@glassly/engine 0.1.0-dev.0

package/build/services/soniox/SonioxTokenAssembler.js

@@ -0,0 +1,394 @@
+/**
+ * SonioxTokenAssembler
+ *
+ * Pure port of the cloud Soniox provider's utterance assembly
+ * (cloud-v2/packages/runtime/src/services/audio/providers/soniox.ts). Soniox
+ * emits a rolling token window that is re-finalized/compacted across results;
+ * this class turns that stream into stable interim/final utterances:
+ *
+ * - finalized text accumulates into a stable prefix so interims never shrink
+ * - a single FINAL per utterance, committed at the endpoint boundary
+ * - the endpoint is a soft signal: the candidate final is held briefly
+ *   (debounce) and merged back if the next rolling window overlaps
+ * - speaker/language are tracked per token but never split an utterance
+ *   (Soniox re-diarizes its window; per-token labels flicker)
+ * - translation sessions split the window: translation-status tokens are the
+ *   emitted text, original-status tokens ride along as `originalText`, and
+ *   untagged tokens are Soniox's same-language passthrough
+ *
+ * Timers are injected for testability; production uses BgTimer so the debounce
+ * survives app backgrounding.
+ */
+import { BgTimer } from "../../utils/timers";
+const MIN_ENDPOINT_OVERLAP_CHARS = 3;
+const MIN_REGRESSION_DROP_CHARS = 12;
+const MIN_REGRESSION_PREFIX_CHARS = 12;
+const DEFAULT_ENDPOINT_DEBOUNCE_MS = 500;
+export class SonioxTokenAssembler {
+    opts;
+    endpointDebounceMs;
+    setTimeoutFn;
+    clearTimeoutFn;
+    // === Per-utterance state ===
+    stablePrefix = "";
+    prevFinalLen = 0;
+    lastSentInterim = "";
+    // Translation subs only: parallel accumulator for original-language tokens.
+    originalStablePrefix = "";
+    originalPrevFinalLen = 0;
+    lastOriginalText = "";
+    currentSourceLanguage = undefined;
+    currentUtteranceId = null;
+    currentSpeakerId = undefined;
+    currentLanguage = undefined;
+    // Deferred final (endpoint debounce).
+    pendingFinalText = "";
+    pendingFinalUtteranceId = null;
+    pendingFinalSpeakerId = undefined;
+    pendingFinalLanguage = undefined;
+    pendingFinalOriginalText = "";
+    pendingFinalSourceLanguage = undefined;
+    pendingFinalStartMs = undefined;
+    pendingFinalEndMs = undefined;
+    pendingFinalTimer = null;
+    // Id minting: an instance nonce prevents collisions across service restarts.
+    utteranceSeq = 0;
+    instanceId = Math.random().toString(36).slice(2, 10);
+    constructor(opts) {
+        this.opts = opts;
+        this.endpointDebounceMs = opts.endpointDebounceMs ?? DEFAULT_ENDPOINT_DEBOUNCE_MS;
+        this.setTimeoutFn = opts.setTimeoutFn ?? ((cb, ms) => BgTimer.setTimeout(cb, ms));
+        this.clearTimeoutFn = opts.clearTimeoutFn ?? ((id) => BgTimer.clearTimeout(id));
+    }
+    mintUtteranceId() {
+        this.utteranceSeq += 1;
+        return `utt_${this.opts.scope}_${this.instanceId}_${this.utteranceSeq}`;
+    }
+    startNewUtterance(speakerId, language) {
+        this.currentUtteranceId = this.mintUtteranceId();
+        this.currentSpeakerId = speakerId;
+        this.currentLanguage = language;
+        this.stablePrefix = "";
+        this.prevFinalLen = 0;
+        this.lastSentInterim = "";
+        this.originalStablePrefix = "";
+        this.originalPrevFinalLen = 0;
+        this.lastOriginalText = "";
+        this.currentSourceLanguage = undefined;
+    }
+    resetActiveUtterance() {
+        this.stablePrefix = "";
+        this.prevFinalLen = 0;
+        this.lastSentInterim = "";
+        this.originalStablePrefix = "";
+        this.originalPrevFinalLen = 0;
+        this.lastOriginalText = "";
+        this.currentSourceLanguage = undefined;
+        this.currentUtteranceId = null;
+        this.currentSpeakerId = undefined;
+        this.currentLanguage = undefined;
+    }
+    clearPendingFinal() {
+        if (this.pendingFinalTimer !== null) {
+            this.clearTimeoutFn(this.pendingFinalTimer);
+            this.pendingFinalTimer = null;
+        }
+        this.pendingFinalText = "";
+        this.pendingFinalUtteranceId = null;
+        this.pendingFinalSpeakerId = undefined;
+        this.pendingFinalLanguage = undefined;
+        this.pendingFinalOriginalText = "";
+        this.pendingFinalSourceLanguage = undefined;
+        this.pendingFinalStartMs = undefined;
+        this.pendingFinalEndMs = undefined;
+    }
+    tryMergeOverlap(a, b) {
+        const left = a.trim();
+        const right = b.trim();
+        if (!left || !right)
+            return null;
+        if (left === right || left.startsWith(right))
+            return left;
+        if (right.startsWith(left))
+            return left + right.slice(left.length);
+        const maxOverlap = Math.min(left.length, right.length);
+        for (let len = maxOverlap; len >= MIN_ENDPOINT_OVERLAP_CHARS; len -= 1) {
+            if (left.endsWith(right.slice(0, len))) {
+                return left + right.slice(len);
+            }
+        }
+        return null;
+    }
+    isLikelyWindowRegression(next, previous) {
+        const normalizedNext = next.trim();
+        const normalizedPrevious = previous.trim();
+        if (!normalizedNext || !normalizedPrevious)
+            return false;
+        const drop = normalizedPrevious.length - normalizedNext.length;
+        if (drop < MIN_REGRESSION_DROP_CHARS)
+            return false;
+        let sharedPrefix = 0;
+        const maxPrefix = Math.min(normalizedNext.length, normalizedPrevious.length);
+        while (sharedPrefix < maxPrefix && normalizedNext[sharedPrefix] === normalizedPrevious[sharedPrefix]) {
+            sharedPrefix += 1;
+        }
+        return sharedPrefix >= Math.min(MIN_REGRESSION_PREFIX_CHARS, normalizedNext.length);
+    }
+    emitFinalEvent(event) {
+        this.opts.onTranscript({
+            text: event.text,
+            isFinal: true,
+            utteranceId: event.utteranceId,
+            speakerId: event.speakerId,
+            language: event.language,
+            originalText: event.originalText || undefined,
+            sourceLanguage: event.sourceLanguage,
+            startMs: event.startMs,
+            endMs: event.endMs,
+        });
+    }
+    commitPendingFinal() {
+        if (!this.pendingFinalText) {
+            this.clearPendingFinal();
+            return;
+        }
+        const committedText = this.pendingFinalText;
+        const committedUtteranceId = this.pendingFinalUtteranceId;
+        this.emitFinalEvent({
+            text: committedText,
+            utteranceId: committedUtteranceId ?? undefined,
+            speakerId: this.pendingFinalSpeakerId,
+            language: this.pendingFinalLanguage,
+            originalText: this.pendingFinalOriginalText,
+            sourceLanguage: this.pendingFinalSourceLanguage,
+            startMs: this.pendingFinalStartMs,
+            endMs: this.pendingFinalEndMs,
+        });
+        if (committedUtteranceId &&
+            this.currentUtteranceId === committedUtteranceId &&
+            this.lastSentInterim.trim() === committedText) {
+            this.resetActiveUtterance();
+        }
+        this.clearPendingFinal();
+    }
+    schedulePendingFinalCommit() {
+        if (this.pendingFinalTimer !== null)
+            this.clearTimeoutFn(this.pendingFinalTimer);
+        this.pendingFinalTimer = this.setTimeoutFn(() => {
+            this.commitPendingFinal();
+        }, this.endpointDebounceMs);
+    }
+    setPendingFinalFromActive(text) {
+        this.pendingFinalText = text;
+        this.pendingFinalUtteranceId = this.currentUtteranceId;
+        this.pendingFinalSpeakerId = this.currentSpeakerId;
+        this.pendingFinalLanguage = this.currentLanguage;
+        this.pendingFinalOriginalText = this.lastOriginalText;
+        this.pendingFinalSourceLanguage = this.currentSourceLanguage;
+    }
+    /**
+     * Commit the current utterance as a single FINAL and reset. Called only at a
+     * real boundary: endpoint (via the debounce path), finalize ack, or close.
+     */
+    emitFinal(startMs, endMs) {
+        const text = this.lastSentInterim.trim();
+        if (!text) {
+            this.resetActiveUtterance();
+            return;
+        }
+        this.emitFinalEvent({
+            text,
+            utteranceId: this.currentUtteranceId ?? undefined,
+            speakerId: this.currentSpeakerId,
+            language: this.currentLanguage,
+            originalText: this.lastOriginalText || undefined,
+            sourceLanguage: this.currentSourceLanguage,
+            startMs,
+            endMs,
+        });
+        this.resetActiveUtterance();
+    }
+    handleResult(result) {
+        const allTokens = result.tokens ?? [];
+        // Translation sessions: translation-status tokens are the emitted text;
+        // original-status tokens feed the parallel originalText accumulator.
+        // Untagged tokens are Soniox's SAME-LANGUAGE PASSTHROUGH (spoken language
+        // already equals the target): keep them as emitted text so target-language
+        // speech surfaces instead of silently vanishing.
+        const tokens = this.opts.isTranslation ? allTokens.filter((t) => t.translation_status !== "original") : allTokens;
+        const originalTokens = this.opts.isTranslation ? allTokens.filter((t) => t.translation_status === "original") : [];
+        let interimText = "";
+        let language;
+        let startMs;
+        let endMs;
+        let currentFinalText = "";
+        for (const t of tokens) {
+            // First token of the stream (or first after a commit): open an
+            // utterance. Deliberately no split on per-token speaker change — Soniox
+            // re-diarizes its rolling window, so speaker labels flicker within one
+            // real utterance; the single FINAL comes from the endpoint boundary.
+            if (!this.currentUtteranceId) {
+                this.startNewUtterance(t.speaker, t.language ?? this.opts.fallbackLanguage);
+            }
+            if (t.speaker)
+                this.currentSpeakerId = t.speaker;
+            if (t.language) {
+                language = t.language;
+                this.currentLanguage = t.language;
+                // Untagged tokens on a translation session are the same-language
+                // passthrough: their language IS the detected source.
+                if (this.opts.isTranslation && t.translation_status == null) {
+                    this.currentSourceLanguage = t.language;
+                }
+            }
+            if (t.is_final) {
+                currentFinalText += t.text;
+                if (t.start_ms != null && startMs == null)
+                    startMs = t.start_ms;
+                if (t.end_ms != null)
+                    endMs = t.end_ms;
+            }
+            else {
+                interimText += t.text;
+            }
+        }
+        if (originalTokens.length > 0) {
+            let originalInterim = "";
+            let originalFinal = "";
+            for (const t of originalTokens) {
+                if (!this.currentUtteranceId) {
+                    this.startNewUtterance(t.speaker, undefined);
+                }
+                if (t.language)
+                    this.currentSourceLanguage = t.language;
+                if (t.is_final)
+                    originalFinal += t.text;
+                else
+                    originalInterim += t.text;
+            }
+            if (originalFinal.length > this.originalPrevFinalLen) {
+                this.originalStablePrefix += originalFinal.slice(this.originalPrevFinalLen);
+            }
+            this.originalPrevFinalLen = originalFinal.length;
+            const originalComposite = this.originalStablePrefix + originalInterim;
+            if (originalComposite.length > 0)
+                this.lastOriginalText = originalComposite;
+        }
+        // Accumulate finalized text: on window growth append the delta; on shrink
+        // (compaction) keep stablePrefix and reset the tracker.
+        if (currentFinalText.length > this.prevFinalLen) {
+            this.stablePrefix += currentFinalText.slice(this.prevFinalLen);
+        }
+        this.prevFinalLen = currentFinalText.length;
+        const compositeText = (this.stablePrefix + interimText).trim();
+        if (compositeText.length === 0)
+            return;
+        if (this.pendingFinalText && this.pendingFinalTimer !== null) {
+            const pendingBefore = this.pendingFinalText;
+            const merged = this.tryMergeOverlap(this.pendingFinalText, compositeText);
+            if (merged !== null) {
+                this.currentUtteranceId = this.pendingFinalUtteranceId ?? this.currentUtteranceId;
+                if (this.currentSpeakerId)
+                    this.pendingFinalSpeakerId = this.currentSpeakerId;
+                else
+                    this.currentSpeakerId = this.pendingFinalSpeakerId;
+                if (this.currentLanguage)
+                    this.pendingFinalLanguage = this.currentLanguage;
+                else
+                    this.currentLanguage = this.pendingFinalLanguage;
+                if (this.lastOriginalText) {
+                    this.pendingFinalOriginalText =
+                        this.tryMergeOverlap(this.pendingFinalOriginalText, this.lastOriginalText) ?? this.lastOriginalText;
+                }
+                if (this.currentSourceLanguage) {
+                    this.pendingFinalSourceLanguage = this.currentSourceLanguage;
+                }
+                this.pendingFinalText = merged;
+                this.lastSentInterim = merged;
+                if (merged !== pendingBefore) {
+                    this.opts.onTranscript({
+                        text: merged,
+                        isFinal: false,
+                        utteranceId: this.currentUtteranceId ?? undefined,
+                        speakerId: this.currentSpeakerId,
+                        language,
+                        originalText: this.pendingFinalOriginalText || undefined,
+                        sourceLanguage: this.pendingFinalSourceLanguage,
+                        startMs,
+                        endMs,
+                    });
+                    this.schedulePendingFinalCommit();
+                }
+                return;
+            }
+            this.commitPendingFinal();
+        }
+        // Emit an INTERIM only; the single final comes at the utterance boundary.
+        if (this.isLikelyWindowRegression(compositeText, this.lastSentInterim)) {
+            return;
+        }
+        if (compositeText !== this.lastSentInterim) {
+            this.opts.onTranscript({
+                text: compositeText,
+                isFinal: false,
+                utteranceId: this.currentUtteranceId ?? undefined,
+                speakerId: this.currentSpeakerId,
+                language,
+                originalText: this.lastOriginalText || undefined,
+                sourceLanguage: this.currentSourceLanguage,
+                startMs,
+                endMs,
+            });
+            this.lastSentInterim = compositeText;
+        }
+    }
+    handleEndpoint() {
+        // Soft boundary: defer the FINAL briefly; if Soniox's next rolling window
+        // overlaps, it merges back into this utterance.
+        const candidate = this.lastSentInterim.trim();
+        if (candidate) {
+            if (this.pendingFinalText && this.pendingFinalTimer !== null) {
+                const merged = this.tryMergeOverlap(this.pendingFinalText, candidate);
+                if (merged !== null) {
+                    this.pendingFinalText = merged;
+                    if (this.currentSpeakerId)
+                        this.pendingFinalSpeakerId = this.currentSpeakerId;
+                    if (this.currentLanguage)
+                        this.pendingFinalLanguage = this.currentLanguage;
+                    if (this.lastOriginalText) {
+                        this.pendingFinalOriginalText =
+                            this.tryMergeOverlap(this.pendingFinalOriginalText, this.lastOriginalText) ?? this.lastOriginalText;
+                    }
+                    if (this.currentSourceLanguage) {
+                        this.pendingFinalSourceLanguage = this.currentSourceLanguage;
+                    }
+                }
+                else {
+                    this.commitPendingFinal();
+                    this.setPendingFinalFromActive(candidate);
+                }
+            }
+            else {
+                this.setPendingFinalFromActive(candidate);
+            }
+            this.schedulePendingFinalCommit();
+        }
+        this.resetActiveUtterance();
+    }
+    /** Finalize ack (`<fin>`): trusted "speech segment ended" signal. */
+    handleFinalized() {
+        this.commitPendingFinal();
+        this.emitFinal();
+    }
+    /** Flush any in-flight utterance as a final (session close/teardown). */
+    flush() {
+        this.commitPendingFinal();
+        this.emitFinal();
+    }
+    /** Drop all state without emitting (e.g. audio-gap resume). */
+    discard() {
+        this.clearPendingFinal();
+        this.resetActiveUtterance();
+    }
+}
+//# sourceMappingURL=SonioxTokenAssembler.js.map

package/build/services/soniox/SonioxTokenAssembler.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"SonioxTokenAssembler.js","sourceRoot":"","sources":["../../../src/services/soniox/SonioxTokenAssembler.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;GAoBG;AAEH,OAAO,EAAC,OAAO,EAAC,MAAM,oBAAoB,CAAA;AAG1C,MAAM,0BAA0B,GAAG,CAAC,CAAA;AACpC,MAAM,yBAAyB,GAAG,EAAE,CAAA;AACpC,MAAM,2BAA2B,GAAG,EAAE,CAAA;AACtC,MAAM,4BAA4B,GAAG,GAAG,CAAA;AA4BxC,MAAM,OAAO,oBAAoB;IACd,IAAI,CAA6B;IACjC,kBAAkB,CAAQ;IAC1B,YAAY,CAAwC;IACpD,cAAc,CAAsB;IAErD,8BAA8B;IACtB,YAAY,GAAG,EAAE,CAAA;IACjB,YAAY,GAAG,CAAC,CAAA;IAChB,eAAe,GAAG,EAAE,CAAA;IAE5B,4EAA4E;IACpE,oBAAoB,GAAG,EAAE,CAAA;IACzB,oBAAoB,GAAG,CAAC,CAAA;IACxB,gBAAgB,GAAG,EAAE,CAAA;IACrB,qBAAqB,GAAuB,SAAS,CAAA;IAErD,kBAAkB,GAAkB,IAAI,CAAA;IACxC,gBAAgB,GAAuB,SAAS,CAAA;IAChD,eAAe,GAAuB,SAAS,CAAA;IAEvD,sCAAsC;IAC9B,gBAAgB,GAAG,EAAE,CAAA;IACrB,uBAAuB,GAAkB,IAAI,CAAA;IAC7C,qBAAqB,GAAuB,SAAS,CAAA;IACrD,oBAAoB,GAAuB,SAAS,CAAA;IACpD,wBAAwB,GAAG,EAAE,CAAA;IAC7B,0BAA0B,GAAuB,SAAS,CAAA;IAC1D,mBAAmB,GAAuB,SAAS,CAAA;IACnD,iBAAiB,GAAuB,SAAS,CAAA;IACjD,iBAAiB,GAAkB,IAAI,CAAA;IAE/C,6EAA6E;IACrE,YAAY,GAAG,CAAC,CAAA;IACP,UAAU,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,CAAA;IAErE,YAAY,IAAiC;QAC3C,IAAI,CAAC,IAAI,GAAG,IAAI,CAAA;QAChB,IAAI,CAAC,kBAAkB,GAAG,IAAI,CAAC,kBAAkB,IAAI,4BAA4B,CAAA;QACjF,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC,YAAY,IAAI,CAAC,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC,OAAO,CAAC,UAAU,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,CAAA;QACjF,IAAI,CAAC,cAAc,GAAG,IAAI,CAAC,cAAc,IAAI,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,OAAO,CAAC,YAAY,CAAC,EAAE,CAAC,CAAC,CAAA;IACjF,CAAC;IAEO,eAAe;QACrB,IAAI,CAAC,YAAY,IAAI,CAAC,CAAA;QACtB,OAAO,OAAO,IAAI,CAAC,IAAI,CAAC,KAAK,IAAI,IAAI,CAAC,UAAU,IAAI,IAAI,CAAC,YAAY,EAAE,CAAA;IACzE,CAAC;IAEO,iBAAiB,CAAC,SAAkB,EAAE,QAAiB;QAC7D,IAAI,CAAC,kBAAkB,GAAG,IAAI,CAAC,eAAe,EAAE,CAAA;QAChD,IAAI,CAAC,gBAAgB,GAAG,SAAS,CAAA;QACjC,IAAI,CAAC,eAAe,GAAG,QAAQ,CAAA;QAC/B,IAAI,CAAC,YAAY,GAAG,EAAE,CAAA;QACtB,IAAI,CAAC,YAAY,GAAG,CAAC,CAAA;QACrB,IAAI,CAAC,eAAe,GAAG,EAAE,CAAA;QACzB,IAAI,CAAC,oBAAoB,GAAG,EAAE,CAAA;QAC9B,IAAI,CAAC,oBAAoB,GAAG,CAAC,CAAA;QAC7B,IAAI,CAAC,gBAAgB,GAAG,EAAE,CAAA;QAC1B,IAAI,CAAC,qBAAqB,GAAG,SAAS,CAAA;IACxC,CAAC;IAEO,oBAAoB;QAC1B,IAAI,CAAC,YAAY,GAAG,EAAE,CAAA;QACtB,IAAI,CAAC,YAAY,GAAG,CAAC,CAAA;QACrB,IAAI,CAAC,eAAe,GAAG,EAAE,CAAA;QACzB,IAAI,CAAC,oBAAoB,GAAG,EAAE,CAAA;QAC9B,IAAI,CAAC,oBAAoB,GAAG,CAAC,CAAA;QAC7B,IAAI,CAAC,gBAAgB,GAAG,EAAE,CAAA;QAC1B,IAAI,CAAC,qBAAqB,GAAG,SAAS,CAAA;QACtC,IAAI,CAAC,kBAAkB,GAAG,IAAI,CAAA;QAC9B,IAAI,CAAC,gBAAgB,GAAG,SAAS,CAAA;QACjC,IAAI,CAAC,eAAe,GAAG,SAAS,CAAA;IAClC,CAAC;IAEO,iBAAiB;QACvB,IAAI,IAAI,CAAC,iBAAiB,KAAK,IAAI,EAAE,CAAC;YACpC,IAAI,CAAC,cAAc,CAAC,IAAI,CAAC,iBAAiB,CAAC,CAAA;YAC3C,IAAI,CAAC,iBAAiB,GAAG,IAAI,CAAA;QAC/B,CAAC;QACD,IAAI,CAAC,gBAAgB,GAAG,EAAE,CAAA;QAC1B,IAAI,CAAC,uBAAuB,GAAG,IAAI,CAAA;QACnC,IAAI,CAAC,qBAAqB,GAAG,SAAS,CAAA;QACtC,IAAI,CAAC,oBAAoB,GAAG,SAAS,CAAA;QACrC,IAAI,CAAC,wBAAwB,GAAG,EAAE,CAAA;QAClC,IAAI,CAAC,0BAA0B,GAAG,SAAS,CAAA;QAC3C,IAAI,CAAC,mBAAmB,GAAG,SAAS,CAAA;QACpC,IAAI,CAAC,iBAAiB,GAAG,SAAS,CAAA;IACpC,CAAC;IAEO,eAAe,CAAC,CAAS,EAAE,CAAS;QAC1C,MAAM,IAAI,GAAG,CAAC,CAAC,IAAI,EAAE,CAAA;QACrB,MAAM,KAAK,GAAG,CAAC,CAAC,IAAI,EAAE,CAAA;QACtB,IAAI,CAAC,IAAI,IAAI,CAAC,KAAK;YAAE,OAAO,IAAI,CAAA;QAChC,IAAI,IAAI,KAAK,KAAK,IAAI,IAAI,CAAC,UAAU,CAAC,KAAK,CAAC;YAAE,OAAO,IAAI,CAAA;QACzD,IAAI,KAAK,CAAC,UAAU,CAAC,IAAI,CAAC;YAAE,OAAO,IAAI,GAAG,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,CAAA;QAElE,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC,MAAM,CAAC,CAAA;QACtD,KAAK,IAAI,GAAG,GAAG,UAAU,EAAE,GAAG,IAAI,0BAA0B,EAAE,GAAG,IAAI,CAAC,EAAE,CAAC;YACvE,IAAI,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,EAAE,CAAC;gBACvC,OAAO,IAAI,GAAG,KAAK,CAAC,KAAK,CAAC,GAAG,CAAC,CAAA;YAChC,CAAC;QACH,CAAC;QACD,OAAO,IAAI,CAAA;IACb,CAAC;IAEO,wBAAwB,CAAC,IAAY,EAAE,QAAgB;QAC7D,MAAM,cAAc,GAAG,IAAI,CAAC,IAAI,EAAE,CAAA;QAClC,MAAM,kBAAkB,GAAG,QAAQ,CAAC,IAAI,EAAE,CAAA;QAC1C,IAAI,CAAC,cAAc,IAAI,CAAC,kBAAkB;YAAE,OAAO,KAAK,CAAA;QACxD,MAAM,IAAI,GAAG,kBAAkB,CAAC,MAAM,GAAG,cAAc,CAAC,MAAM,CAAA;QAC9D,IAAI,IAAI,GAAG,yBAAyB;YAAE,OAAO,KAAK,CAAA;QAElD,IAAI,YAAY,GAAG,CAAC,CAAA;QACpB,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,cAAc,CAAC,MAAM,EAAE,kBAAkB,CAAC,MAAM,CAAC,CAAA;QAC5E,OAAO,YAAY,GAAG,SAAS,IAAI,cAAc,CAAC,YAAY,CAAC,KAAK,kBAAkB,CAAC,YAAY,CAAC,EAAE,CAAC;YACrG,YAAY,IAAI,CAAC,CAAA;QACnB,CAAC;QAED,OAAO,YAAY,IAAI,IAAI,CAAC,GAAG,CAAC,2BAA2B,EAAE,cAAc,CAAC,MAAM,CAAC,CAAA;IACrF,CAAC;IAEO,cAAc,CAAC,KAStB;QACC,IAAI,CAAC,IAAI,CAAC,YAAY,CAAC;YACrB,IAAI,EAAE,KAAK,CAAC,IAAI;YAChB,OAAO,EAAE,IAAI;YACb,WAAW,EAAE,KAAK,CAAC,WAAW;YAC9B,SAAS,EAAE,KAAK,CAAC,SAAS;YAC1B,QAAQ,EAAE,KAAK,CAAC,QAAQ;YACxB,YAAY,EAAE,KAAK,CAAC,YAAY,IAAI,SAAS;YAC7C,cAAc,EAAE,KAAK,CAAC,cAAc;YACpC,OAAO,EAAE,KAAK,CAAC,OAAO;YACtB,KAAK,EAAE,KAAK,CAAC,KAAK;SACnB,CAAC,CAAA;IACJ,CAAC;IAEO,kBAAkB;QACxB,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC3B,IAAI,CAAC,iBAAiB,EAAE,CAAA;YACxB,OAAM;QACR,CAAC;QAED,MAAM,aAAa,GAAG,IAAI,CAAC,gBAAgB,CAAA;QAC3C,MAAM,oBAAoB,GAAG,IAAI,CAAC,uBAAuB,CAAA;QACzD,IAAI,CAAC,cAAc,CAAC;YAClB,IAAI,EAAE,aAAa;YACnB,WAAW,EAAE,oBAAoB,IAAI,SAAS;YAC9C,SAAS,EAAE,IAAI,CAAC,qBAAqB;YACrC,QAAQ,EAAE,IAAI,CAAC,oBAAoB;YACnC,YAAY,EAAE,IAAI,CAAC,wBAAwB;YAC3C,cAAc,EAAE,IAAI,CAAC,0BAA0B;YAC/C,OAAO,EAAE,IAAI,CAAC,mBAAmB;YACjC,KAAK,EAAE,IAAI,CAAC,iBAAiB;SAC9B,CAAC,CAAA;QACF,IACE,oBAAoB;YACpB,IAAI,CAAC,kBAAkB,KAAK,oBAAoB;YAChD,IAAI,CAAC,eAAe,CAAC,IAAI,EAAE,KAAK,aAAa,EAC7C,CAAC;YACD,IAAI,CAAC,oBAAoB,EAAE,CAAA;QAC7B,CAAC;QACD,IAAI,CAAC,iBAAiB,EAAE,CAAA;IAC1B,CAAC;IAEO,0BAA0B;QAChC,IAAI,IAAI,CAAC,iBAAiB,KAAK,IAAI;YAAE,IAAI,CAAC,cAAc,CAAC,IAAI,CAAC,iBAAiB,CAAC,CAAA;QAChF,IAAI,CAAC,iBAAiB,GAAG,IAAI,CAAC,YAAY,CAAC,GAAG,EAAE;YAC9C,IAAI,CAAC,kBAAkB,EAAE,CAAA;QAC3B,CAAC,EAAE,IAAI,CAAC,kBAAkB,CAAC,CAAA;IAC7B,CAAC;IAEO,yBAAyB,CAAC,IAAY;QAC5C,IAAI,CAAC,gBAAgB,GAAG,IAAI,CAAA;QAC5B,IAAI,CAAC,uBAAuB,GAAG,IAAI,CAAC,kBAAkB,CAAA;QACtD,IAAI,CAAC,qBAAqB,GAAG,IAAI,CAAC,gBAAgB,CAAA;QAClD,IAAI,CAAC,oBAAoB,GAAG,IAAI,CAAC,eAAe,CAAA;QAChD,IAAI,CAAC,wBAAwB,GAAG,IAAI,CAAC,gBAAgB,CAAA;QACrD,IAAI,CAAC,0BAA0B,GAAG,IAAI,CAAC,qBAAqB,CAAA;IAC9D,CAAC;IAED;;;OAGG;IACK,SAAS,CAAC,OAAgB,EAAE,KAAc;QAChD,MAAM,IAAI,GAAG,IAAI,CAAC,eAAe,CAAC,IAAI,EAAE,CAAA;QACxC,IAAI,CAAC,IAAI,EAAE,CAAC;YACV,IAAI,CAAC,oBAAoB,EAAE,CAAA;YAC3B,OAAM;QACR,CAAC;QACD,IAAI,CAAC,cAAc,CAAC;YAClB,IAAI;YACJ,WAAW,EAAE,IAAI,CAAC,kBAAkB,IAAI,SAAS;YACjD,SAAS,EAAE,IAAI,CAAC,gBAAgB;YAChC,QAAQ,EAAE,IAAI,CAAC,eAAe;YAC9B,YAAY,EAAE,IAAI,CAAC,gBAAgB,IAAI,SAAS;YAChD,cAAc,EAAE,IAAI,CAAC,qBAAqB;YAC1C,OAAO;YACP,KAAK;SACN,CAAC,CAAA;QACF,IAAI,CAAC,oBAAoB,EAAE,CAAA;IAC7B,CAAC;IAED,YAAY,CAAC,MAAoB;QAC/B,MAAM,SAAS,GAAG,MAAM,CAAC,MAAM,IAAI,EAAE,CAAA;QAErC,wEAAwE;QACxE,qEAAqE;QACrE,0EAA0E;QAC1E,2EAA2E;QAC3E,iDAAiD;QACjD,MAAM,MAAM,GAAG,IAAI,CAAC,IAAI,CAAC,aAAa,CAAC,CAAC,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,kBAAkB,KAAK,UAAU,CAAC,CAAC,CAAC,CAAC,SAAS,CAAA;QACjH,MAAM,cAAc,GAAG,IAAI,CAAC,IAAI,CAAC,aAAa,CAAC,CAAC,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,kBAAkB,KAAK,UAAU,CAAC,CAAC,CAAC,CAAC,EAAE,CAAA;QAElH,IAAI,WAAW,GAAG,EAAE,CAAA;QACpB,IAAI,QAA4B,CAAA;QAChC,IAAI,OAA2B,CAAA;QAC/B,IAAI,KAAyB,CAAA;QAC7B,IAAI,gBAAgB,GAAG,EAAE,CAAA;QAEzB,KAAK,MAAM,CAAC,IAAI,MAAM,EAAE,CAAC;YACvB,+DAA+D;YAC/D,wEAAwE;YACxE,uEAAuE;YACvE,qEAAqE;YACrE,IAAI,CAAC,IAAI,CAAC,kBAAkB,EAAE,CAAC;gBAC7B,IAAI,CAAC,iBAAiB,CAAC,CAAC,CAAC,OAAO,EAAE,CAAC,CAAC,QAAQ,IAAI,IAAI,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAA;YAC7E,CAAC;YAED,IAAI,CAAC,CAAC,OAAO;gBAAE,IAAI,CAAC,gBAAgB,GAAG,CAAC,CAAC,OAAO,CAAA;YAChD,IAAI,CAAC,CAAC,QAAQ,EAAE,CAAC;gBACf,QAAQ,GAAG,CAAC,CAAC,QAAQ,CAAA;gBACrB,IAAI,CAAC,eAAe,GAAG,CAAC,CAAC,QAAQ,CAAA;gBACjC,iEAAiE;gBACjE,sDAAsD;gBACtD,IAAI,IAAI,CAAC,IAAI,CAAC,aAAa,IAAI,CAAC,CAAC,kBAAkB,IAAI,IAAI,EAAE,CAAC;oBAC5D,IAAI,CAAC,qBAAqB,GAAG,CAAC,CAAC,QAAQ,CAAA;gBACzC,CAAC;YACH,CAAC;YAED,IAAI,CAAC,CAAC,QAAQ,EAAE,CAAC;gBACf,gBAAgB,IAAI,CAAC,CAAC,IAAI,CAAA;gBAC1B,IAAI,CAAC,CAAC,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,IAAI;oBAAE,OAAO,GAAG,CAAC,CAAC,QAAQ,CAAA;gBAC/D,IAAI,CAAC,CAAC,MAAM,IAAI,IAAI;oBAAE,KAAK,GAAG,CAAC,CAAC,MAAM,CAAA;YACxC,CAAC;iBAAM,CAAC;gBACN,WAAW,IAAI,CAAC,CAAC,IAAI,CAAA;YACvB,CAAC;QACH,CAAC;QAED,IAAI,cAAc,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAC9B,IAAI,eAAe,GAAG,EAAE,CAAA;YACxB,IAAI,aAAa,GAAG,EAAE,CAAA;YACtB,KAAK,MAAM,CAAC,IAAI,cAAc,EAAE,CAAC;gBAC/B,IAAI,CAAC,IAAI,CAAC,kBAAkB,EAAE,CAAC;oBAC7B,IAAI,CAAC,iBAAiB,CAAC,CAAC,CAAC,OAAO,EAAE,SAAS,CAAC,CAAA;gBAC9C,CAAC;gBACD,IAAI,CAAC,CAAC,QAAQ;oBAAE,IAAI,CAAC,qBAAqB,GAAG,CAAC,CAAC,QAAQ,CAAA;gBACvD,IAAI,CAAC,CAAC,QAAQ;oBAAE,aAAa,IAAI,CAAC,CAAC,IAAI,CAAA;;oBAClC,eAAe,IAAI,CAAC,CAAC,IAAI,CAAA;YAChC,CAAC;YACD,IAAI,aAAa,CAAC,MAAM,GAAG,IAAI,CAAC,oBAAoB,EAAE,CAAC;gBACrD,IAAI,CAAC,oBAAoB,IAAI,aAAa,CAAC,KAAK,CAAC,IAAI,CAAC,oBAAoB,CAAC,CAAA;YAC7E,CAAC;YACD,IAAI,CAAC,oBAAoB,GAAG,aAAa,CAAC,MAAM,CAAA;YAChD,MAAM,iBAAiB,GAAG,IAAI,CAAC,oBAAoB,GAAG,eAAe,CAAA;YACrE,IAAI,iBAAiB,CAAC,MAAM,GAAG,CAAC;gBAAE,IAAI,CAAC,gBAAgB,GAAG,iBAAiB,CAAA;QAC7E,CAAC;QAED,0EAA0E;QAC1E,wDAAwD;QACxD,IAAI,gBAAgB,CAAC,MAAM,GAAG,IAAI,CAAC,YAAY,EAAE,CAAC;YAChD,IAAI,CAAC,YAAY,IAAI,gBAAgB,CAAC,KAAK,CAAC,IAAI,CAAC,YAAY,CAAC,CAAA;QAChE,CAAC;QACD,IAAI,CAAC,YAAY,GAAG,gBAAgB,CAAC,MAAM,CAAA;QAE3C,MAAM,aAAa,GAAG,CAAC,IAAI,CAAC,YAAY,GAAG,WAAW,CAAC,CAAC,IAAI,EAAE,CAAA;QAC9D,IAAI,aAAa,CAAC,MAAM,KAAK,CAAC;YAAE,OAAM;QAEtC,IAAI,IAAI,CAAC,gBAAgB,IAAI,IAAI,CAAC,iBAAiB,KAAK,IAAI,EAAE,CAAC;YAC7D,MAAM,aAAa,GAAG,IAAI,CAAC,gBAAgB,CAAA;YAC3C,MAAM,MAAM,GAAG,IAAI,CAAC,eAAe,CAAC,IAAI,CAAC,gBAAgB,EAAE,aAAa,CAAC,CAAA;YACzE,IAAI,MAAM,KAAK,IAAI,EAAE,CAAC;gBACpB,IAAI,CAAC,kBAAkB,GAAG,IAAI,CAAC,uBAAuB,IAAI,IAAI,CAAC,kBAAkB,CAAA;gBACjF,IAAI,IAAI,CAAC,gBAAgB;oBAAE,IAAI,CAAC,qBAAqB,GAAG,IAAI,CAAC,gBAAgB,CAAA;;oBACxE,IAAI,CAAC,gBAAgB,GAAG,IAAI,CAAC,qBAAqB,CAAA;gBACvD,IAAI,IAAI,CAAC,eAAe;oBAAE,IAAI,CAAC,oBAAoB,GAAG,IAAI,CAAC,eAAe,CAAA;;oBACrE,IAAI,CAAC,eAAe,GAAG,IAAI,CAAC,oBAAoB,CAAA;gBACrD,IAAI,IAAI,CAAC,gBAAgB,EAAE,CAAC;oBAC1B,IAAI,CAAC,wBAAwB;wBAC3B,IAAI,CAAC,eAAe,CAAC,IAAI,CAAC,wBAAwB,EAAE,IAAI,CAAC,gBAAgB,CAAC,IAAI,IAAI,CAAC,gBAAgB,CAAA;gBACvG,CAAC;gBACD,IAAI,IAAI,CAAC,qBAAqB,EAAE,CAAC;oBAC/B,IAAI,CAAC,0BAA0B,GAAG,IAAI,CAAC,qBAAqB,CAAA;gBAC9D,CAAC;gBAED,IAAI,CAAC,gBAAgB,GAAG,MAAM,CAAA;gBAC9B,IAAI,CAAC,eAAe,GAAG,MAAM,CAAA;gBAE7B,IAAI,MAAM,KAAK,aAAa,EAAE,CAAC;oBAC7B,IAAI,CAAC,IAAI,CAAC,YAAY,CAAC;wBACrB,IAAI,EAAE,MAAM;wBACZ,OAAO,EAAE,KAAK;wBACd,WAAW,EAAE,IAAI,CAAC,kBAAkB,IAAI,SAAS;wBACjD,SAAS,EAAE,IAAI,CAAC,gBAAgB;wBAChC,QAAQ;wBACR,YAAY,EAAE,IAAI,CAAC,wBAAwB,IAAI,SAAS;wBACxD,cAAc,EAAE,IAAI,CAAC,0BAA0B;wBAC/C,OAAO;wBACP,KAAK;qBACN,CAAC,CAAA;oBACF,IAAI,CAAC,0BAA0B,EAAE,CAAA;gBACnC,CAAC;gBACD,OAAM;YACR,CAAC;YAED,IAAI,CAAC,kBAAkB,EAAE,CAAA;QAC3B,CAAC;QAED,0EAA0E;QAC1E,IAAI,IAAI,CAAC,wBAAwB,CAAC,aAAa,EAAE,IAAI,CAAC,eAAe,CAAC,EAAE,CAAC;YACvE,OAAM;QACR,CAAC;QAED,IAAI,aAAa,KAAK,IAAI,CAAC,eAAe,EAAE,CAAC;YAC3C,IAAI,CAAC,IAAI,CAAC,YAAY,CAAC;gBACrB,IAAI,EAAE,aAAa;gBACnB,OAAO,EAAE,KAAK;gBACd,WAAW,EAAE,IAAI,CAAC,kBAAkB,IAAI,SAAS;gBACjD,SAAS,EAAE,IAAI,CAAC,gBAAgB;gBAChC,QAAQ;gBACR,YAAY,EAAE,IAAI,CAAC,gBAAgB,IAAI,SAAS;gBAChD,cAAc,EAAE,IAAI,CAAC,qBAAqB;gBAC1C,OAAO;gBACP,KAAK;aACN,CAAC,CAAA;YACF,IAAI,CAAC,eAAe,GAAG,aAAa,CAAA;QACtC,CAAC;IACH,CAAC;IAED,cAAc;QACZ,0EAA0E;QAC1E,gDAAgD;QAChD,MAAM,SAAS,GAAG,IAAI,CAAC,eAAe,CAAC,IAAI,EAAE,CAAA;QAC7C,IAAI,SAAS,EAAE,CAAC;YACd,IAAI,IAAI,CAAC,gBAAgB,IAAI,IAAI,CAAC,iBAAiB,KAAK,IAAI,EAAE,CAAC;gBAC7D,MAAM,MAAM,GAAG,IAAI,CAAC,eAAe,CAAC,IAAI,CAAC,gBAAgB,EAAE,SAAS,CAAC,CAAA;gBACrE,IAAI,MAAM,KAAK,IAAI,EAAE,CAAC;oBACpB,IAAI,CAAC,gBAAgB,GAAG,MAAM,CAAA;oBAC9B,IAAI,IAAI,CAAC,gBAAgB;wBAAE,IAAI,CAAC,qBAAqB,GAAG,IAAI,CAAC,gBAAgB,CAAA;oBAC7E,IAAI,IAAI,CAAC,eAAe;wBAAE,IAAI,CAAC,oBAAoB,GAAG,IAAI,CAAC,eAAe,CAAA;oBAC1E,IAAI,IAAI,CAAC,gBAAgB,EAAE,CAAC;wBAC1B,IAAI,CAAC,wBAAwB;4BAC3B,IAAI,CAAC,eAAe,CAAC,IAAI,CAAC,wBAAwB,EAAE,IAAI,CAAC,gBAAgB,CAAC,IAAI,IAAI,CAAC,gBAAgB,CAAA;oBACvG,CAAC;oBACD,IAAI,IAAI,CAAC,qBAAqB,EAAE,CAAC;wBAC/B,IAAI,CAAC,0BAA0B,GAAG,IAAI,CAAC,qBAAqB,CAAA;oBAC9D,CAAC;gBACH,CAAC;qBAAM,CAAC;oBACN,IAAI,CAAC,kBAAkB,EAAE,CAAA;oBACzB,IAAI,CAAC,yBAAyB,CAAC,SAAS,CAAC,CAAA;gBAC3C,CAAC;YACH,CAAC;iBAAM,CAAC;gBACN,IAAI,CAAC,yBAAyB,CAAC,SAAS,CAAC,CAAA;YAC3C,CAAC;YACD,IAAI,CAAC,0BAA0B,EAAE,CAAA;QACnC,CAAC;QACD,IAAI,CAAC,oBAAoB,EAAE,CAAA;IAC7B,CAAC;IAED,qEAAqE;IACrE,eAAe;QACb,IAAI,CAAC,kBAAkB,EAAE,CAAA;QACzB,IAAI,CAAC,SAAS,EAAE,CAAA;IAClB,CAAC;IAED,yEAAyE;IACzE,KAAK;QACH,IAAI,CAAC,kBAAkB,EAAE,CAAA;QACzB,IAAI,CAAC,SAAS,EAAE,CAAA;IAClB,CAAC;IAED,+DAA+D;IAC/D,OAAO;QACL,IAAI,CAAC,iBAAiB,EAAE,CAAA;QACxB,IAAI,CAAC,oBAAoB,EAAE,CAAA;IAC7B,CAAC;CACF","sourcesContent":["/**\n * SonioxTokenAssembler\n *\n * Pure port of the cloud Soniox provider's utterance assembly\n * (cloud-v2/packages/runtime/src/services/audio/providers/soniox.ts). Soniox\n * emits a rolling token window that is re-finalized/compacted across results;\n * this class turns that stream into stable interim/final utterances:\n *\n * - finalized text accumulates into a stable prefix so interims never shrink\n * - a single FINAL per utterance, committed at the endpoint boundary\n * - the endpoint is a soft signal: the candidate final is held briefly\n *   (debounce) and merged back if the next rolling window overlaps\n * - speaker/language are tracked per token but never split an utterance\n *   (Soniox re-diarizes its window; per-token labels flicker)\n * - translation sessions split the window: translation-status tokens are the\n *   emitted text, original-status tokens ride along as `originalText`, and\n *   untagged tokens are Soniox's same-language passthrough\n *\n * Timers are injected for testability; production uses BgTimer so the debounce\n * survives app backgrounding.\n */\n\nimport {BgTimer} from \"../../utils/timers\"\nimport type {SonioxResult} from \"./SonioxRealtimeClient\"\n\nconst MIN_ENDPOINT_OVERLAP_CHARS = 3\nconst MIN_REGRESSION_DROP_CHARS = 12\nconst MIN_REGRESSION_PREFIX_CHARS = 12\nconst DEFAULT_ENDPOINT_DEBOUNCE_MS = 500\n\nexport interface AssembledTranscript {\n  text: string\n  isFinal: boolean\n  utteranceId?: string\n  speakerId?: string\n  language?: string\n  originalText?: string\n  sourceLanguage?: string\n  startMs?: number\n  endMs?: number\n}\n\nexport interface SonioxTokenAssemblerOptions {\n  /** Stable id used in minted utterance ids (readable per subscription). */\n  scope: string\n  /** True for translation sessions: split original/translation token halves. */\n  isTranslation: boolean\n  /** Subscription language, used when a token carries no detected language. */\n  fallbackLanguage?: string\n  onTranscript: (t: AssembledTranscript) => void\n  endpointDebounceMs?: number\n  /** Test seams; default to background-safe timers. */\n  setTimeoutFn?: (cb: () => void, ms: number) => number\n  clearTimeoutFn?: (id: number) => void\n}\n\nexport class SonioxTokenAssembler {\n  private readonly opts: SonioxTokenAssemblerOptions\n  private readonly endpointDebounceMs: number\n  private readonly setTimeoutFn: (cb: () => void, ms: number) => number\n  private readonly clearTimeoutFn: (id: number) => void\n\n  // === Per-utterance state ===\n  private stablePrefix = \"\"\n  private prevFinalLen = 0\n  private lastSentInterim = \"\"\n\n  // Translation subs only: parallel accumulator for original-language tokens.\n  private originalStablePrefix = \"\"\n  private originalPrevFinalLen = 0\n  private lastOriginalText = \"\"\n  private currentSourceLanguage: string | undefined = undefined\n\n  private currentUtteranceId: string | null = null\n  private currentSpeakerId: string | undefined = undefined\n  private currentLanguage: string | undefined = undefined\n\n  // Deferred final (endpoint debounce).\n  private pendingFinalText = \"\"\n  private pendingFinalUtteranceId: string | null = null\n  private pendingFinalSpeakerId: string | undefined = undefined\n  private pendingFinalLanguage: string | undefined = undefined\n  private pendingFinalOriginalText = \"\"\n  private pendingFinalSourceLanguage: string | undefined = undefined\n  private pendingFinalStartMs: number | undefined = undefined\n  private pendingFinalEndMs: number | undefined = undefined\n  private pendingFinalTimer: number | null = null\n\n  // Id minting: an instance nonce prevents collisions across service restarts.\n  private utteranceSeq = 0\n  private readonly instanceId = Math.random().toString(36).slice(2, 10)\n\n  constructor(opts: SonioxTokenAssemblerOptions) {\n    this.opts = opts\n    this.endpointDebounceMs = opts.endpointDebounceMs ?? DEFAULT_ENDPOINT_DEBOUNCE_MS\n    this.setTimeoutFn = opts.setTimeoutFn ?? ((cb, ms) => BgTimer.setTimeout(cb, ms))\n    this.clearTimeoutFn = opts.clearTimeoutFn ?? ((id) => BgTimer.clearTimeout(id))\n  }\n\n  private mintUtteranceId(): string {\n    this.utteranceSeq += 1\n    return `utt_${this.opts.scope}_${this.instanceId}_${this.utteranceSeq}`\n  }\n\n  private startNewUtterance(speakerId?: string, language?: string): void {\n    this.currentUtteranceId = this.mintUtteranceId()\n    this.currentSpeakerId = speakerId\n    this.currentLanguage = language\n    this.stablePrefix = \"\"\n    this.prevFinalLen = 0\n    this.lastSentInterim = \"\"\n    this.originalStablePrefix = \"\"\n    this.originalPrevFinalLen = 0\n    this.lastOriginalText = \"\"\n    this.currentSourceLanguage = undefined\n  }\n\n  private resetActiveUtterance(): void {\n    this.stablePrefix = \"\"\n    this.prevFinalLen = 0\n    this.lastSentInterim = \"\"\n    this.originalStablePrefix = \"\"\n    this.originalPrevFinalLen = 0\n    this.lastOriginalText = \"\"\n    this.currentSourceLanguage = undefined\n    this.currentUtteranceId = null\n    this.currentSpeakerId = undefined\n    this.currentLanguage = undefined\n  }\n\n  private clearPendingFinal(): void {\n    if (this.pendingFinalTimer !== null) {\n      this.clearTimeoutFn(this.pendingFinalTimer)\n      this.pendingFinalTimer = null\n    }\n    this.pendingFinalText = \"\"\n    this.pendingFinalUtteranceId = null\n    this.pendingFinalSpeakerId = undefined\n    this.pendingFinalLanguage = undefined\n    this.pendingFinalOriginalText = \"\"\n    this.pendingFinalSourceLanguage = undefined\n    this.pendingFinalStartMs = undefined\n    this.pendingFinalEndMs = undefined\n  }\n\n  private tryMergeOverlap(a: string, b: string): string | null {\n    const left = a.trim()\n    const right = b.trim()\n    if (!left || !right) return null\n    if (left === right || left.startsWith(right)) return left\n    if (right.startsWith(left)) return left + right.slice(left.length)\n\n    const maxOverlap = Math.min(left.length, right.length)\n    for (let len = maxOverlap; len >= MIN_ENDPOINT_OVERLAP_CHARS; len -= 1) {\n      if (left.endsWith(right.slice(0, len))) {\n        return left + right.slice(len)\n      }\n    }\n    return null\n  }\n\n  private isLikelyWindowRegression(next: string, previous: string): boolean {\n    const normalizedNext = next.trim()\n    const normalizedPrevious = previous.trim()\n    if (!normalizedNext || !normalizedPrevious) return false\n    const drop = normalizedPrevious.length - normalizedNext.length\n    if (drop < MIN_REGRESSION_DROP_CHARS) return false\n\n    let sharedPrefix = 0\n    const maxPrefix = Math.min(normalizedNext.length, normalizedPrevious.length)\n    while (sharedPrefix < maxPrefix && normalizedNext[sharedPrefix] === normalizedPrevious[sharedPrefix]) {\n      sharedPrefix += 1\n    }\n\n    return sharedPrefix >= Math.min(MIN_REGRESSION_PREFIX_CHARS, normalizedNext.length)\n  }\n\n  private emitFinalEvent(event: {\n    text: string\n    utteranceId?: string\n    speakerId?: string\n    language?: string\n    originalText?: string\n    sourceLanguage?: string\n    startMs?: number\n    endMs?: number\n  }): void {\n    this.opts.onTranscript({\n      text: event.text,\n      isFinal: true,\n      utteranceId: event.utteranceId,\n      speakerId: event.speakerId,\n      language: event.language,\n      originalText: event.originalText || undefined,\n      sourceLanguage: event.sourceLanguage,\n      startMs: event.startMs,\n      endMs: event.endMs,\n    })\n  }\n\n  private commitPendingFinal(): void {\n    if (!this.pendingFinalText) {\n      this.clearPendingFinal()\n      return\n    }\n\n    const committedText = this.pendingFinalText\n    const committedUtteranceId = this.pendingFinalUtteranceId\n    this.emitFinalEvent({\n      text: committedText,\n      utteranceId: committedUtteranceId ?? undefined,\n      speakerId: this.pendingFinalSpeakerId,\n      language: this.pendingFinalLanguage,\n      originalText: this.pendingFinalOriginalText,\n      sourceLanguage: this.pendingFinalSourceLanguage,\n      startMs: this.pendingFinalStartMs,\n      endMs: this.pendingFinalEndMs,\n    })\n    if (\n      committedUtteranceId &&\n      this.currentUtteranceId === committedUtteranceId &&\n      this.lastSentInterim.trim() === committedText\n    ) {\n      this.resetActiveUtterance()\n    }\n    this.clearPendingFinal()\n  }\n\n  private schedulePendingFinalCommit(): void {\n    if (this.pendingFinalTimer !== null) this.clearTimeoutFn(this.pendingFinalTimer)\n    this.pendingFinalTimer = this.setTimeoutFn(() => {\n      this.commitPendingFinal()\n    }, this.endpointDebounceMs)\n  }\n\n  private setPendingFinalFromActive(text: string): void {\n    this.pendingFinalText = text\n    this.pendingFinalUtteranceId = this.currentUtteranceId\n    this.pendingFinalSpeakerId = this.currentSpeakerId\n    this.pendingFinalLanguage = this.currentLanguage\n    this.pendingFinalOriginalText = this.lastOriginalText\n    this.pendingFinalSourceLanguage = this.currentSourceLanguage\n  }\n\n  /**\n   * Commit the current utterance as a single FINAL and reset. Called only at a\n   * real boundary: endpoint (via the debounce path), finalize ack, or close.\n   */\n  private emitFinal(startMs?: number, endMs?: number): void {\n    const text = this.lastSentInterim.trim()\n    if (!text) {\n      this.resetActiveUtterance()\n      return\n    }\n    this.emitFinalEvent({\n      text,\n      utteranceId: this.currentUtteranceId ?? undefined,\n      speakerId: this.currentSpeakerId,\n      language: this.currentLanguage,\n      originalText: this.lastOriginalText || undefined,\n      sourceLanguage: this.currentSourceLanguage,\n      startMs,\n      endMs,\n    })\n    this.resetActiveUtterance()\n  }\n\n  handleResult(result: SonioxResult): void {\n    const allTokens = result.tokens ?? []\n\n    // Translation sessions: translation-status tokens are the emitted text;\n    // original-status tokens feed the parallel originalText accumulator.\n    // Untagged tokens are Soniox's SAME-LANGUAGE PASSTHROUGH (spoken language\n    // already equals the target): keep them as emitted text so target-language\n    // speech surfaces instead of silently vanishing.\n    const tokens = this.opts.isTranslation ? allTokens.filter((t) => t.translation_status !== \"original\") : allTokens\n    const originalTokens = this.opts.isTranslation ? allTokens.filter((t) => t.translation_status === \"original\") : []\n\n    let interimText = \"\"\n    let language: string | undefined\n    let startMs: number | undefined\n    let endMs: number | undefined\n    let currentFinalText = \"\"\n\n    for (const t of tokens) {\n      // First token of the stream (or first after a commit): open an\n      // utterance. Deliberately no split on per-token speaker change — Soniox\n      // re-diarizes its rolling window, so speaker labels flicker within one\n      // real utterance; the single FINAL comes from the endpoint boundary.\n      if (!this.currentUtteranceId) {\n        this.startNewUtterance(t.speaker, t.language ?? this.opts.fallbackLanguage)\n      }\n\n      if (t.speaker) this.currentSpeakerId = t.speaker\n      if (t.language) {\n        language = t.language\n        this.currentLanguage = t.language\n        // Untagged tokens on a translation session are the same-language\n        // passthrough: their language IS the detected source.\n        if (this.opts.isTranslation && t.translation_status == null) {\n          this.currentSourceLanguage = t.language\n        }\n      }\n\n      if (t.is_final) {\n        currentFinalText += t.text\n        if (t.start_ms != null && startMs == null) startMs = t.start_ms\n        if (t.end_ms != null) endMs = t.end_ms\n      } else {\n        interimText += t.text\n      }\n    }\n\n    if (originalTokens.length > 0) {\n      let originalInterim = \"\"\n      let originalFinal = \"\"\n      for (const t of originalTokens) {\n        if (!this.currentUtteranceId) {\n          this.startNewUtterance(t.speaker, undefined)\n        }\n        if (t.language) this.currentSourceLanguage = t.language\n        if (t.is_final) originalFinal += t.text\n        else originalInterim += t.text\n      }\n      if (originalFinal.length > this.originalPrevFinalLen) {\n        this.originalStablePrefix += originalFinal.slice(this.originalPrevFinalLen)\n      }\n      this.originalPrevFinalLen = originalFinal.length\n      const originalComposite = this.originalStablePrefix + originalInterim\n      if (originalComposite.length > 0) this.lastOriginalText = originalComposite\n    }\n\n    // Accumulate finalized text: on window growth append the delta; on shrink\n    // (compaction) keep stablePrefix and reset the tracker.\n    if (currentFinalText.length > this.prevFinalLen) {\n      this.stablePrefix += currentFinalText.slice(this.prevFinalLen)\n    }\n    this.prevFinalLen = currentFinalText.length\n\n    const compositeText = (this.stablePrefix + interimText).trim()\n    if (compositeText.length === 0) return\n\n    if (this.pendingFinalText && this.pendingFinalTimer !== null) {\n      const pendingBefore = this.pendingFinalText\n      const merged = this.tryMergeOverlap(this.pendingFinalText, compositeText)\n      if (merged !== null) {\n        this.currentUtteranceId = this.pendingFinalUtteranceId ?? this.currentUtteranceId\n        if (this.currentSpeakerId) this.pendingFinalSpeakerId = this.currentSpeakerId\n        else this.currentSpeakerId = this.pendingFinalSpeakerId\n        if (this.currentLanguage) this.pendingFinalLanguage = this.currentLanguage\n        else this.currentLanguage = this.pendingFinalLanguage\n        if (this.lastOriginalText) {\n          this.pendingFinalOriginalText =\n            this.tryMergeOverlap(this.pendingFinalOriginalText, this.lastOriginalText) ?? this.lastOriginalText\n        }\n        if (this.currentSourceLanguage) {\n          this.pendingFinalSourceLanguage = this.currentSourceLanguage\n        }\n\n        this.pendingFinalText = merged\n        this.lastSentInterim = merged\n\n        if (merged !== pendingBefore) {\n          this.opts.onTranscript({\n            text: merged,\n            isFinal: false,\n            utteranceId: this.currentUtteranceId ?? undefined,\n            speakerId: this.currentSpeakerId,\n            language,\n            originalText: this.pendingFinalOriginalText || undefined,\n            sourceLanguage: this.pendingFinalSourceLanguage,\n            startMs,\n            endMs,\n          })\n          this.schedulePendingFinalCommit()\n        }\n        return\n      }\n\n      this.commitPendingFinal()\n    }\n\n    // Emit an INTERIM only; the single final comes at the utterance boundary.\n    if (this.isLikelyWindowRegression(compositeText, this.lastSentInterim)) {\n      return\n    }\n\n    if (compositeText !== this.lastSentInterim) {\n      this.opts.onTranscript({\n        text: compositeText,\n        isFinal: false,\n        utteranceId: this.currentUtteranceId ?? undefined,\n        speakerId: this.currentSpeakerId,\n        language,\n        originalText: this.lastOriginalText || undefined,\n        sourceLanguage: this.currentSourceLanguage,\n        startMs,\n        endMs,\n      })\n      this.lastSentInterim = compositeText\n    }\n  }\n\n  handleEndpoint(): void {\n    // Soft boundary: defer the FINAL briefly; if Soniox's next rolling window\n    // overlaps, it merges back into this utterance.\n    const candidate = this.lastSentInterim.trim()\n    if (candidate) {\n      if (this.pendingFinalText && this.pendingFinalTimer !== null) {\n        const merged = this.tryMergeOverlap(this.pendingFinalText, candidate)\n        if (merged !== null) {\n          this.pendingFinalText = merged\n          if (this.currentSpeakerId) this.pendingFinalSpeakerId = this.currentSpeakerId\n          if (this.currentLanguage) this.pendingFinalLanguage = this.currentLanguage\n          if (this.lastOriginalText) {\n            this.pendingFinalOriginalText =\n              this.tryMergeOverlap(this.pendingFinalOriginalText, this.lastOriginalText) ?? this.lastOriginalText\n          }\n          if (this.currentSourceLanguage) {\n            this.pendingFinalSourceLanguage = this.currentSourceLanguage\n          }\n        } else {\n          this.commitPendingFinal()\n          this.setPendingFinalFromActive(candidate)\n        }\n      } else {\n        this.setPendingFinalFromActive(candidate)\n      }\n      this.schedulePendingFinalCommit()\n    }\n    this.resetActiveUtterance()\n  }\n\n  /** Finalize ack (`<fin>`): trusted \"speech segment ended\" signal. */\n  handleFinalized(): void {\n    this.commitPendingFinal()\n    this.emitFinal()\n  }\n\n  /** Flush any in-flight utterance as a final (session close/teardown). */\n  flush(): void {\n    this.commitPendingFinal()\n    this.emitFinal()\n  }\n\n  /** Drop all state without emitting (e.g. audio-gap resume). */\n  discard(): void {\n    this.clearPendingFinal()\n    this.resetActiveUtterance()\n  }\n}\n"]}

package/build/services/speakerProfileStoreInstance.d.ts

@@ -0,0 +1,8 @@
+/**
+ * The engine's singleton SpeakerProfileStore — device-local voice profiles
+ * behind engine.speakers. Lives in its own module (not SpeakerProfileStore.ts)
+ * so the store class stays dependency-free for tests.
+ */
+import { SpeakerProfileStore } from "./SpeakerProfileStore";
+export declare const speakerProfileStore: SpeakerProfileStore;
+//# sourceMappingURL=speakerProfileStoreInstance.d.ts.map

package/build/services/speakerProfileStoreInstance.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"speakerProfileStoreInstance.d.ts","sourceRoot":"","sources":["../../src/services/speakerProfileStoreInstance.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAIH,OAAO,EAAC,mBAAmB,EAAC,MAAM,uBAAuB,CAAA;AAEzD,eAAO,MAAM,mBAAmB,qBAc9B,CAAA"}

package/build/services/speakerProfileStoreInstance.js

@@ -0,0 +1,24 @@
+/**
+ * The engine's singleton SpeakerProfileStore — device-local voice profiles
+ * behind engine.speakers. Lives in its own module (not SpeakerProfileStore.ts)
+ * so the store class stays dependency-free for tests.
+ */
+import { storage as mmkvStorage } from "../utils/storage/storage";
+import { cloudClientService } from "./CloudClientService";
+import { SpeakerProfileStore } from "./SpeakerProfileStore";
+export const speakerProfileStore = new SpeakerProfileStore({
+    getUserId: () => cloudClientService.resolveGlasslyUserId(),
+    backend: {
+        get(key) {
+            const result = mmkvStorage.load(key);
+            return result.is_ok() ? result.value : null;
+        },
+        set(key, value) {
+            mmkvStorage.save(key, value);
+        },
+        remove(key) {
+            mmkvStorage.remove(key);
+        },
+    },
+});
+//# sourceMappingURL=speakerProfileStoreInstance.js.map

package/build/services/speakerProfileStoreInstance.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"speakerProfileStoreInstance.js","sourceRoot":"","sources":["../../src/services/speakerProfileStoreInstance.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,OAAO,EAAC,OAAO,IAAI,WAAW,EAAC,MAAM,0BAA0B,CAAA;AAC/D,OAAO,EAAC,kBAAkB,EAAC,MAAM,sBAAsB,CAAA;AACvD,OAAO,EAAC,mBAAmB,EAAC,MAAM,uBAAuB,CAAA;AAEzD,MAAM,CAAC,MAAM,mBAAmB,GAAG,IAAI,mBAAmB,CAAC;IACzD,SAAS,EAAE,GAAG,EAAE,CAAC,kBAAkB,CAAC,oBAAoB,EAAE;IAC1D,OAAO,EAAE;QACP,GAAG,CAAC,GAAW;YACb,MAAM,MAAM,GAAG,WAAW,CAAC,IAAI,CAAU,GAAG,CAAC,CAAA;YAC7C,OAAO,MAAM,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAA;QAC7C,CAAC;QACD,GAAG,CAAC,GAAW,EAAE,KAAc;YAC7B,WAAW,CAAC,IAAI,CAAC,GAAG,EAAE,KAAK,CAAC,CAAA;QAC9B,CAAC;QACD,MAAM,CAAC,GAAW;YAChB,WAAW,CAAC,MAAM,CAAC,GAAG,CAAC,CAAA;QACzB,CAAC;KACF;CACF,CAAC,CAAA","sourcesContent":["/**\n * The engine's singleton SpeakerProfileStore — device-local voice profiles\n * behind engine.speakers. Lives in its own module (not SpeakerProfileStore.ts)\n * so the store class stays dependency-free for tests.\n */\n\nimport {storage as mmkvStorage} from \"../utils/storage/storage\"\nimport {cloudClientService} from \"./CloudClientService\"\nimport {SpeakerProfileStore} from \"./SpeakerProfileStore\"\n\nexport const speakerProfileStore = new SpeakerProfileStore({\n  getUserId: () => cloudClientService.resolveGlasslyUserId(),\n  backend: {\n    get(key: string): unknown | null {\n      const result = mmkvStorage.load<unknown>(key)\n      return result.is_ok() ? result.value : null\n    },\n    set(key: string, value: unknown): void {\n      mmkvStorage.save(key, value)\n    },\n    remove(key: string): void {\n      mmkvStorage.remove(key)\n    },\n  },\n})\n"]}

package/build/services/uiShim.d.ts

@@ -0,0 +1,38 @@
+/**
+ * window.glassly — WebView-side bridge to the per-miniapp background JSContext.
+ *
+ * Injected into every UI WebView via `injectedJavaScriptBeforeContentLoaded`
+ * (iOS) or `injectedJavaScriptObject` + a bundler-emitted inline `<script>`
+ * (Android, where the `BeforeContentLoaded` prop is documented unreliable
+ * per react-native-webview issues #1099 / #1609).
+ *
+ * The shim:
+ *   - Buffers `glassly.send(channel, payload)` calls until `glassly.ready()`
+ *     fires + the host acks the channel is open. WebView-side BUFFERS (the
+ *     WebView is the short-lived side and shouldn't drop user input).
+ *   - Installs `window.__glassly.recv(envelope)` for the host to push
+ *     background → WebView messages via `webview.injectJavaScript(...)`.
+ *   - Sends `__heartbeat__` every 5s; host considers the WebView gone
+ *     after 15s of silence.
+ *   - Maintains monotonic seq numbers; the host's per-WebView dedup window
+ *     drops replays during reconnect.
+ *
+ * This file exports a `buildUiShim()` function returning the JS
+ * source as a string. The host's UIRouter wraps it inside the
+ * existing WebView setup so a single bundler-time string emission covers
+ * both platforms.
+ *
+ * IMPORTANT: the shim runs in the WebView's bare ECMAScript scope before
+ * the miniapp's bundle. It must not import anything; the source is one
+ * IIFE. The only injected dependencies are
+ * `window.ReactNativeWebView.postMessage` (provided by react-native-webview)
+ * and (on Android) `window.ReactNativeWebView.injectedObjectJson()`
+ * (provided when the host sets `injectedJavaScriptObject`).
+ */
+export interface UiShimOptions {
+    packageName: string;
+    /** @deprecated No longer used — heartbeat removed since WebViews are foreground-only. */
+    heartbeatIntervalMs?: number;
+}
+export declare function buildUiShim(options: UiShimOptions): string;
+//# sourceMappingURL=uiShim.d.ts.map

package/build/services/uiShim.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"uiShim.d.ts","sourceRoot":"","sources":["../../src/services/uiShim.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AAEH,MAAM,WAAW,aAAa;IAC5B,WAAW,EAAE,MAAM,CAAA;IACnB,yFAAyF;IACzF,mBAAmB,CAAC,EAAE,MAAM,CAAA;CAC7B;AAED,wBAAgB,WAAW,CAAC,OAAO,EAAE,aAAa,GAAG,MAAM,CAgV1D"}