@teamlearners/clawops 0.11.0 → 0.13.0

package/dist/agent/index.d.cts

@@ -759,6 +759,10 @@ interface GeminiRealtimeOptions {
     language?: string;
     /** Send initial greeting. Default: true */
     greeting?: boolean;
+    /** Gemini VAD config. @google/genai RealtimeInputConfig 구조 그대로 전달. */
+    realtimeInputConfig?: Record<string, unknown>;
+    /** Input audio transcription config. @google/genai AudioTranscriptionConfig 구조 그대로 전달. */
+    inputAudioTranscription?: Record<string, unknown>;
 }
 declare class GeminiRealtime implements Session {
     private _apiKey;
@@ -767,6 +771,8 @@ declare class GeminiRealtime implements Session {
     private _voice;
     private _language;
     private _greeting;
+    private _realtimeInputConfig;
+    private _inputAudioTranscription;
     private _session;
     private _call;
     private _tools;

package/dist/agent/index.d.ts

@@ -759,6 +759,10 @@ interface GeminiRealtimeOptions {
     language?: string;
     /** Send initial greeting. Default: true */
     greeting?: boolean;
+    /** Gemini VAD config. @google/genai RealtimeInputConfig 구조 그대로 전달. */
+    realtimeInputConfig?: Record<string, unknown>;
+    /** Input audio transcription config. @google/genai AudioTranscriptionConfig 구조 그대로 전달. */
+    inputAudioTranscription?: Record<string, unknown>;
 }
 declare class GeminiRealtime implements Session {
     private _apiKey;
@@ -767,6 +771,8 @@ declare class GeminiRealtime implements Session {
     private _voice;
     private _language;
     private _greeting;
+    private _realtimeInputConfig;
+    private _inputAudioTranscription;
     private _session;
     private _call;
     private _tools;

package/dist/agent/index.js

@@ -1,4 +1,4 @@
-import { DEFAULT_BASE_URL, AgentError, AgentConnectionError, VERSION } from '../chunk-5IY2ZR3F.js';
+import { DEFAULT_BASE_URL, AgentError, AgentConnectionError, VERSION } from '../chunk-WH65DGWI.js';
 import pino from 'pino';
 import * as fs from 'fs';
 import * as path from 'path';
@@ -2986,6 +2986,8 @@ var GeminiRealtime = class _GeminiRealtime {
   _voice;
   _language;
   _greeting;
+  _realtimeInputConfig;
+  _inputAudioTranscription;
   // eslint-disable-next-line @typescript-eslint/no-explicit-any
   _session = null;
   _call = null;
@@ -3007,6 +3009,8 @@ var GeminiRealtime = class _GeminiRealtime {
     this._voice = options.voice ?? "Kore";
     this._language = options.language ?? "ko";
     this._greeting = options.greeting ?? true;
+    this._realtimeInputConfig = options.realtimeInputConfig ?? null;
+    this._inputAudioTranscription = options.inputAudioTranscription ?? {};
   }
   /** Inject per-call ToolRegistry. */
   setToolRegistry(registry) {
@@ -3047,11 +3051,8 @@ var GeminiRealtime = class _GeminiRealtime {
     this._closed = false;
     this._sentAudioChunks = 0;
     this._audioRemainder = Buffer.alloc(0);
-    if (!this._apiKey) {
-      throw new Error("Google API key is required. Set GOOGLE_API_KEY or pass apiKey option.");
-    }
     const { GoogleGenAI } = await import('@google/genai/node');
-    const client = new GoogleGenAI({ apiKey: this._apiKey });
+    const client = this._apiKey ? new GoogleGenAI({ apiKey: this._apiKey }) : new GoogleGenAI({});
     const config = {
       responseModalities: ["AUDIO"],
       speechConfig: {
@@ -3061,9 +3062,12 @@ var GeminiRealtime = class _GeminiRealtime {
           }
         }
       },
-      inputAudioTranscription: {},
+      inputAudioTranscription: this._inputAudioTranscription,
       outputAudioTranscription: {}
     };
+    if (this._realtimeInputConfig) {
+      config["realtimeInputConfig"] = this._realtimeInputConfig;
+    }
     if (this._systemPrompt) {
       config["systemInstruction"] = {
         parts: [{ text: this._systemPrompt }]
@@ -3073,6 +3077,7 @@ var GeminiRealtime = class _GeminiRealtime {
     if (toolSchemas.length > 0) {
       config["tools"] = [{ functionDeclarations: toolSchemas }];
     }
+    this._log.debug({ config }, "Gemini SDK final config");
     this._session = await client.live.connect({
       model: this._model,
       config,
@@ -3818,9 +3823,8 @@ var GeminiLLM = class {
   }
   async *generate(messages, options) {
     const { GoogleGenAI } = await import('@google/genai');
-    const client = new GoogleGenAI({
-      apiKey: this._options.apiKey ?? process.env["GOOGLE_API_KEY"]
-    });
+    const apiKey = this._options.apiKey ?? process.env["GOOGLE_API_KEY"];
+    const client = apiKey ? new GoogleGenAI({ apiKey }) : new GoogleGenAI();
     let systemInstruction;
     const contents = [];
     for (const msg of messages) {