@desktalk/core 0.1.0-alpha.1

package/dist/services/voice/openai-whisper-adapter.js

@@ -0,0 +1,61 @@
+/**
+ * OpenAI Whisper STT adapter.
+ *
+ * Uses the /v1/audio/transcriptions endpoint to transcribe complete
+ * audio buffers. Audio is sent as WAV (PCM s16le, 16kHz, mono) since
+ * Whisper requires a file-like format.
+ */
+import { createWavBuffer } from './audio-format.js';
+export class OpenAIWhisperAdapter {
+    name = 'openai-whisper';
+    apiKey;
+    model;
+    baseUrl;
+    constructor(options) {
+        this.apiKey = options.apiKey;
+        this.model = options.model ?? 'whisper-1';
+        this.baseUrl = options.baseUrl ?? 'https://api.openai.com/v1';
+    }
+    async transcribe(audioBuffer, sampleRate, language) {
+        const wavBuffer = createWavBuffer(audioBuffer, sampleRate);
+        const durationMs = Math.round((audioBuffer.length / 2 / sampleRate) * 1000);
+        // Build multipart/form-data manually to avoid heavy dependencies
+        const boundary = `----DeskTalkBoundary${Date.now()}`;
+        const parts = [];
+        // model field
+        parts.push(Buffer.from(`--${boundary}\r\nContent-Disposition: form-data; name="model"\r\n\r\n${this.model}\r\n`));
+        // response_format field
+        parts.push(Buffer.from(`--${boundary}\r\nContent-Disposition: form-data; name="response_format"\r\n\r\njson\r\n`));
+        console.log("Language:", language);
+        // language field (ISO-639-1 code, e.g. 'en', 'zh')
+        if (language) {
+            parts.push(Buffer.from(`--${boundary}\r\nContent-Disposition: form-data; name="language"\r\n\r\n${language}\r\n`));
+        }
+        // file field
+        parts.push(Buffer.from(`--${boundary}\r\nContent-Disposition: form-data; name="file"; filename="audio.wav"\r\nContent-Type: audio/wav\r\n\r\n`));
+        parts.push(wavBuffer);
+        parts.push(Buffer.from(`\r\n--${boundary}--\r\n`));
+        const body = Buffer.concat(parts);
+        const response = await fetch(`${this.baseUrl}/audio/transcriptions`, {
+            method: 'POST',
+            headers: {
+                Authorization: `Bearer ${this.apiKey}`,
+                'Content-Type': `multipart/form-data; boundary=${boundary}`,
+            },
+            body,
+        });
+        if (!response.ok) {
+            const errorText = await response.text();
+            throw new Error(`OpenAI Whisper API error (${response.status}): ${errorText}`);
+        }
+        const result = (await response.json());
+        return {
+            text: result.text.trim(),
+            durationMs,
+        };
+    }
+    close() {
+        // No persistent resources to clean up
+    }
+}
+//# sourceMappingURL=openai-whisper-adapter.js.map

package/dist/services/voice/openai-whisper-adapter.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"openai-whisper-adapter.js","sourceRoot":"","sources":["../../../src/services/voice/openai-whisper-adapter.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAGH,OAAO,EAAE,eAAe,EAAE,MAAM,gBAAgB,CAAC;AAEjD,MAAM,OAAO,oBAAoB;IACtB,IAAI,GAAG,gBAAgB,CAAC;IACzB,MAAM,CAAS;IACf,KAAK,CAAS;IACd,OAAO,CAAS;IAExB,YAAY,OAA6D;QACvE,IAAI,CAAC,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC;QAC7B,IAAI,CAAC,KAAK,GAAG,OAAO,CAAC,KAAK,IAAI,WAAW,CAAC;QAC1C,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC,OAAO,IAAI,2BAA2B,CAAC;IAChE,CAAC;IAED,KAAK,CAAC,UAAU,CACd,WAAmB,EACnB,UAAkB,EAClB,QAAiB;QAEjB,MAAM,SAAS,GAAG,eAAe,CAAC,WAAW,EAAE,UAAU,CAAC,CAAC;QAC3D,MAAM,UAAU,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,WAAW,CAAC,MAAM,GAAG,CAAC,GAAG,UAAU,CAAC,GAAG,IAAI,CAAC,CAAC;QAE5E,iEAAiE;QACjE,MAAM,QAAQ,GAAG,uBAAuB,IAAI,CAAC,GAAG,EAAE,EAAE,CAAC;QACrD,MAAM,KAAK,GAAa,EAAE,CAAC;QAE3B,cAAc;QACd,KAAK,CAAC,IAAI,CACR,MAAM,CAAC,IAAI,CACT,KAAK,QAAQ,2DAA2D,IAAI,CAAC,KAAK,MAAM,CACzF,CACF,CAAC;QAEF,wBAAwB;QACxB,KAAK,CAAC,IAAI,CACR,MAAM,CAAC,IAAI,CACT,KAAK,QAAQ,4EAA4E,CAC1F,CACF,CAAC;QACF,OAAO,CAAC,GAAG,CAAC,WAAW,EAAE,QAAQ,CAAC,CAAC;QAEnC,mDAAmD;QACnD,IAAI,QAAQ,EAAE,CAAC;YACb,KAAK,CAAC,IAAI,CACR,MAAM,CAAC,IAAI,CACT,KAAK,QAAQ,8DAA8D,QAAQ,MAAM,CAC1F,CACF,CAAC;QACJ,CAAC;QAED,aAAa;QACb,KAAK,CAAC,IAAI,CACR,MAAM,CAAC,IAAI,CACT,KAAK,QAAQ,0GAA0G,CACxH,CACF,CAAC;QACF,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;QACtB,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,SAAS,QAAQ,QAAQ,CAAC,CAAC,CAAC;QAEnD,MAAM,IAAI,GAAG,MAAM,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;QAElC,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,IAAI,CAAC,OAAO,uBAAuB,EAAE;YACnE,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,aAAa,EAAE,UAAU,IAAI,CAAC,MAAM,EAAE;gBACtC,cAAc,EAAE,iCAAiC,QAAQ,EAAE;aAC5D;YACD,IAAI;SACL,CAAC,CAAC;QAEH,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,SAAS,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACxC,MAAM,IAAI,KAAK,CAAC,6BAA6B,QAAQ,CAAC,MAAM,MAAM,SAAS,EAAE,CAAC,CAAC;QACjF,CAAC;QAED,MAAM,MAAM,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAqB,CAAC;QAE3D,OAAO;YACL,IAAI,EAAE,MAAM,CAAC,IAAI,CAAC,IAAI,EAAE;YACxB,UAAU;SACX,CAAC;IACJ,CAAC;IAED,KAAK;QACH,sCAAsC;IACxC,CAAC;CACF"}

package/dist/services/voice/stt-adapter.d.ts

@@ -0,0 +1,31 @@
+/**
+ * Speech-to-Text adapter interface.
+ *
+ * Abstracts the STT provider so the voice pipeline can swap providers
+ * without changing the session manager or segmenter logic.
+ */
+export interface SttTranscript {
+    text: string;
+    /** Confidence score 0-1 if available from provider */
+    confidence?: number;
+    /** Duration of the audio in milliseconds */
+    durationMs?: number;
+}
+export interface SttAdapter {
+    /** Human-readable name of the provider */
+    readonly name: string;
+    /**
+     * Transcribe a complete audio buffer.
+     *
+     * @param audioBuffer - Raw PCM s16le audio at 16kHz mono
+     * @param sampleRate - Sample rate (always 16000 for v1)
+     * @param language - Optional ISO-639-1 language code (e.g. 'en', 'zh') to hint the STT engine
+     * @returns Transcript result
+     */
+    transcribe(audioBuffer: Buffer, sampleRate: number, language?: string): Promise<SttTranscript>;
+    /**
+     * Clean up any resources held by the adapter.
+     */
+    close(): void;
+}
+//# sourceMappingURL=stt-adapter.d.ts.map

package/dist/services/voice/stt-adapter.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"stt-adapter.d.ts","sourceRoot":"","sources":["../../../src/services/voice/stt-adapter.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAEH,MAAM,WAAW,aAAa;IAC5B,IAAI,EAAE,MAAM,CAAC;IACb,sDAAsD;IACtD,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,4CAA4C;IAC5C,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAED,MAAM,WAAW,UAAU;IACzB,0CAA0C;IAC1C,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IAEtB;;;;;;;OAOG;IACH,UAAU,CAAC,WAAW,EAAE,MAAM,EAAE,UAAU,EAAE,MAAM,EAAE,QAAQ,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,aAAa,CAAC,CAAC;IAE/F;;OAEG;IACH,KAAK,IAAI,IAAI,CAAC;CACf"}

package/dist/services/voice/stt-adapter.js

@@ -0,0 +1,8 @@
+/**
+ * Speech-to-Text adapter interface.
+ *
+ * Abstracts the STT provider so the voice pipeline can swap providers
+ * without changing the session manager or segmenter logic.
+ */
+export {};
+//# sourceMappingURL=stt-adapter.js.map

package/dist/services/voice/stt-adapter.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"stt-adapter.js","sourceRoot":"","sources":["../../../src/services/voice/stt-adapter.ts"],"names":[],"mappings":"AAAA;;;;;GAKG"}

package/dist/services/voice/vad-segmenter.d.ts

@@ -0,0 +1,68 @@
+/**
+ * Voice Activity Detection (VAD) and utterance segmenter.
+ *
+ * Uses energy-based detection on raw PCM s16le audio to determine
+ * speech vs silence boundaries. Manages state transitions for
+ * utterance segmentation with configurable thresholds.
+ *
+ * Design decisions:
+ * - Energy-based VAD rather than ML-based for v1 simplicity
+ * - RMS energy with configurable threshold
+ * - Silence timeout finalizes utterances (800ms default)
+ * - Minimum speech duration prevents noise spikes (300ms default)
+ * - Maximum utterance duration caps long segments (15s default)
+ */
+export interface VadConfig {
+    /** RMS energy threshold to consider as speech (0-32767 range). Default: 500 */
+    energyThreshold: number;
+    /** Silence duration (ms) before finalizing utterance. Default: 800 */
+    silenceTimeoutMs: number;
+    /** Minimum speech duration (ms) to count as valid utterance. Default: 300 */
+    minSpeechDurationMs: number;
+    /** Maximum utterance duration (ms). Default: 15000 */
+    maxUtteranceDurationMs: number;
+}
+export declare const DEFAULT_VAD_CONFIG: VadConfig;
+export type VadState = 'LISTENING' | 'IN_SPEECH' | 'WAITING_FOR_FINAL_SILENCE' | 'PROCESSING';
+export interface VadEvent {
+    type: 'speech_start' | 'speech_end' | 'silence_timeout' | 'max_duration';
+    timestamp: number;
+}
+export declare class VadSegmenter {
+    private config;
+    private state;
+    private speechStartTime;
+    private lastVoiceActivityTime;
+    private speechBuffer;
+    private speechBufferBytes;
+    constructor(config?: Partial<VadConfig>);
+    getState(): VadState;
+    /**
+     * Feed a chunk of PCM s16le audio into the segmenter.
+     *
+     * @param chunk - Raw PCM s16le audio buffer at 16kHz
+     * @returns Array of events triggered by this chunk, plus the
+     *          complete utterance buffer if the utterance was finalized.
+     */
+    processChunk(chunk: Buffer): {
+        events: VadEvent[];
+        finalizedUtterance: Buffer | null;
+    };
+    /**
+     * Concatenate the speech buffer and transition to PROCESSING.
+     */
+    private finalizeUtterance;
+    /**
+     * Mark processing as complete and return to listening.
+     */
+    doneProcessing(): void;
+    /**
+     * Reset the segmenter to the initial listening state.
+     */
+    reset(): void;
+    /**
+     * Get the current speech buffer size in bytes (for observability).
+     */
+    getBufferSizeBytes(): number;
+}
+//# sourceMappingURL=vad-segmenter.d.ts.map

package/dist/services/voice/vad-segmenter.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"vad-segmenter.d.ts","sourceRoot":"","sources":["../../../src/services/voice/vad-segmenter.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;GAaG;AAEH,MAAM,WAAW,SAAS;IACxB,+EAA+E;IAC/E,eAAe,EAAE,MAAM,CAAC;IACxB,sEAAsE;IACtE,gBAAgB,EAAE,MAAM,CAAC;IACzB,6EAA6E;IAC7E,mBAAmB,EAAE,MAAM,CAAC;IAC5B,sDAAsD;IACtD,sBAAsB,EAAE,MAAM,CAAC;CAChC;AAED,eAAO,MAAM,kBAAkB,EAAE,SAKhC,CAAC;AAEF,MAAM,MAAM,QAAQ,GAAG,WAAW,GAAG,WAAW,GAAG,2BAA2B,GAAG,YAAY,CAAC;AAE9F,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,cAAc,GAAG,YAAY,GAAG,iBAAiB,GAAG,cAAc,CAAC;IACzE,SAAS,EAAE,MAAM,CAAC;CACnB;AAkBD,qBAAa,YAAY;IACvB,OAAO,CAAC,MAAM,CAAY;IAC1B,OAAO,CAAC,KAAK,CAAyB;IACtC,OAAO,CAAC,eAAe,CAAK;IAC5B,OAAO,CAAC,qBAAqB,CAAK;IAClC,OAAO,CAAC,YAAY,CAAgB;IACpC,OAAO,CAAC,iBAAiB,CAAK;gBAElB,MAAM,GAAE,OAAO,CAAC,SAAS,CAAM;IAI3C,QAAQ,IAAI,QAAQ;IAIpB;;;;;;OAMG;IACH,YAAY,CAAC,KAAK,EAAE,MAAM,GAAG;QAC3B,MAAM,EAAE,QAAQ,EAAE,CAAC;QACnB,kBAAkB,EAAE,MAAM,GAAG,IAAI,CAAC;KACnC;IAkFD;;OAEG;IACH,OAAO,CAAC,iBAAiB;IAQzB;;OAEG;IACH,cAAc,IAAI,IAAI;IAMtB;;OAEG;IACH,KAAK,IAAI,IAAI;IAQb;;OAEG;IACH,kBAAkB,IAAI,MAAM;CAG7B"}

package/dist/services/voice/vad-segmenter.js

@@ -0,0 +1,159 @@
+/**
+ * Voice Activity Detection (VAD) and utterance segmenter.
+ *
+ * Uses energy-based detection on raw PCM s16le audio to determine
+ * speech vs silence boundaries. Manages state transitions for
+ * utterance segmentation with configurable thresholds.
+ *
+ * Design decisions:
+ * - Energy-based VAD rather than ML-based for v1 simplicity
+ * - RMS energy with configurable threshold
+ * - Silence timeout finalizes utterances (800ms default)
+ * - Minimum speech duration prevents noise spikes (300ms default)
+ * - Maximum utterance duration caps long segments (15s default)
+ */
+export const DEFAULT_VAD_CONFIG = {
+    energyThreshold: 500,
+    silenceTimeoutMs: 800,
+    minSpeechDurationMs: 300,
+    maxUtteranceDurationMs: 15000,
+};
+/**
+ * Compute RMS (Root Mean Square) energy of a PCM s16le buffer.
+ */
+function computeRmsEnergy(pcmBuffer) {
+    const sampleCount = pcmBuffer.length / 2;
+    if (sampleCount === 0)
+        return 0;
+    let sumSquares = 0;
+    for (let i = 0; i < pcmBuffer.length; i += 2) {
+        const sample = pcmBuffer.readInt16LE(i);
+        sumSquares += sample * sample;
+    }
+    return Math.sqrt(sumSquares / sampleCount);
+}
+export class VadSegmenter {
+    config;
+    state = 'LISTENING';
+    speechStartTime = 0;
+    lastVoiceActivityTime = 0;
+    speechBuffer = [];
+    speechBufferBytes = 0;
+    constructor(config = {}) {
+        this.config = { ...DEFAULT_VAD_CONFIG, ...config };
+    }
+    getState() {
+        return this.state;
+    }
+    /**
+     * Feed a chunk of PCM s16le audio into the segmenter.
+     *
+     * @param chunk - Raw PCM s16le audio buffer at 16kHz
+     * @returns Array of events triggered by this chunk, plus the
+     *          complete utterance buffer if the utterance was finalized.
+     */
+    processChunk(chunk) {
+        const now = Date.now();
+        const events = [];
+        let finalizedUtterance = null;
+        const energy = computeRmsEnergy(chunk);
+        const isSpeech = energy >= this.config.energyThreshold;
+        switch (this.state) {
+            case 'LISTENING':
+                if (isSpeech) {
+                    this.state = 'IN_SPEECH';
+                    this.speechStartTime = now;
+                    this.lastVoiceActivityTime = now;
+                    this.speechBuffer = [chunk];
+                    this.speechBufferBytes = chunk.length;
+                    events.push({ type: 'speech_start', timestamp: now });
+                }
+                break;
+            case 'IN_SPEECH':
+                this.speechBuffer.push(chunk);
+                this.speechBufferBytes += chunk.length;
+                if (isSpeech) {
+                    this.lastVoiceActivityTime = now;
+                }
+                // Check max duration
+                if (now - this.speechStartTime >= this.config.maxUtteranceDurationMs) {
+                    events.push({ type: 'max_duration', timestamp: now });
+                    finalizedUtterance = this.finalizeUtterance();
+                    break;
+                }
+                // Transition to waiting if we detect silence
+                if (!isSpeech) {
+                    this.state = 'WAITING_FOR_FINAL_SILENCE';
+                }
+                break;
+            case 'WAITING_FOR_FINAL_SILENCE':
+                this.speechBuffer.push(chunk);
+                this.speechBufferBytes += chunk.length;
+                if (isSpeech) {
+                    // Speech resumed — go back to IN_SPEECH
+                    this.state = 'IN_SPEECH';
+                    this.lastVoiceActivityTime = now;
+                    break;
+                }
+                // Check max duration
+                if (now - this.speechStartTime >= this.config.maxUtteranceDurationMs) {
+                    events.push({ type: 'max_duration', timestamp: now });
+                    finalizedUtterance = this.finalizeUtterance();
+                    break;
+                }
+                // Check if silence has lasted long enough
+                if (now - this.lastVoiceActivityTime >= this.config.silenceTimeoutMs) {
+                    const speechDuration = this.lastVoiceActivityTime - this.speechStartTime;
+                    if (speechDuration >= this.config.minSpeechDurationMs) {
+                        events.push({ type: 'silence_timeout', timestamp: now });
+                        finalizedUtterance = this.finalizeUtterance();
+                    }
+                    else {
+                        // Too short to be real speech — discard and go back to listening
+                        this.reset();
+                    }
+                }
+                break;
+            case 'PROCESSING':
+                // While processing, buffer audio for the next utterance
+                // In v1, just drop it and wait for reset
+                break;
+        }
+        return { events, finalizedUtterance };
+    }
+    /**
+     * Concatenate the speech buffer and transition to PROCESSING.
+     */
+    finalizeUtterance() {
+        this.state = 'PROCESSING';
+        const result = Buffer.concat(this.speechBuffer);
+        this.speechBuffer = [];
+        this.speechBufferBytes = 0;
+        return result;
+    }
+    /**
+     * Mark processing as complete and return to listening.
+     */
+    doneProcessing() {
+        this.state = 'LISTENING';
+        this.speechBuffer = [];
+        this.speechBufferBytes = 0;
+    }
+    /**
+     * Reset the segmenter to the initial listening state.
+     */
+    reset() {
+        this.state = 'LISTENING';
+        this.speechStartTime = 0;
+        this.lastVoiceActivityTime = 0;
+        this.speechBuffer = [];
+        this.speechBufferBytes = 0;
+    }
+    /**
+     * Get the current speech buffer size in bytes (for observability).
+     */
+    getBufferSizeBytes() {
+        return this.speechBufferBytes;
+    }
+}
+//# sourceMappingURL=vad-segmenter.js.map

package/dist/services/voice/vad-segmenter.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"vad-segmenter.js","sourceRoot":"","sources":["../../../src/services/voice/vad-segmenter.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;GAaG;AAaH,MAAM,CAAC,MAAM,kBAAkB,GAAc;IAC3C,eAAe,EAAE,GAAG;IACpB,gBAAgB,EAAE,GAAG;IACrB,mBAAmB,EAAE,GAAG;IACxB,sBAAsB,EAAE,KAAK;CAC9B,CAAC;AASF;;GAEG;AACH,SAAS,gBAAgB,CAAC,SAAiB;IACzC,MAAM,WAAW,GAAG,SAAS,CAAC,MAAM,GAAG,CAAC,CAAC;IACzC,IAAI,WAAW,KAAK,CAAC;QAAE,OAAO,CAAC,CAAC;IAEhC,IAAI,UAAU,GAAG,CAAC,CAAC;IACnB,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,SAAS,CAAC,MAAM,EAAE,CAAC,IAAI,CAAC,EAAE,CAAC;QAC7C,MAAM,MAAM,GAAG,SAAS,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC;QACxC,UAAU,IAAI,MAAM,GAAG,MAAM,CAAC;IAChC,CAAC;IAED,OAAO,IAAI,CAAC,IAAI,CAAC,UAAU,GAAG,WAAW,CAAC,CAAC;AAC7C,CAAC;AAED,MAAM,OAAO,YAAY;IACf,MAAM,CAAY;IAClB,KAAK,GAAa,WAAW,CAAC;IAC9B,eAAe,GAAG,CAAC,CAAC;IACpB,qBAAqB,GAAG,CAAC,CAAC;IAC1B,YAAY,GAAa,EAAE,CAAC;IAC5B,iBAAiB,GAAG,CAAC,CAAC;IAE9B,YAAY,SAA6B,EAAE;QACzC,IAAI,CAAC,MAAM,GAAG,EAAE,GAAG,kBAAkB,EAAE,GAAG,MAAM,EAAE,CAAC;IACrD,CAAC;IAED,QAAQ;QACN,OAAO,IAAI,CAAC,KAAK,CAAC;IACpB,CAAC;IAED;;;;;;OAMG;IACH,YAAY,CAAC,KAAa;QAIxB,MAAM,GAAG,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QACvB,MAAM,MAAM,GAAe,EAAE,CAAC;QAC9B,IAAI,kBAAkB,GAAkB,IAAI,CAAC;QAE7C,MAAM,MAAM,GAAG,gBAAgB,CAAC,KAAK,CAAC,CAAC;QACvC,MAAM,QAAQ,GAAG,MAAM,IAAI,IAAI,CAAC,MAAM,CAAC,eAAe,CAAC;QAEvD,QAAQ,IAAI,CAAC,KAAK,EAAE,CAAC;YACnB,KAAK,WAAW;gBACd,IAAI,QAAQ,EAAE,CAAC;oBACb,IAAI,CAAC,KAAK,GAAG,WAAW,CAAC;oBACzB,IAAI,CAAC,eAAe,GAAG,GAAG,CAAC;oBAC3B,IAAI,CAAC,qBAAqB,GAAG,GAAG,CAAC;oBACjC,IAAI,CAAC,YAAY,GAAG,CAAC,KAAK,CAAC,CAAC;oBAC5B,IAAI,CAAC,iBAAiB,GAAG,KAAK,CAAC,MAAM,CAAC;oBACtC,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,cAAc,EAAE,SAAS,EAAE,GAAG,EAAE,CAAC,CAAC;gBACxD,CAAC;gBACD,MAAM;YAER,KAAK,WAAW;gBACd,IAAI,CAAC,YAAY,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;gBAC9B,IAAI,CAAC,iBAAiB,IAAI,KAAK,CAAC,MAAM,CAAC;gBAEvC,IAAI,QAAQ,EAAE,CAAC;oBACb,IAAI,CAAC,qBAAqB,GAAG,GAAG,CAAC;gBACnC,CAAC;gBAED,qBAAqB;gBACrB,IAAI,GAAG,GAAG,IAAI,CAAC,eAAe,IAAI,IAAI,CAAC,MAAM,CAAC,sBAAsB,EAAE,CAAC;oBACrE,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,cAAc,EAAE,SAAS,EAAE,GAAG,EAAE,CAAC,CAAC;oBACtD,kBAAkB,GAAG,IAAI,CAAC,iBAAiB,EAAE,CAAC;oBAC9C,MAAM;gBACR,CAAC;gBAED,6CAA6C;gBAC7C,IAAI,CAAC,QAAQ,EAAE,CAAC;oBACd,IAAI,CAAC,KAAK,GAAG,2BAA2B,CAAC;gBAC3C,CAAC;gBACD,MAAM;YAER,KAAK,2BAA2B;gBAC9B,IAAI,CAAC,YAAY,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;gBAC9B,IAAI,CAAC,iBAAiB,IAAI,KAAK,CAAC,MAAM,CAAC;gBAEvC,IAAI,QAAQ,EAAE,CAAC;oBACb,wCAAwC;oBACxC,IAAI,CAAC,KAAK,GAAG,WAAW,CAAC;oBACzB,IAAI,CAAC,qBAAqB,GAAG,GAAG,CAAC;oBACjC,MAAM;gBACR,CAAC;gBAED,qBAAqB;gBACrB,IAAI,GAAG,GAAG,IAAI,CAAC,eAAe,IAAI,IAAI,CAAC,MAAM,CAAC,sBAAsB,EAAE,CAAC;oBACrE,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,cAAc,EAAE,SAAS,EAAE,GAAG,EAAE,CAAC,CAAC;oBACtD,kBAAkB,GAAG,IAAI,CAAC,iBAAiB,EAAE,CAAC;oBAC9C,MAAM;gBACR,CAAC;gBAED,0CAA0C;gBAC1C,IAAI,GAAG,GAAG,IAAI,CAAC,qBAAqB,IAAI,IAAI,CAAC,MAAM,CAAC,gBAAgB,EAAE,CAAC;oBACrE,MAAM,cAAc,GAAG,IAAI,CAAC,qBAAqB,GAAG,IAAI,CAAC,eAAe,CAAC;oBAEzE,IAAI,cAAc,IAAI,IAAI,CAAC,MAAM,CAAC,mBAAmB,EAAE,CAAC;wBACtD,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,iBAAiB,EAAE,SAAS,EAAE,GAAG,EAAE,CAAC,CAAC;wBACzD,kBAAkB,GAAG,IAAI,CAAC,iBAAiB,EAAE,CAAC;oBAChD,CAAC;yBAAM,CAAC;wBACN,iEAAiE;wBACjE,IAAI,CAAC,KAAK,EAAE,CAAC;oBACf,CAAC;gBACH,CAAC;gBACD,MAAM;YAER,KAAK,YAAY;gBACf,wDAAwD;gBACxD,yCAAyC;gBACzC,MAAM;QACV,CAAC;QAED,OAAO,EAAE,MAAM,EAAE,kBAAkB,EAAE,CAAC;IACxC,CAAC;IAED;;OAEG;IACK,iBAAiB;QACvB,IAAI,CAAC,KAAK,GAAG,YAAY,CAAC;QAC1B,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,YAAY,CAAC,CAAC;QAChD,IAAI,CAAC,YAAY,GAAG,EAAE,CAAC;QACvB,IAAI,CAAC,iBAAiB,GAAG,CAAC,CAAC;QAC3B,OAAO,MAAM,CAAC;IAChB,CAAC;IAED;;OAEG;IACH,cAAc;QACZ,IAAI,CAAC,KAAK,GAAG,WAAW,CAAC;QACzB,IAAI,CAAC,YAAY,GAAG,EAAE,CAAC;QACvB,IAAI,CAAC,iBAAiB,GAAG,CAAC,CAAC;IAC7B,CAAC;IAED;;OAEG;IACH,KAAK;QACH,IAAI,CAAC,KAAK,GAAG,WAAW,CAAC;QACzB,IAAI,CAAC,eAAe,GAAG,CAAC,CAAC;QACzB,IAAI,CAAC,qBAAqB,GAAG,CAAC,CAAC;QAC/B,IAAI,CAAC,YAAY,GAAG,EAAE,CAAC;QACvB,IAAI,CAAC,iBAAiB,GAAG,CAAC,CAAC;IAC7B,CAAC;IAED;;OAEG;IACH,kBAAkB;QAChB,OAAO,IAAI,CAAC,iBAAiB,CAAC;IAChC,CAAC;CACF"}

package/dist/services/voice/voice-session.d.ts

@@ -0,0 +1,54 @@
+/**
+ * Voice session manager.
+ *
+ * Manages per-session state: audio ingestion, VAD segmentation,
+ * STT transcription, and event publishing back to the client.
+ *
+ * Each voice WebSocket connection gets one VoiceSession instance.
+ */
+import type { WebSocket } from 'ws';
+import type pino from 'pino';
+import { type VadConfig } from './vad-segmenter.js';
+import type { SttAdapter } from './stt-adapter.js';
+export type SessionState = 'LISTENING' | 'IN_SPEECH' | 'WAITING_FOR_FINAL_SILENCE' | 'PROCESSING' | 'CLOSED';
+export interface VoiceSessionConfig {
+    sampleRate: number;
+    channels: number;
+    format: string;
+    vad?: Partial<VadConfig>;
+    /** ISO-639-1 language code (e.g. 'en', 'zh') to hint the STT engine */
+    language?: string;
+}
+export interface TranscriptEntry {
+    utteranceId: string;
+    text: string;
+    timestamp: number;
+    durationMs?: number;
+}
+export declare class VoiceSession {
+    readonly sessionId: string;
+    private ws;
+    private sttAdapter;
+    private segmenter;
+    private config;
+    private state;
+    private transcripts;
+    private chunkCount;
+    private createdAt;
+    private currentUtteranceId;
+    private logger;
+    constructor(sessionId: string, ws: WebSocket, sttAdapter: SttAdapter, config: VoiceSessionConfig, logger: pino.Logger);
+    /**
+     * Process an incoming binary audio chunk from the client.
+     */
+    processAudioChunk(chunk: Buffer): Promise<void>;
+    /**
+     * Close the session and clean up resources.
+     */
+    close(): void;
+    getState(): SessionState;
+    getTranscripts(): TranscriptEntry[];
+    getChunkCount(): number;
+    private sendEvent;
+}
+//# sourceMappingURL=voice-session.d.ts.map

package/dist/services/voice/voice-session.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"voice-session.d.ts","sourceRoot":"","sources":["../../../src/services/voice/voice-session.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAEH,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,IAAI,CAAC;AACpC,OAAO,KAAK,IAAI,MAAM,MAAM,CAAC;AAC7B,OAAO,EAAgB,KAAK,SAAS,EAAE,MAAM,iBAAiB,CAAC;AAC/D,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,eAAe,CAAC;AAEhD,MAAM,MAAM,YAAY,GACpB,WAAW,GACX,WAAW,GACX,2BAA2B,GAC3B,YAAY,GACZ,QAAQ,CAAC;AAEb,MAAM,WAAW,kBAAkB;IACjC,UAAU,EAAE,MAAM,CAAC;IACnB,QAAQ,EAAE,MAAM,CAAC;IACjB,MAAM,EAAE,MAAM,CAAC;IACf,GAAG,CAAC,EAAE,OAAO,CAAC,SAAS,CAAC,CAAC;IACzB,uEAAuE;IACvE,QAAQ,CAAC,EAAE,MAAM,CAAC;CACnB;AAED,MAAM,WAAW,eAAe;IAC9B,WAAW,EAAE,MAAM,CAAC;IACpB,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAcD,qBAAa,YAAY;IACvB,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAC3B,OAAO,CAAC,EAAE,CAAY;IACtB,OAAO,CAAC,UAAU,CAAa;IAC/B,OAAO,CAAC,SAAS,CAAe;IAChC,OAAO,CAAC,MAAM,CAAqB;IACnC,OAAO,CAAC,KAAK,CAA6B;IAC1C,OAAO,CAAC,WAAW,CAAyB;IAC5C,OAAO,CAAC,UAAU,CAAK;IACvB,OAAO,CAAC,SAAS,CAAS;IAC1B,OAAO,CAAC,kBAAkB,CAAuB;IACjD,OAAO,CAAC,MAAM,CAAc;gBAG1B,SAAS,EAAE,MAAM,EACjB,EAAE,EAAE,SAAS,EACb,UAAU,EAAE,UAAU,EACtB,MAAM,EAAE,kBAAkB,EAC1B,MAAM,EAAE,IAAI,CAAC,MAAM;IAWrB;;OAEG;IACG,iBAAiB,CAAC,KAAK,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAoFrD;;OAEG;IACH,KAAK,IAAI,IAAI;IAab,QAAQ,IAAI,YAAY;IAIxB,cAAc,IAAI,eAAe,EAAE;IAInC,aAAa,IAAI,MAAM;IAIvB,OAAO,CAAC,SAAS;CAKlB"}

package/dist/services/voice/voice-session.js

@@ -0,0 +1,137 @@
+/**
+ * Voice session manager.
+ *
+ * Manages per-session state: audio ingestion, VAD segmentation,
+ * STT transcription, and event publishing back to the client.
+ *
+ * Each voice WebSocket connection gets one VoiceSession instance.
+ */
+import { VadSegmenter } from './vad-segmenter.js';
+let utteranceCounter = 0;
+function nextUtteranceId() {
+    return `u${++utteranceCounter}`;
+}
+export class VoiceSession {
+    sessionId;
+    ws;
+    sttAdapter;
+    segmenter;
+    config;
+    state = 'LISTENING';
+    transcripts = [];
+    chunkCount = 0;
+    createdAt;
+    currentUtteranceId = null;
+    logger;
+    constructor(sessionId, ws, sttAdapter, config, logger) {
+        this.sessionId = sessionId;
+        this.ws = ws;
+        this.sttAdapter = sttAdapter;
+        this.config = config;
+        this.segmenter = new VadSegmenter(config.vad);
+        this.createdAt = Date.now();
+        this.logger = logger;
+    }
+    /**
+     * Process an incoming binary audio chunk from the client.
+     */
+    async processAudioChunk(chunk) {
+        if (this.state === 'CLOSED')
+            return;
+        this.chunkCount++;
+        const { events, finalizedUtterance } = this.segmenter.processChunk(chunk);
+        // Update our state to mirror the segmenter
+        const segmenterState = this.segmenter.getState();
+        if (segmenterState !== 'PROCESSING') {
+            this.state = segmenterState;
+        }
+        // Emit VAD events to the client
+        for (const event of events) {
+            switch (event.type) {
+                case 'speech_start':
+                    this.currentUtteranceId = nextUtteranceId();
+                    this.sendEvent({
+                        type: 'speech.start',
+                        sessionId: this.sessionId,
+                        utteranceId: this.currentUtteranceId,
+                        timestamp: event.timestamp,
+                    });
+                    break;
+                case 'silence_timeout':
+                case 'max_duration':
+                    // Speech ended — utterance will be finalized
+                    break;
+            }
+        }
+        // If the segmenter finalized an utterance, transcribe it
+        if (finalizedUtterance) {
+            this.state = 'PROCESSING';
+            const utteranceId = this.currentUtteranceId ?? nextUtteranceId();
+            try {
+                const transcript = await this.sttAdapter.transcribe(finalizedUtterance, this.config.sampleRate, this.config.language);
+                const entry = {
+                    utteranceId,
+                    text: transcript.text,
+                    timestamp: Date.now(),
+                    durationMs: transcript.durationMs,
+                };
+                this.transcripts.push(entry);
+                // Send final transcript
+                this.sendEvent({
+                    type: 'transcript.final',
+                    sessionId: this.sessionId,
+                    utteranceId,
+                    text: transcript.text,
+                    durationMs: transcript.durationMs,
+                });
+                // Send speech end
+                this.sendEvent({
+                    type: 'speech.end',
+                    sessionId: this.sessionId,
+                    utteranceId,
+                    timestamp: Date.now(),
+                });
+            }
+            catch (err) {
+                this.sendEvent({
+                    type: 'error',
+                    sessionId: this.sessionId,
+                    code: 'PROVIDER_ERROR',
+                    message: err.message,
+                });
+            }
+            finally {
+                this.segmenter.doneProcessing();
+                this.state = 'LISTENING';
+                this.currentUtteranceId = null;
+            }
+        }
+    }
+    /**
+     * Close the session and clean up resources.
+     */
+    close() {
+        this.state = 'CLOSED';
+        this.segmenter.reset();
+        this.logger.info({
+            chunks: this.chunkCount,
+            transcripts: this.transcripts.length,
+            durationMs: Date.now() - this.createdAt,
+        }, 'voice session closed');
+    }
+    getState() {
+        return this.state;
+    }
+    getTranscripts() {
+        return [...this.transcripts];
+    }
+    getChunkCount() {
+        return this.chunkCount;
+    }
+    sendEvent(event) {
+        if (this.ws.readyState === 1 /* OPEN */) {
+            this.ws.send(JSON.stringify(event));
+        }
+    }
+}
+//# sourceMappingURL=voice-session.js.map

package/dist/services/voice/voice-session.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"voice-session.js","sourceRoot":"","sources":["../../../src/services/voice/voice-session.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAIH,OAAO,EAAE,YAAY,EAAkB,MAAM,iBAAiB,CAAC;AAgC/D,IAAI,gBAAgB,GAAG,CAAC,CAAC;AAEzB,SAAS,eAAe;IACtB,OAAO,IAAI,EAAE,gBAAgB,EAAE,CAAC;AAClC,CAAC;AAED,MAAM,OAAO,YAAY;IACd,SAAS,CAAS;IACnB,EAAE,CAAY;IACd,UAAU,CAAa;IACvB,SAAS,CAAe;IACxB,MAAM,CAAqB;IAC3B,KAAK,GAAiB,WAAW,CAAC;IAClC,WAAW,GAAsB,EAAE,CAAC;IACpC,UAAU,GAAG,CAAC,CAAC;IACf,SAAS,CAAS;IAClB,kBAAkB,GAAkB,IAAI,CAAC;IACzC,MAAM,CAAc;IAE5B,YACE,SAAiB,EACjB,EAAa,EACb,UAAsB,EACtB,MAA0B,EAC1B,MAAmB;QAEnB,IAAI,CAAC,SAAS,GAAG,SAAS,CAAC;QAC3B,IAAI,CAAC,EAAE,GAAG,EAAE,CAAC;QACb,IAAI,CAAC,UAAU,GAAG,UAAU,CAAC;QAC7B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;QACrB,IAAI,CAAC,SAAS,GAAG,IAAI,YAAY,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;QAC9C,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAC5B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;IACvB,CAAC;IAED;;OAEG;IACH,KAAK,CAAC,iBAAiB,CAAC,KAAa;QACnC,IAAI,IAAI,CAAC,KAAK,KAAK,QAAQ;YAAE,OAAO;QAEpC,IAAI,CAAC,UAAU,EAAE,CAAC;QAElB,MAAM,EAAE,MAAM,EAAE,kBAAkB,EAAE,GAAG,IAAI,CAAC,SAAS,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC;QAE1E,2CAA2C;QAC3C,MAAM,cAAc,GAAG,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,CAAC;QACjD,IAAI,cAAc,KAAK,YAAY,EAAE,CAAC;YACpC,IAAI,CAAC,KAAK,GAAG,cAAc,CAAC;QAC9B,CAAC;QAED,gCAAgC;QAChC,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;YAC3B,QAAQ,KAAK,CAAC,IAAI,EAAE,CAAC;gBACnB,KAAK,cAAc;oBACjB,IAAI,CAAC,kBAAkB,GAAG,eAAe,EAAE,CAAC;oBAC5C,IAAI,CAAC,SAAS,CAAC;wBACb,IAAI,EAAE,cAAc;wBACpB,SAAS,EAAE,IAAI,CAAC,SAAS;wBACzB,WAAW,EAAE,IAAI,CAAC,kBAAkB;wBACpC,SAAS,EAAE,KAAK,CAAC,SAAS;qBAC3B,CAAC,CAAC;oBACH,MAAM;gBAER,KAAK,iBAAiB,CAAC;gBACvB,KAAK,cAAc;oBACjB,6CAA6C;oBAC7C,MAAM;YACV,CAAC;QACH,CAAC;QAED,yDAAyD;QACzD,IAAI,kBAAkB,EAAE,CAAC;YACvB,IAAI,CAAC,KAAK,GAAG,YAAY,CAAC;YAC1B,MAAM,WAAW,GAAG,IAAI,CAAC,kBAAkB,IAAI,eAAe,EAAE,CAAC;YAEjE,IAAI,CAAC;gBACH,MAAM,UAAU,GAAG,MAAM,IAAI,CAAC,UAAU,CAAC,UAAU,CACjD,kBAAkB,EAClB,IAAI,CAAC,MAAM,CAAC,UAAU,EACtB,IAAI,CAAC,MAAM,CAAC,QAAQ,CACrB,CAAC;gBAEF,MAAM,KAAK,GAAoB;oBAC7B,WAAW;oBACX,IAAI,EAAE,UAAU,CAAC,IAAI;oBACrB,SAAS,EAAE,IAAI,CAAC,GAAG,EAAE;oBACrB,UAAU,EAAE,UAAU,CAAC,UAAU;iBAClC,CAAC;gBACF,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;gBAE7B,wBAAwB;gBACxB,IAAI,CAAC,SAAS,CAAC;oBACb,IAAI,EAAE,kBAAkB;oBACxB,SAAS,EAAE,IAAI,CAAC,SAAS;oBACzB,WAAW;oBACX,IAAI,EAAE,UAAU,CAAC,IAAI;oBACrB,UAAU,EAAE,UAAU,CAAC,UAAU;iBAClC,CAAC,CAAC;gBAEH,kBAAkB;gBAClB,IAAI,CAAC,SAAS,CAAC;oBACb,IAAI,EAAE,YAAY;oBAClB,SAAS,EAAE,IAAI,CAAC,SAAS;oBACzB,WAAW;oBACX,SAAS,EAAE,IAAI,CAAC,GAAG,EAAE;iBACtB,CAAC,CAAC;YACL,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACb,IAAI,CAAC,SAAS,CAAC;oBACb,IAAI,EAAE,OAAO;oBACb,SAAS,EAAE,IAAI,CAAC,SAAS;oBACzB,IAAI,EAAE,gBAAgB;oBACtB,OAAO,EAAG,GAAa,CAAC,OAAO;iBAChC,CAAC,CAAC;YACL,CAAC;oBAAS,CAAC;gBACT,IAAI,CAAC,SAAS,CAAC,cAAc,EAAE,CAAC;gBAChC,IAAI,CAAC,KAAK,GAAG,WAAW,CAAC;gBACzB,IAAI,CAAC,kBAAkB,GAAG,IAAI,CAAC;YACjC,CAAC;QACH,CAAC;IACH,CAAC;IAED;;OAEG;IACH,KAAK;QACH,IAAI,CAAC,KAAK,GAAG,QAAQ,CAAC;QACtB,IAAI,CAAC,SAAS,CAAC,KAAK,EAAE,CAAC;QACvB,IAAI,CAAC,MAAM,CAAC,IAAI,CACd;YACE,MAAM,EAAE,IAAI,CAAC,UAAU;YACvB,WAAW,EAAE,IAAI,CAAC,WAAW,CAAC,MAAM;YACpC,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,IAAI,CAAC,SAAS;SACxC,EACD,sBAAsB,CACvB,CAAC;IACJ,CAAC;IAED,QAAQ;QACN,OAAO,IAAI,CAAC,KAAK,CAAC;IACpB,CAAC;IAED,cAAc;QACZ,OAAO,CAAC,GAAG,IAAI,CAAC,WAAW,CAAC,CAAC;IAC/B,CAAC;IAED,aAAa;QACX,OAAO,IAAI,CAAC,UAAU,CAAC;IACzB,CAAC;IAEO,SAAS,CAAC,KAAkB;QAClC,IAAI,IAAI,CAAC,EAAE,CAAC,UAAU,KAAK,CAAC,CAAC,UAAU,EAAE,CAAC;YACxC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,CAAC;QACtC,CAAC;IACH,CAAC;CACF"}

package/dist/services/window-manager.d.ts

@@ -0,0 +1,94 @@
+export interface SerializableActionDefinition {
+    name: string;
+    description: string;
+    params?: Record<string, {
+        type: 'string' | 'number' | 'boolean';
+        description?: string;
+        required?: boolean;
+    }>;
+}
+/**
+ * Tiling tree node types (mirrored from frontend tiling-tree.ts for persistence).
+ */
+export interface LeafNode {
+    type: 'leaf';
+    windowId: string;
+}
+export interface ContainerNode {
+    type: 'container';
+    split: 'horizontal' | 'vertical';
+    ratio: number;
+    children: [TilingNode, TilingNode];
+}
+export type TilingNode = LeafNode | ContainerNode;
+export interface PersistedWindow {
+    id: string;
+    miniAppId: string;
+    title: string;
+    args?: Record<string, unknown>;
+}
+export interface PersistedWindowState {
+    version: 2;
+    windows: PersistedWindow[];
+    focusedWindowId: string | null;
+    fullscreenWindowId: string | null;
+    windowIdCounter: number;
+    nextSplitDirection: 'horizontal' | 'vertical' | 'auto';
+    tree: TilingNode | null;
+}
+/**
+ * Backend window manager service — persistence-only.
+ *
+ * The frontend Zustand store is the live source of truth.
+ * This service:
+ *  - Loads persisted state on startup for the initial `window:state` message
+ *  - Receives synced state from the frontend and persists it
+ *  - Keeps window actions metadata for the AI system prompt
+ *  - Provides `getSystemPromptContext()` for dynamic AI prompt injection
+ */
+export declare class WindowManagerService {
+    private state;
+    private readonly windowActions;
+    private filePath;
+    constructor(filePath: string);
+    /**
+     * Switch the backing file path (e.g. when a different user connects)
+     * and reload persisted state from the new location.
+     */
+    switchUser(newFilePath: string): void;
+    private load;
+    /**
+     * Get the persisted state to send to the frontend on initial connect.
+     */
+    getPersistedState(): PersistedWindowState;
+    /**
+     * Receive the full state from the frontend and persist it.
+     * Called when the frontend sends `window:sync`.
+     */
+    syncState(payload: PersistedWindowState): void;
+    /**
+     * Get the current windows (from last synced state).
+     */
+    getWindows(): PersistedWindow[];
+    getFocusedWindow(): PersistedWindow | undefined;
+    getWindowActions(windowId: string): SerializableActionDefinition[];
+    setWindowActions(windowId: string, actions: SerializableActionDefinition[]): void;
+    /**
+     * Activate persisted MiniApps on startup.
+     */
+    activatePersistedMiniApps(activate: (miniAppId: string, launchArgs: Array<Record<string, unknown>>) => Promise<void>): Promise<void>;
+    /**
+     * Build a dynamic desktop context block to prepend to user messages.
+     *
+     * This is injected per-prompt so the AI always sees the latest state
+     * without polluting the cacheable system prompt.
+     */
+    getDesktopContext(availableMiniApps: Array<{
+        id: string;
+        name: string;
+    }>, availableLiveApps: Array<{
+        id: string;
+        name: string;
+    }>, userHomeDir: string): string;
+}
+//# sourceMappingURL=window-manager.d.ts.map

package/dist/services/window-manager.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"window-manager.d.ts","sourceRoot":"","sources":["../../src/services/window-manager.ts"],"names":[],"mappings":"AAGA,MAAM,WAAW,4BAA4B;IAC3C,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,MAAM,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CACb,MAAM,EACN;QACE,IAAI,EAAE,QAAQ,GAAG,QAAQ,GAAG,SAAS,CAAC;QACtC,WAAW,CAAC,EAAE,MAAM,CAAC;QACrB,QAAQ,CAAC,EAAE,OAAO,CAAC;KACpB,CACF,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,MAAM,CAAC;IACb,QAAQ,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,aAAa;IAC5B,IAAI,EAAE,WAAW,CAAC;IAClB,KAAK,EAAE,YAAY,GAAG,UAAU,CAAC;IACjC,KAAK,EAAE,MAAM,CAAC;IACd,QAAQ,EAAE,CAAC,UAAU,EAAE,UAAU,CAAC,CAAC;CACpC;AAED,MAAM,MAAM,UAAU,GAAG,QAAQ,GAAG,aAAa,CAAC;AAElD,MAAM,WAAW,eAAe;IAC9B,EAAE,EAAE,MAAM,CAAC;IACX,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,EAAE,MAAM,CAAC;IACd,IAAI,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CAChC;AAED,MAAM,WAAW,oBAAoB;IACnC,OAAO,EAAE,CAAC,CAAC;IACX,OAAO,EAAE,eAAe,EAAE,CAAC;IAC3B,eAAe,EAAE,MAAM,GAAG,IAAI,CAAC;IAC/B,kBAAkB,EAAE,MAAM,GAAG,IAAI,CAAC;IAClC,eAAe,EAAE,MAAM,CAAC;IACxB,kBAAkB,EAAE,YAAY,GAAG,UAAU,GAAG,MAAM,CAAC;IACvD,IAAI,EAAE,UAAU,GAAG,IAAI,CAAC;CACzB;AAkJD;;;;;;;;;GASG;AACH,qBAAa,oBAAoB;IAC/B,OAAO,CAAC,KAAK,CAA+C;IAC5D,OAAO,CAAC,QAAQ,CAAC,aAAa,CAAsD;IACpF,OAAO,CAAC,QAAQ,CAAS;gBAEb,QAAQ,EAAE,MAAM;IAK5B;;;OAGG;IACH,UAAU,CAAC,WAAW,EAAE,MAAM,GAAG,IAAI;IAMrC,OAAO,CAAC,IAAI;IAuCZ;;OAEG;IACH,iBAAiB,IAAI,oBAAoB;IAIzC;;;OAGG;IACH,SAAS,CAAC,OAAO,EAAE,oBAAoB,GAAG,IAAI;IAmB9C;;OAEG;IACH,UAAU,IAAI,eAAe,EAAE;IAI/B,gBAAgB,IAAI,eAAe,GAAG,SAAS;IAI/C,gBAAgB,CAAC,QAAQ,EAAE,MAAM,GAAG,4BAA4B,EAAE;IAIlE,gBAAgB,CAAC,QAAQ,EAAE,MAAM,EAAE,OAAO,EAAE,4BAA4B,EAAE,GAAG,IAAI;IAIjF;;OAEG;IACG,yBAAyB,CAC7B,QAAQ,EAAE,CAAC,SAAS,EAAE,MAAM,EAAE,UAAU,EAAE,KAAK,CAAC,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,KAAK,OAAO,CAAC,IAAI,CAAC,GACzF,OAAO,CAAC,IAAI,CAAC;IAchB;;;;;OAKG;IACH,iBAAiB,CACf,iBAAiB,EAAE,KAAK,CAAC;QAAE,EAAE,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC,EACtD,iBAAiB,EAAE,KAAK,CAAC;QAAE,EAAE,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC,EACtD,WAAW,EAAE,MAAM,GAClB,MAAM;CA8DV"}