStore, type VectorUpsertOptions, type VectorUpsertResponse, WebhookModule, type Workspace };
+export { ActivitiesModule, type Activity, type Agent, ApiModule, type ChatMessage, type ChatOptions, type ChatResponse, type EmbedOptions, type EmbedResponse, LLMModule, LLMPermissionError, LLMProviderError, PermissionDeniedError, PermissionRequiredError, PortModule, type Provider, type ProvidersResponse, RealtimeXSDK, type SDKConfig, type STTListenOptions, type STTModel, type STTModelsResponse, STTModule, type STTProvider, type STTProvidersResponse, type STTResponse, type StreamChunk, type TTSChunk, type TTSChunkEvent, TTSModule, type TTSOptions, type TTSProvider, type TTSProviderConfig, type TTSProvidersResponse, type Task, TaskModule, type TaskRun, type Thread, type TriggerAgentPayload, type TriggerAgentResponse, type VectorDeleteOptions, type VectorDeleteResponse, type VectorQueryOptions, type VectorQueryResponse, type VectorQueryResult, type VectorRecord, VectorStore, type VectorUpsertOptions, type VectorUpsertResponse, WebhookModule, type Workspace };

package/dist/index.js

@@ -39,6 +39,7 @@ __export(index_exports, {
   PermissionRequiredError: () => PermissionRequiredError,
   PortModule: () => PortModule,
   RealtimeXSDK: () => RealtimeXSDK,
+  STTModule: () => STTModule,
   TTSModule: () => TTSModule,
   TaskModule: () => TaskModule,
   VectorStore: () => VectorStore,
@@ -1166,6 +1167,41 @@ var TTSModule = class {
   }
 };
 
+// src/modules/stt.ts
+var STTModule = class extends ApiModule {
+  /**
+   * Get available STT providers and their models.
+   * @returns Promise resolving to list of providers
+   */
+  async listProviders() {
+    try {
+      const response = await this.apiCall("GET", "/sdk/stt/providers");
+      if (!response.success) {
+        throw new Error(response.error || "Failed to fetch providers");
+      }
+      return response.providers;
+    } catch (error) {
+      throw new Error(`STT providers fetch failed: ${error.message}`);
+    }
+  }
+  /**
+   * Listen to microphone and transcribe speech to text.
+   * Performed on the client device (Electron) via the RealtimeX Hub.
+   * 
+   * @param options Configuration options for listening
+   * @returns Promise resolving to the transcribed text
+   */
+  async listen(options) {
+    try {
+      return await this.apiCall("POST", "/sdk/stt/listen", {
+        body: JSON.stringify(options)
+      });
+    } catch (error) {
+      throw new Error(`STT listen failed: ${error.message}`);
+    }
+  }
+};
+
 // src/index.ts
 var _RealtimeXSDK = class _RealtimeXSDK {
   constructor(config = {}) {
@@ -1184,6 +1220,7 @@ var _RealtimeXSDK = class _RealtimeXSDK {
     this.port = new PortModule(config.defaultPort);
     this.llm = new LLMModule(this.realtimexUrl, this.appId, this.appName, this.apiKey);
     this.tts = new TTSModule(this.realtimexUrl, this.appId, this.appName, this.apiKey);
+    this.stt = new STTModule(this.realtimexUrl, this.appId, this.appName, this.apiKey);
     if (this.permissions.length > 0 && this.appId && !this.apiKey) {
       this.register().catch((err) => {
         console.error("[RealtimeX SDK] Auto-registration failed:", err.message);
@@ -1294,6 +1331,7 @@ var RealtimeXSDK = _RealtimeXSDK;
   PermissionRequiredError,
   PortModule,
   RealtimeXSDK,
+  STTModule,
   TTSModule,
   TaskModule,
   VectorStore,

package/dist/index.mjs

@@ -1118,6 +1118,41 @@ var TTSModule = class {
   }
 };
 
+// src/modules/stt.ts
+var STTModule = class extends ApiModule {
+  /**
+   * Get available STT providers and their models.
+   * @returns Promise resolving to list of providers
+   */
+  async listProviders() {
+    try {
+      const response = await this.apiCall("GET", "/sdk/stt/providers");
+      if (!response.success) {
+        throw new Error(response.error || "Failed to fetch providers");
+      }
+      return response.providers;
+    } catch (error) {
+      throw new Error(`STT providers fetch failed: ${error.message}`);
+    }
+  }
+  /**
+   * Listen to microphone and transcribe speech to text.
+   * Performed on the client device (Electron) via the RealtimeX Hub.
+   * 
+   * @param options Configuration options for listening
+   * @returns Promise resolving to the transcribed text
+   */
+  async listen(options) {
+    try {
+      return await this.apiCall("POST", "/sdk/stt/listen", {
+        body: JSON.stringify(options)
+      });
+    } catch (error) {
+      throw new Error(`STT listen failed: ${error.message}`);
+    }
+  }
+};
+
 // src/index.ts
 var _RealtimeXSDK = class _RealtimeXSDK {
   constructor(config = {}) {
@@ -1136,6 +1171,7 @@ var _RealtimeXSDK = class _RealtimeXSDK {
     this.port = new PortModule(config.defaultPort);
     this.llm = new LLMModule(this.realtimexUrl, this.appId, this.appName, this.apiKey);
     this.tts = new TTSModule(this.realtimexUrl, this.appId, this.appName, this.apiKey);
+    this.stt = new STTModule(this.realtimexUrl, this.appId, this.appName, this.apiKey);
     if (this.permissions.length > 0 && this.appId && !this.apiKey) {
       this.register().catch((err) => {
         console.error("[RealtimeX SDK] Auto-registration failed:", err.message);
@@ -1245,6 +1281,7 @@ export {
   PermissionRequiredError,
   PortModule,
   RealtimeXSDK,
+  STTModule,
   TTSModule,
   TaskModule,
   VectorStore,

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@realtimex/sdk",
-    "version": "1.2.5",
+    "version": "1.3.1",
     "description": "SDK for building Local Apps that integrate with RealtimeX",
     "main": "dist/index.js",
     "module": "dist/index.mjs",