@dtelecom/agents-js 0.1.1 → 0.1.4

package/dist/index.d.mts

@@ -1,8 +1,8 @@
 import * as _dtelecom_server_sdk_node from '@dtelecom/server-sdk-node';
 import { Room, AudioSource, RemoteAudioTrack, AudioFrame } from '@dtelecom/server-sdk-node';
 import { EventEmitter } from 'events';
-import { A as AgentConfig, a as AgentStartOptions, M as Message, L as LLMPlugin, P as PipelineOptions, b as AgentState, S as STTStream, T as TranscriptionResult } from './types-Cs5uUoTC.mjs';
-export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, E as Embedder, e as LLMChunk, f as MemoryConfig, g as PipelineEvents, R as RespondMode, h as RoomMemory, i as RoomMemoryConfig, j as STTPlugin, k as STTStreamOptions, l as TTSPlugin } from './types-Cs5uUoTC.mjs';
+import { A as AgentConfig, a as AgentStartOptions, M as Message, L as LLMPlugin, P as PipelineOptions, b as AgentState, S as STTStream, T as TranscriptionResult } from './types-EvtHMokR.mjs';
+export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, e as LLMChunk, f as MemoryConfig, g as PipelineEvents, R as RespondMode, h as STTPlugin, i as STTStreamOptions, j as TTSPlugin } from './types-EvtHMokR.mjs';
 
 declare class VoiceAgent extends EventEmitter {
     private readonly config;
@@ -111,6 +111,9 @@ declare class Pipeline extends EventEmitter {
     /** Queued turn while current one is still processing */
     private pendingTurn;
     constructor(options: PipelineOptions);
+    /** One-shot warmup — safe to call from constructor, resolves when both LLM and TTS are ready. */
+    private _warmupPromise;
+    private warmup;
     get processing(): boolean;
     get running(): boolean;
     get agentState(): AgentState;
@@ -263,72 +266,6 @@ declare class AudioInput {
     close(): void;
 }
 
-/**
- * MemoryStore — SQLite + sqlite-vec database layer for room memory.
- *
- * Single .db file stores:
- * - turns: every spoken turn (full transcript)
- * - sessions: meeting metadata + LLM-generated summaries
- * - turn_vectors: embedding index for semantic turn search
- * - session_vectors: embedding index for session summary search
- */
-interface TurnRow {
-    id: number;
-    room: string;
-    session_id: string;
-    speaker: string;
-    text: string;
-    is_agent: number;
-    created_at: number;
-}
-interface SessionRow {
-    id: string;
-    room: string;
-    started_at: number;
-    ended_at: number | null;
-    participants: string | null;
-    summary: string | null;
-    turn_count: number;
-}
-interface SearchResult {
-    speaker: string;
-    text: string;
-    created_at: number;
-    session_id: string;
-    distance: number;
-}
-interface SessionSearchResult {
-    session_id: string;
-    summary: string;
-    started_at: number;
-    distance: number;
-}
-declare class MemoryStore {
-    private db;
-    constructor(dbPath: string);
-    private createTables;
-    /** Insert a turn and its embedding vector. */
-    insertTurn(room: string, sessionId: string, speaker: string, text: string, isAgent: boolean, embedding: Float32Array): number;
-    /** Create a new session record. */
-    insertSession(id: string, room: string): void;
-    /** Update a session with summary and end time. */
-    updateSessionSummary(sessionId: string, summary: string, turnCount: number, participants: string[], embedding: Float32Array): void;
-    /** End a session without summary (e.g., too few turns). */
-    endSession(sessionId: string, turnCount: number, participants: string[]): void;
-    /** KNN search turns by embedding similarity. */
-    searchTurns(room: string, queryEmbedding: Float32Array, limit: number): SearchResult[];
-    /** KNN search session summaries by embedding similarity. */
-    searchSessions(room: string, queryEmbedding: Float32Array, limit: number): SessionSearchResult[];
-    /** Get the last N turns from a specific session. */
-    getRecentTurns(room: string, sessionId: string, limit: number): TurnRow[];
-    /** Get all turns for a session (for summarization). */
-    getSessionTurns(sessionId: string): TurnRow[];
-    /** Get total turn count for a session. */
-    getSessionTurnCount(sessionId: string): number;
-    /** Close the database. */
-    close(): void;
-}
-
 type LogLevel = 'debug' | 'info' | 'warn' | 'error' | 'silent';
 declare function setLogLevel(level: LogLevel): void;
 declare function getLogLevel(): LogLevel;
@@ -340,4 +277,4 @@ interface Logger {
 }
 declare function createLogger(tag: string): Logger;
 
-export { AgentConfig, AgentStartOptions, AgentState, AudioInput, BargeIn, BaseSTTStream, ContextManager, type ContextManagerOptions, LLMPlugin, type LogLevel, type Logger, MemoryStore, Message, Pipeline, PipelineOptions, RoomConnection, type RoomConnectionOptions, STTStream, type SearchResult, SentenceSplitter, type SessionRow, type SessionSearchResult, TranscriptionResult, TurnDetector, type TurnDetectorOptions, type TurnRow, VoiceAgent, createLogger, getLogLevel, setLogLevel };
+export { AgentConfig, AgentStartOptions, AgentState, AudioInput, BargeIn, BaseSTTStream, ContextManager, type ContextManagerOptions, LLMPlugin, type LogLevel, type Logger, Message, Pipeline, PipelineOptions, RoomConnection, type RoomConnectionOptions, STTStream, SentenceSplitter, TranscriptionResult, TurnDetector, type TurnDetectorOptions, VoiceAgent, createLogger, getLogLevel, setLogLevel };

package/dist/index.d.ts

@@ -1,8 +1,8 @@
 import * as _dtelecom_server_sdk_node from '@dtelecom/server-sdk-node';
 import { Room, AudioSource, RemoteAudioTrack, AudioFrame } from '@dtelecom/server-sdk-node';
 import { EventEmitter } from 'events';
-import { A as AgentConfig, a as AgentStartOptions, M as Message, L as LLMPlugin, P as PipelineOptions, b as AgentState, S as STTStream, T as TranscriptionResult } from './types-Cs5uUoTC.js';
-export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, E as Embedder, e as LLMChunk, f as MemoryConfig, g as PipelineEvents, R as RespondMode, h as RoomMemory, i as RoomMemoryConfig, j as STTPlugin, k as STTStreamOptions, l as TTSPlugin } from './types-Cs5uUoTC.js';
+import { A as AgentConfig, a as AgentStartOptions, M as Message, L as LLMPlugin, P as PipelineOptions, b as AgentState, S as STTStream, T as TranscriptionResult } from './types-EvtHMokR.js';
+export { c as AgentEvents, d as AudioOutput, D as DataMessageHandler, e as LLMChunk, f as MemoryConfig, g as PipelineEvents, R as RespondMode, h as STTPlugin, i as STTStreamOptions, j as TTSPlugin } from './types-EvtHMokR.js';
 
 declare class VoiceAgent extends EventEmitter {
     private readonly config;
@@ -111,6 +111,9 @@ declare class Pipeline extends EventEmitter {
     /** Queued turn while current one is still processing */
     private pendingTurn;
     constructor(options: PipelineOptions);
+    /** One-shot warmup — safe to call from constructor, resolves when both LLM and TTS are ready. */
+    private _warmupPromise;
+    private warmup;
     get processing(): boolean;
     get running(): boolean;
     get agentState(): AgentState;
@@ -263,72 +266,6 @@ declare class AudioInput {
     close(): void;
 }
 
-/**
- * MemoryStore — SQLite + sqlite-vec database layer for room memory.
- *
- * Single .db file stores:
- * - turns: every spoken turn (full transcript)
- * - sessions: meeting metadata + LLM-generated summaries
- * - turn_vectors: embedding index for semantic turn search
- * - session_vectors: embedding index for session summary search
- */
-interface TurnRow {
-    id: number;
-    room: string;
-    session_id: string;
-    speaker: string;
-    text: string;
-    is_agent: number;
-    created_at: number;
-}
-interface SessionRow {
-    id: string;
-    room: string;
-    started_at: number;
-    ended_at: number | null;
-    participants: string | null;
-    summary: string | null;
-    turn_count: number;
-}
-interface SearchResult {
-    speaker: string;
-    text: string;
-    created_at: number;
-    session_id: string;
-    distance: number;
-}
-interface SessionSearchResult {
-    session_id: string;
-    summary: string;
-    started_at: number;
-    distance: number;
-}
-declare class MemoryStore {
-    private db;
-    constructor(dbPath: string);
-    private createTables;
-    /** Insert a turn and its embedding vector. */
-    insertTurn(room: string, sessionId: string, speaker: string, text: string, isAgent: boolean, embedding: Float32Array): number;
-    /** Create a new session record. */
-    insertSession(id: string, room: string): void;
-    /** Update a session with summary and end time. */
-    updateSessionSummary(sessionId: string, summary: string, turnCount: number, participants: string[], embedding: Float32Array): void;
-    /** End a session without summary (e.g., too few turns). */
-    endSession(sessionId: string, turnCount: number, participants: string[]): void;
-    /** KNN search turns by embedding similarity. */
-    searchTurns(room: string, queryEmbedding: Float32Array, limit: number): SearchResult[];
-    /** KNN search session summaries by embedding similarity. */
-    searchSessions(room: string, queryEmbedding: Float32Array, limit: number): SessionSearchResult[];
-    /** Get the last N turns from a specific session. */
-    getRecentTurns(room: string, sessionId: string, limit: number): TurnRow[];
-    /** Get all turns for a session (for summarization). */
-    getSessionTurns(sessionId: string): TurnRow[];
-    /** Get total turn count for a session. */
-    getSessionTurnCount(sessionId: string): number;
-    /** Close the database. */
-    close(): void;
-}
-
 type LogLevel = 'debug' | 'info' | 'warn' | 'error' | 'silent';
 declare function setLogLevel(level: LogLevel): void;
 declare function getLogLevel(): LogLevel;
@@ -340,4 +277,4 @@ interface Logger {
 }
 declare function createLogger(tag: string): Logger;
 
-export { AgentConfig, AgentStartOptions, AgentState, AudioInput, BargeIn, BaseSTTStream, ContextManager, type ContextManagerOptions, LLMPlugin, type LogLevel, type Logger, MemoryStore, Message, Pipeline, PipelineOptions, RoomConnection, type RoomConnectionOptions, STTStream, type SearchResult, SentenceSplitter, type SessionRow, type SessionSearchResult, TranscriptionResult, TurnDetector, type TurnDetectorOptions, type TurnRow, VoiceAgent, createLogger, getLogLevel, setLogLevel };
+export { AgentConfig, AgentStartOptions, AgentState, AudioInput, BargeIn, BaseSTTStream, ContextManager, type ContextManagerOptions, LLMPlugin, type LogLevel, type Logger, Message, Pipeline, PipelineOptions, RoomConnection, type RoomConnectionOptions, STTStream, SentenceSplitter, TranscriptionResult, TurnDetector, type TurnDetectorOptions, VoiceAgent, createLogger, getLogLevel, setLogLevel };

package/dist/index.js

@@ -519,11 +519,8 @@ __export(src_exports, {
   BargeIn: () => BargeIn,
   BaseSTTStream: () => BaseSTTStream,
   ContextManager: () => ContextManager,
-  Embedder: () => Embedder,
-  MemoryStore: () => MemoryStore,
   Pipeline: () => Pipeline,
   RoomConnection: () => RoomConnection,
-  RoomMemory: () => RoomMemory,
   SentenceSplitter: () => SentenceSplitter,
   TurnDetector: () => TurnDetector,
   VoiceAgent: () => VoiceAgent,
@@ -705,6 +702,9 @@ var AudioOutput = class {
   startSilence() {
     if (this.silenceInterval) return;
     log3.debug("Starting silence keepalive (sparse, 3s interval)");
+    const immediate = new import_server_sdk_node2.AudioFrame(SILENCE, SAMPLE_RATE, CHANNELS, SAMPLES_PER_FRAME);
+    this.source.captureFrame(immediate).catch(() => {
+    });
     this.silenceInterval = setInterval(() => {
       if (!this._playing && !this._responding && !this._stopped) {
         const f = new import_server_sdk_node2.AudioFrame(SILENCE, SAMPLE_RATE, CHANNELS, SAMPLES_PER_FRAME);
@@ -1176,16 +1176,27 @@ var Pipeline = class extends import_events.EventEmitter {
       this.splitter.reset();
       this.setAgentState("idle");
     };
+    this._warmupPromise = this.warmup(options.instructions);
+  }
+  /** One-shot warmup — safe to call from constructor, resolves when both LLM and TTS are ready. */
+  _warmupPromise;
+  async warmup(instructions) {
+    const tasks = [];
     if (this.llm.warmup) {
-      this.llm.warmup(options.instructions).catch((err) => {
-        log7.warn("LLM warmup failed:", err);
-      });
+      tasks.push(
+        this.llm.warmup(instructions).catch((err) => {
+          log7.warn("LLM warmup failed:", err);
+        })
+      );
     }
     if (this.tts?.warmup) {
-      this.tts.warmup().catch((err) => {
-        log7.warn("TTS warmup failed:", err);
-      });
+      tasks.push(
+        this.tts.warmup().catch((err) => {
+          log7.warn("TTS warmup failed:", err);
+        })
+      );
     }
+    await Promise.all(tasks);
   }
   get processing() {
     return this._processing;
@@ -1292,6 +1303,7 @@ var Pipeline = class extends import_events.EventEmitter {
       return;
     }
     this._processing = true;
+    await this._warmupPromise;
     const tSpeechEnd = this.lastFinalAt;
     const sttDuration = this.lastSttDuration;
     let tLlmFirstToken = 0;
@@ -1445,6 +1457,7 @@ var Pipeline = class extends import_events.EventEmitter {
       return;
     }
     this._processing = true;
+    await this._warmupPromise;
     log7.info(`say(): "${text.slice(0, 60)}"`);
     try {
       const signal = this.bargeIn.startCycle();
@@ -1724,9 +1737,6 @@ var BaseSTTStream = class extends import_events3.EventEmitter {
 };
 
 // src/index.ts
-init_embedder();
-init_memory_store();
-init_room_memory();
 init_logger();
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
@@ -1735,11 +1745,8 @@ init_logger();
   BargeIn,
   BaseSTTStream,
   ContextManager,
-  Embedder,
-  MemoryStore,
   Pipeline,
   RoomConnection,
-  RoomMemory,
   SentenceSplitter,
   TurnDetector,
   VoiceAgent,