@teamlearners/clawops 0.5.4 → 0.5.8

package/dist/agent/index.d.cts

@@ -470,6 +470,72 @@ declare class AudioRecorder {
     stop(): void;
 }
 
+/**
+ * PipelineSession orchestrates STT -> LLM -> TTS for a voice call.
+ */
+
+interface PipelineSessionOptions {
+    stt: STT;
+    llm: LLM;
+    tts: TTS;
+    /** System prompt for the LLM. */
+    systemPrompt?: string;
+    /** Send initial greeting. Default: true */
+    greeting?: boolean;
+    /** Language code. Default: 'ko' */
+    language?: string;
+    /** Tool registry for function calling. */
+    toolRegistry?: ToolRegistry;
+    /** Audio recorder for the session. */
+    recorder?: AudioRecorder;
+    /** LLM temperature. */
+    temperature?: number;
+    /** Max tokens for LLM generation. */
+    maxTokens?: number;
+    /** Sample rate for audio. Default: 8000 */
+    sampleRate?: number;
+    /** Whether to interrupt TTS when user starts speaking. Default: true */
+    interruptOnSpeech?: boolean;
+}
+declare class PipelineSession implements Session {
+    private _stt;
+    private _llm;
+    private _tts;
+    private _systemPrompt;
+    private _greeting;
+    private _language;
+    private _temperature;
+    private _maxTokens;
+    private _sampleRate;
+    private _interruptOnSpeech;
+    private _callSession;
+    private _tools;
+    private _recorder;
+    private _conversation;
+    private _audioBuffer;
+    private _running;
+    private _speaking;
+    private _builtinTools;
+    private _log;
+    constructor(options: PipelineSessionOptions);
+    setToolRegistry(registry: ToolRegistry): void;
+    setRecorder(recorder: AudioRecorder): void;
+    setBuiltinTools(tools: Set<BuiltinTool>): void;
+    setLogger(logger: Logger): void;
+    start(callSession: CallSession, tools?: ToolRegistry): Promise<void>;
+    feedAudio(audio: Buffer): void;
+    feedDtmf(digits: string): Promise<void>;
+    stop(): Promise<void>;
+    private _runSttLoop;
+    private _createAudioStream;
+    private _generateGreeting;
+    private _handleUserSpeech;
+    private _buildEffectiveTools;
+    private _respond;
+    private _handleToolCall;
+    private _synthesizeAndSend;
+}
+
 /**
  * OpenAI Realtime API session (speech-to-speech).
  *
@@ -487,8 +553,13 @@ interface OpenAIRealtimeOptions {
     voice?: string;
     /** Language code (BCP 47). Default: 'ko' */
     language?: string;
-    /** VAD eagerness: 'low', 'medium', 'high'. Default: 'high' */
-    eagerness?: string;
+    /**
+     * Turn detection configuration. Default: semantic_vad with eagerness 'medium'.
+     * - `{ type: 'semantic_vad', eagerness?: 'low'|'medium'|'high'|'auto', interrupt_response?: boolean }`
+     * - `{ type: 'server_vad', threshold?: number, silence_duration_ms?: number, prefix_padding_ms?: number }`
+     * - `null` to disable turn detection
+     */
+    turnDetection?: Record<string, unknown> | null;
     /** Send initial greeting. Default: true */
     greeting?: boolean;
 }
@@ -498,7 +569,7 @@ declare class OpenAIRealtime implements Session {
     private _model;
     private _voice;
     private _language;
-    private _eagerness;
+    private _turnDetection;
     private _greeting;
     private _log;
     private _builtinTools;
@@ -509,10 +580,9 @@ declare class OpenAIRealtime implements Session {
     private _tools;
     private _recorder;
     private _closed;
-    private _lastAssistantItem;
-    private _responseStartTs;
-    private _sentAudioChunks;
-    private _audioRemainder;
+    private _playback;
+    private _latestMediaTs;
+    private _pendingToolCalls;
     private _responseInProgress;
     private _onResponseDone;
     constructor(options?: OpenAIRealtimeOptions);
@@ -589,72 +659,6 @@ declare class GeminiRealtime implements Session {
     private _handleToolCall;
 }
 
-/**
- * PipelineSession orchestrates STT -> LLM -> TTS for a voice call.
- */
-
-interface PipelineSessionOptions {
-    stt: STT;
-    llm: LLM;
-    tts: TTS;
-    /** System prompt for the LLM. */
-    systemPrompt?: string;
-    /** Send initial greeting. Default: true */
-    greeting?: boolean;
-    /** Language code. Default: 'ko' */
-    language?: string;
-    /** Tool registry for function calling. */
-    toolRegistry?: ToolRegistry;
-    /** Audio recorder for the session. */
-    recorder?: AudioRecorder;
-    /** LLM temperature. */
-    temperature?: number;
-    /** Max tokens for LLM generation. */
-    maxTokens?: number;
-    /** Sample rate for audio. Default: 8000 */
-    sampleRate?: number;
-    /** Whether to interrupt TTS when user starts speaking. Default: true */
-    interruptOnSpeech?: boolean;
-}
-declare class PipelineSession implements Session {
-    private _stt;
-    private _llm;
-    private _tts;
-    private _systemPrompt;
-    private _greeting;
-    private _language;
-    private _temperature;
-    private _maxTokens;
-    private _sampleRate;
-    private _interruptOnSpeech;
-    private _callSession;
-    private _tools;
-    private _recorder;
-    private _conversation;
-    private _audioBuffer;
-    private _running;
-    private _speaking;
-    private _builtinTools;
-    private _log;
-    constructor(options: PipelineSessionOptions);
-    setToolRegistry(registry: ToolRegistry): void;
-    setRecorder(recorder: AudioRecorder): void;
-    setBuiltinTools(tools: Set<BuiltinTool>): void;
-    setLogger(logger: Logger): void;
-    start(callSession: CallSession, tools?: ToolRegistry): Promise<void>;
-    feedAudio(audio: Buffer): void;
-    feedDtmf(digits: string): Promise<void>;
-    stop(): Promise<void>;
-    private _runSttLoop;
-    private _createAudioStream;
-    private _generateGreeting;
-    private _handleUserSpeech;
-    private _buildEffectiveTools;
-    private _respond;
-    private _handleToolCall;
-    private _synthesizeAndSend;
-}
-
 /**
  * Deepgram STT (Speech-to-Text) provider.
  */
@@ -960,6 +964,27 @@ declare class XaiLLM implements LLM {
     }): AsyncGenerator<LLMChunk>;
 }
 
+/**
+ * Built-in tool 스키마 정의, 포맷 변환, 실행 헬퍼.
+ *
+ * 모든 세션(PipelineSession, OpenAIRealtime, GeminiRealtime)이 공통으로 사용하는
+ * 내장 도구 스키마를 한 곳에서 관리한다.
+ */
+
+declare const BUILTIN_TOOL_NAMES: Set<string>;
+/**
+ * 활성화된 builtin tool 스키마를 요청한 포맷으로 반환.
+ */
+declare function getBuiltinToolSchemas(builtinTools: Set<BuiltinTool> | null, fmt: 'chat' | 'realtime' | 'gemini'): Array<Record<string, unknown>>;
+/**
+ * Builtin tool을 실행하고 결과 문자열을 반환한다.
+ *
+ * `funcName`이 builtin tool이 아니면 `null`을 반환한다.
+ * `hang_up`의 경우 빈 문자열 `""`을 반환한다 (호출자가 종료 처리).
+ */
+declare function isBuiltinTool(name: string): boolean;
+declare function executeBuiltinTool(funcName: string, args: Record<string, unknown>, call: CallSession): Promise<string | null>;
+
 /**
  * Logger factory for the ClawOps agent module.
  *
@@ -979,4 +1004,4 @@ declare function createAgentLogger(userLogger?: Logger): Logger;
  */
 declare function createPipelineLogger(parent: Logger): Logger;
 
-export { type AgentEventType, AnthropicLLM, type AnthropicLLMOptions, AudioRecorder, BuiltinTool, type CallDirection, CallSession, type CallStatus, ClawOpsAgent, type ClawOpsAgentOptions, type ConversationMessage, DECODE_TABLE, DeepSeekLLM, type DeepSeekLLMOptions, DeepgramSTT, type DeepgramSTTOptions, ElevenLabsTTS, type ElevenLabsTTSOptions, FireworksLLM, type FireworksLLMOptions, type FunctionTool, GeminiLLM, type GeminiLLMOptions, GeminiRealtime, type GeminiRealtimeOptions, GroqLLM, type GroqLLMOptions, type LLM, type LLMChunk, MCPClient, type MCPServerConfig, type MCPServerHTTP, type MCPServerStdio, MistralLLM, type MistralLLMOptions, OllamaLLM, type OllamaLLMOptions, OpenAICompatLLM, type OpenAICompatLLMOptions, OpenAILLM, type OpenAILLMOptions, OpenAIRealtime, type OpenAIRealtimeOptions, type OpenAIToolDefinition, PerplexityLLM, type PerplexityLLMOptions, PipelineSession, type PipelineSessionOptions, type STT, type Session, type SpeechEvent, type TTS, TogetherLLM, type TogetherLLMOptions, ToolRegistry, type TracingConfig, XaiLLM, type XaiLLMOptions, createAgentLogger, createPipelineLogger, functionTool, getTracingConfig, mcpServerHTTP, mcpServerStdio, pcm16ToUlaw, resamplePcm16, resetTracingConfig, setTracingConfig, ulawToPcm16, zodToToolParams };
+export { type AgentEventType, AnthropicLLM, type AnthropicLLMOptions, AudioRecorder, BUILTIN_TOOL_NAMES, BuiltinTool, type CallDirection, CallSession, type CallStatus, ClawOpsAgent, type ClawOpsAgentOptions, type ConversationMessage, DECODE_TABLE, DeepSeekLLM, type DeepSeekLLMOptions, DeepgramSTT, type DeepgramSTTOptions, ElevenLabsTTS, type ElevenLabsTTSOptions, FireworksLLM, type FireworksLLMOptions, type FunctionTool, GeminiLLM, type GeminiLLMOptions, GeminiRealtime, type GeminiRealtimeOptions, GroqLLM, type GroqLLMOptions, type LLM, type LLMChunk, MCPClient, type MCPServerConfig, type MCPServerHTTP, type MCPServerStdio, MistralLLM, type MistralLLMOptions, OllamaLLM, type OllamaLLMOptions, OpenAICompatLLM, type OpenAICompatLLMOptions, OpenAILLM, type OpenAILLMOptions, OpenAIRealtime, type OpenAIRealtimeOptions, type OpenAIToolDefinition, PerplexityLLM, type PerplexityLLMOptions, PipelineSession, type PipelineSessionOptions, type STT, type Session, type SpeechEvent, type TTS, TogetherLLM, type TogetherLLMOptions, type ToolCallRequest, ToolRegistry, type TracingConfig, XaiLLM, type XaiLLMOptions, createAgentLogger, createPipelineLogger, executeBuiltinTool, functionTool, getBuiltinToolSchemas, getTracingConfig, isBuiltinTool, mcpServerHTTP, mcpServerStdio, pcm16ToUlaw, resamplePcm16, resetTracingConfig, setTracingConfig, ulawToPcm16, zodToToolParams };

package/dist/agent/index.d.ts

@@ -470,6 +470,72 @@ declare class AudioRecorder {
     stop(): void;
 }
 
+/**
+ * PipelineSession orchestrates STT -> LLM -> TTS for a voice call.
+ */
+
+interface PipelineSessionOptions {
+    stt: STT;
+    llm: LLM;
+    tts: TTS;
+    /** System prompt for the LLM. */
+    systemPrompt?: string;
+    /** Send initial greeting. Default: true */
+    greeting?: boolean;
+    /** Language code. Default: 'ko' */
+    language?: string;
+    /** Tool registry for function calling. */
+    toolRegistry?: ToolRegistry;
+    /** Audio recorder for the session. */
+    recorder?: AudioRecorder;
+    /** LLM temperature. */
+    temperature?: number;
+    /** Max tokens for LLM generation. */
+    maxTokens?: number;
+    /** Sample rate for audio. Default: 8000 */
+    sampleRate?: number;
+    /** Whether to interrupt TTS when user starts speaking. Default: true */
+    interruptOnSpeech?: boolean;
+}
+declare class PipelineSession implements Session {
+    private _stt;
+    private _llm;
+    private _tts;
+    private _systemPrompt;
+    private _greeting;
+    private _language;
+    private _temperature;
+    private _maxTokens;
+    private _sampleRate;
+    private _interruptOnSpeech;
+    private _callSession;
+    private _tools;
+    private _recorder;
+    private _conversation;
+    private _audioBuffer;
+    private _running;
+    private _speaking;
+    private _builtinTools;
+    private _log;
+    constructor(options: PipelineSessionOptions);
+    setToolRegistry(registry: ToolRegistry): void;
+    setRecorder(recorder: AudioRecorder): void;
+    setBuiltinTools(tools: Set<BuiltinTool>): void;
+    setLogger(logger: Logger): void;
+    start(callSession: CallSession, tools?: ToolRegistry): Promise<void>;
+    feedAudio(audio: Buffer): void;
+    feedDtmf(digits: string): Promise<void>;
+    stop(): Promise<void>;
+    private _runSttLoop;
+    private _createAudioStream;
+    private _generateGreeting;
+    private _handleUserSpeech;
+    private _buildEffectiveTools;
+    private _respond;
+    private _handleToolCall;
+    private _synthesizeAndSend;
+}
+
 /**
  * OpenAI Realtime API session (speech-to-speech).
  *
@@ -487,8 +553,13 @@ interface OpenAIRealtimeOptions {
     voice?: string;
     /** Language code (BCP 47). Default: 'ko' */
     language?: string;
-    /** VAD eagerness: 'low', 'medium', 'high'. Default: 'high' */
-    eagerness?: string;
+    /**
+     * Turn detection configuration. Default: semantic_vad with eagerness 'medium'.
+     * - `{ type: 'semantic_vad', eagerness?: 'low'|'medium'|'high'|'auto', interrupt_response?: boolean }`
+     * - `{ type: 'server_vad', threshold?: number, silence_duration_ms?: number, prefix_padding_ms?: number }`
+     * - `null` to disable turn detection
+     */
+    turnDetection?: Record<string, unknown> | null;
     /** Send initial greeting. Default: true */
     greeting?: boolean;
 }
@@ -498,7 +569,7 @@ declare class OpenAIRealtime implements Session {
     private _model;
     private _voice;
     private _language;
-    private _eagerness;
+    private _turnDetection;
     private _greeting;
     private _log;
     private _builtinTools;
@@ -509,10 +580,9 @@ declare class OpenAIRealtime implements Session {
     private _tools;
     private _recorder;
     private _closed;
-    private _lastAssistantItem;
-    private _responseStartTs;
-    private _sentAudioChunks;
-    private _audioRemainder;
+    private _playback;
+    private _latestMediaTs;
+    private _pendingToolCalls;
     private _responseInProgress;
     private _onResponseDone;
     constructor(options?: OpenAIRealtimeOptions);
@@ -589,72 +659,6 @@ declare class GeminiRealtime implements Session {
     private _handleToolCall;
 }
 
-/**
- * PipelineSession orchestrates STT -> LLM -> TTS for a voice call.
- */
-
-interface PipelineSessionOptions {
-    stt: STT;
-    llm: LLM;
-    tts: TTS;
-    /** System prompt for the LLM. */
-    systemPrompt?: string;
-    /** Send initial greeting. Default: true */
-    greeting?: boolean;
-    /** Language code. Default: 'ko' */
-    language?: string;
-    /** Tool registry for function calling. */
-    toolRegistry?: ToolRegistry;
-    /** Audio recorder for the session. */
-    recorder?: AudioRecorder;
-    /** LLM temperature. */
-    temperature?: number;
-    /** Max tokens for LLM generation. */
-    maxTokens?: number;
-    /** Sample rate for audio. Default: 8000 */
-    sampleRate?: number;
-    /** Whether to interrupt TTS when user starts speaking. Default: true */
-    interruptOnSpeech?: boolean;
-}
-declare class PipelineSession implements Session {
-    private _stt;
-    private _llm;
-    private _tts;
-    private _systemPrompt;
-    private _greeting;
-    private _language;
-    private _temperature;
-    private _maxTokens;
-    private _sampleRate;
-    private _interruptOnSpeech;
-    private _callSession;
-    private _tools;
-    private _recorder;
-    private _conversation;
-    private _audioBuffer;
-    private _running;
-    private _speaking;
-    private _builtinTools;
-    private _log;
-    constructor(options: PipelineSessionOptions);
-    setToolRegistry(registry: ToolRegistry): void;
-    setRecorder(recorder: AudioRecorder): void;
-    setBuiltinTools(tools: Set<BuiltinTool>): void;
-    setLogger(logger: Logger): void;
-    start(callSession: CallSession, tools?: ToolRegistry): Promise<void>;
-    feedAudio(audio: Buffer): void;
-    feedDtmf(digits: string): Promise<void>;
-    stop(): Promise<void>;
-    private _runSttLoop;
-    private _createAudioStream;
-    private _generateGreeting;
-    private _handleUserSpeech;
-    private _buildEffectiveTools;
-    private _respond;
-    private _handleToolCall;
-    private _synthesizeAndSend;
-}
-
 /**
  * Deepgram STT (Speech-to-Text) provider.
  */
@@ -960,6 +964,27 @@ declare class XaiLLM implements LLM {
     }): AsyncGenerator<LLMChunk>;
 }
 
+/**
+ * Built-in tool 스키마 정의, 포맷 변환, 실행 헬퍼.
+ *
+ * 모든 세션(PipelineSession, OpenAIRealtime, GeminiRealtime)이 공통으로 사용하는
+ * 내장 도구 스키마를 한 곳에서 관리한다.
+ */
+
+declare const BUILTIN_TOOL_NAMES: Set<string>;
+/**
+ * 활성화된 builtin tool 스키마를 요청한 포맷으로 반환.
+ */
+declare function getBuiltinToolSchemas(builtinTools: Set<BuiltinTool> | null, fmt: 'chat' | 'realtime' | 'gemini'): Array<Record<string, unknown>>;
+/**
+ * Builtin tool을 실행하고 결과 문자열을 반환한다.
+ *
+ * `funcName`이 builtin tool이 아니면 `null`을 반환한다.
+ * `hang_up`의 경우 빈 문자열 `""`을 반환한다 (호출자가 종료 처리).
+ */
+declare function isBuiltinTool(name: string): boolean;
+declare function executeBuiltinTool(funcName: string, args: Record<string, unknown>, call: CallSession): Promise<string | null>;
+
 /**
  * Logger factory for the ClawOps agent module.
  *
@@ -979,4 +1004,4 @@ declare function createAgentLogger(userLogger?: Logger): Logger;
  */
 declare function createPipelineLogger(parent: Logger): Logger;
 
-export { type AgentEventType, AnthropicLLM, type AnthropicLLMOptions, AudioRecorder, BuiltinTool, type CallDirection, CallSession, type CallStatus, ClawOpsAgent, type ClawOpsAgentOptions, type ConversationMessage, DECODE_TABLE, DeepSeekLLM, type DeepSeekLLMOptions, DeepgramSTT, type DeepgramSTTOptions, ElevenLabsTTS, type ElevenLabsTTSOptions, FireworksLLM, type FireworksLLMOptions, type FunctionTool, GeminiLLM, type GeminiLLMOptions, GeminiRealtime, type GeminiRealtimeOptions, GroqLLM, type GroqLLMOptions, type LLM, type LLMChunk, MCPClient, type MCPServerConfig, type MCPServerHTTP, type MCPServerStdio, MistralLLM, type MistralLLMOptions, OllamaLLM, type OllamaLLMOptions, OpenAICompatLLM, type OpenAICompatLLMOptions, OpenAILLM, type OpenAILLMOptions, OpenAIRealtime, type OpenAIRealtimeOptions, type OpenAIToolDefinition, PerplexityLLM, type PerplexityLLMOptions, PipelineSession, type PipelineSessionOptions, type STT, type Session, type SpeechEvent, type TTS, TogetherLLM, type TogetherLLMOptions, ToolRegistry, type TracingConfig, XaiLLM, type XaiLLMOptions, createAgentLogger, createPipelineLogger, functionTool, getTracingConfig, mcpServerHTTP, mcpServerStdio, pcm16ToUlaw, resamplePcm16, resetTracingConfig, setTracingConfig, ulawToPcm16, zodToToolParams };
+export { type AgentEventType, AnthropicLLM, type AnthropicLLMOptions, AudioRecorder, BUILTIN_TOOL_NAMES, BuiltinTool, type CallDirection, CallSession, type CallStatus, ClawOpsAgent, type ClawOpsAgentOptions, type ConversationMessage, DECODE_TABLE, DeepSeekLLM, type DeepSeekLLMOptions, DeepgramSTT, type DeepgramSTTOptions, ElevenLabsTTS, type ElevenLabsTTSOptions, FireworksLLM, type FireworksLLMOptions, type FunctionTool, GeminiLLM, type GeminiLLMOptions, GeminiRealtime, type GeminiRealtimeOptions, GroqLLM, type GroqLLMOptions, type LLM, type LLMChunk, MCPClient, type MCPServerConfig, type MCPServerHTTP, type MCPServerStdio, MistralLLM, type MistralLLMOptions, OllamaLLM, type OllamaLLMOptions, OpenAICompatLLM, type OpenAICompatLLMOptions, OpenAILLM, type OpenAILLMOptions, OpenAIRealtime, type OpenAIRealtimeOptions, type OpenAIToolDefinition, PerplexityLLM, type PerplexityLLMOptions, PipelineSession, type PipelineSessionOptions, type STT, type Session, type SpeechEvent, type TTS, TogetherLLM, type TogetherLLMOptions, type ToolCallRequest, ToolRegistry, type TracingConfig, XaiLLM, type XaiLLMOptions, createAgentLogger, createPipelineLogger, executeBuiltinTool, functionTool, getBuiltinToolSchemas, getTracingConfig, isBuiltinTool, mcpServerHTTP, mcpServerStdio, pcm16ToUlaw, resamplePcm16, resetTracingConfig, setTracingConfig, ulawToPcm16, zodToToolParams };