@xopcai/xopc 0.0.278 → 0.0.282

package/dist/src/discussions/repository.js

@@ -1,5 +1,5 @@
 import { getSqliteDatabase, runSqliteWriteTransaction } from "../storage/sqlite/transaction.js";
-import { DISCUSSION_STATUSES } from "./types.js";
+import { DISCUSSION_STATUSES } from "../../packages/gateway-contract/src/discussions.js";
 import { randomUUID } from "node:crypto";
 //#region src/discussions/repository.ts
 function optional(condition, value) {
@@ -17,6 +17,7 @@ function discussionFromRow(row) {
 		createdAt: row.created_at,
 		updatedAt: row.updated_at
 	};
+	capture.template = row.organization_template;
 	if (row.project_id) capture.projectId = row.project_id;
 	if (row.audio_attachment_id) capture.audioAttachmentId = row.audio_attachment_id;
 	if (row.duration_ms != null) capture.durationMs = row.duration_ms;
@@ -71,6 +72,7 @@ function organizationFromRow(row) {
 		revision: row.revision,
 		inputTranscriptSha256: row.input_transcript_sha256,
 		promptVersion: row.prompt_version,
+		transcriptRevision: row.transcript_revision,
 		modelRef: row.model_ref,
 		...optional(row.organization_json, { organization: JSON.parse(row.organization_json) }),
 		status: row.status,
@@ -161,7 +163,7 @@ function updateDiscussionCapture(id, patch, expectedStatuses) {
     mime_type=?, audio_size_bytes=?, audio_sha256=?, canonical_transcript=?, canonical_transcript_sha256=?,
     transcript_language=?, transcript_revision=?, generated_title=?, project_inference_score=?,
     project_inference_source=?, failure_stage=?, failure_code=?, failure_message=?, recording_started_at=?,
-    recording_stopped_at=?, updated_at=?, completed_at=?, audio_deleted_at=? WHERE id=?${statusGuard}`).run(next.projectId ?? null, next.audioAttachmentId ?? null, next.source, next.status, next.durationMs ?? null, next.expectedLastSequence ?? null, next.mimeType ?? null, next.audioSizeBytes ?? null, next.audioSha256 ?? null, next.canonicalTranscript ?? null, next.canonicalTranscriptSha256 ?? null, next.transcriptLanguage ?? null, next.transcriptRevision, next.generatedTitle ?? null, next.projectInferenceScore ?? null, next.projectInferenceSource ?? null, next.failureStage ?? null, next.failureCode ?? null, next.failureMessage ?? null, next.recordingStartedAt, next.recordingStoppedAt ?? null, next.updatedAt, next.completedAt ?? null, next.audioDeletedAt ?? null, id, ...expectedStatuses ?? []).changes) ? getDiscussionCapture(id) : null;
+    recording_stopped_at=?, updated_at=?, completed_at=?, audio_deleted_at=?, organization_template=? WHERE id=?${statusGuard}`).run(next.projectId ?? null, next.audioAttachmentId ?? null, next.source, next.status, next.durationMs ?? null, next.expectedLastSequence ?? null, next.mimeType ?? null, next.audioSizeBytes ?? null, next.audioSha256 ?? null, next.canonicalTranscript ?? null, next.canonicalTranscriptSha256 ?? null, next.transcriptLanguage ?? null, next.transcriptRevision, next.generatedTitle ?? null, next.projectInferenceScore ?? null, next.projectInferenceSource ?? null, next.failureStage ?? null, next.failureCode ?? null, next.failureMessage ?? null, next.recordingStartedAt, next.recordingStoppedAt ?? null, next.updatedAt, next.completedAt ?? null, next.audioDeletedAt ?? null, next.template ?? "general", id, ...expectedStatuses ?? []).changes) ? getDiscussionCapture(id) : null;
 }
 function claimNextDiscussionCapture(owner, now = Date.now(), leaseMs = 5 * 6e4) {
 	return runSqliteWriteTransaction((db) => {
@@ -173,8 +175,8 @@ function claimNextDiscussionCapture(owner, now = Date.now(), leaseMs = 5 * 6e4)
       WHERE id=? AND status='organizing' AND (work_lease_owner IS NULL OR work_lease_expires_at <= ?)`).run(owner, now + leaseMs, now, row.id, now).changes ? getDiscussionCapture(row.id) : null;
 	});
 }
-function releaseDiscussionWorkLease(id) {
-	runSqliteWriteTransaction((db) => db.prepare("UPDATE discussion_captures SET work_lease_owner=NULL, work_lease_expires_at=NULL WHERE id=?").run(id));
+function releaseDiscussionWorkLease(id, owner) {
+	runSqliteWriteTransaction((db) => db.prepare("UPDATE discussion_captures SET work_lease_owner=NULL, work_lease_expires_at=NULL WHERE id=? AND work_lease_owner=?").run(id, owner));
 }
 function createDiscussionTranscriptSegment(input) {
 	const existing = getDiscussionTranscriptSegment(input.discussionId, input.sequence);
@@ -228,12 +230,12 @@ function completeDiscussionTranscriptSegment(discussionId, sequence, owner, text
 		return changed;
 	}) ? getDiscussionTranscriptSegment(discussionId, sequence) : null;
 }
-function correctDiscussionTranscriptSegment(discussionId, sequence, displayText, expectedRevision) {
+function correctDiscussionTranscriptSegment(discussionId, sequence, displayText, expectedRevision, speakerLabel) {
 	const now = Date.now();
 	return runSqliteWriteTransaction((db) => {
-		const changed = db.prepare(`UPDATE discussion_transcript_segments SET display_text=?, corrected_by_user=1,
+		const changed = db.prepare(`UPDATE discussion_transcript_segments SET display_text=?, speaker_label=CASE WHEN ? THEN ? ELSE speaker_label END, corrected_by_user=1,
       corrected_at=?, revision=revision+1, updated_at=? WHERE discussion_id=? AND sequence=?
-      AND status='confirmed' AND revision=?`).run(displayText, now, now, discussionId, sequence, expectedRevision).changes;
+      AND status='confirmed' AND revision=?`).run(displayText, speakerLabel === void 0 ? 0 : 1, speakerLabel?.trim() || null, now, now, discussionId, sequence, expectedRevision).changes;
 		if (changed) db.prepare("UPDATE discussion_captures SET transcript_revision=transcript_revision+1, updated_at=? WHERE id=?").run(now, discussionId);
 		return changed;
 	}) ? getDiscussionTranscriptSegment(discussionId, sequence) : null;
@@ -253,8 +255,8 @@ function createDiscussionOrganization(input) {
 	const revision = (getLatestDiscussionOrganization(input.discussionId)?.revision ?? 0) + 1;
 	const id = randomUUID();
 	runSqliteWriteTransaction((db) => db.prepare(`INSERT INTO discussion_organizations (
-    id, discussion_id, revision, input_transcript_sha256, prompt_version, model_ref, status, created_at
-  ) VALUES (?, ?, ?, ?, ?, ?, 'running', ?)`).run(id, input.discussionId, revision, input.inputTranscriptSha256, input.promptVersion, input.modelRef, now));
+    id, discussion_id, revision, input_transcript_sha256, prompt_version, model_ref, transcript_revision, status, created_at
+  ) VALUES (?, ?, ?, ?, ?, ?, ?, 'running', ?)`).run(id, input.discussionId, revision, input.inputTranscriptSha256, input.promptVersion, input.modelRef, input.transcriptRevision, now));
 	return getLatestDiscussionOrganization(input.discussionId);
 }
 function getLatestDiscussionOrganization(discussionId) {
@@ -285,5 +287,8 @@ function acknowledgeDiscussionCaptureConsent(policyVersion) {
     SET consent_acknowledged_at=?, updated_at=? WHERE workspace_id='default' AND consent_policy_version=?`).run(now, now, policyVersion));
 	return getDiscussionCaptureSettings();
 }
+function renewDiscussionWorkLease(id, owner) {
+	return Number(getSqliteDatabase().prepare("UPDATE discussion_captures SET work_lease_expires_at=? WHERE id=? AND work_lease_owner=? AND status='organizing'").run(Date.now() + 5 * 6e4, id, owner).changes) === 1;
+}
 //#endregion
-export { acknowledgeDiscussionCaptureConsent, claimNextDiscussionCapture, claimNextDiscussionTranscriptSegment, completeDiscussionOrganization, completeDiscussionTranscriptSegment, correctDiscussionTranscriptSegment, createDiscussionCapture, createDiscussionOrganization, createDiscussionTranscriptSegment, deleteDiscussionSegmentAudio, failDiscussionOrganization, getDiscussionCapture, getDiscussionCaptureByClientRequestId, getDiscussionCaptureByNoteId, getDiscussionCaptureSettings, getDiscussionMetrics, getDiscussionTranscriptSegment, getLatestDiscussionOrganization, listDiscussionCaptures, listDiscussionTranscriptSegments, releaseDiscussionWorkLease, retryOrFailDiscussionTranscriptSegment, updateDiscussionCapture };
+export { acknowledgeDiscussionCaptureConsent, claimNextDiscussionCapture, claimNextDiscussionTranscriptSegment, completeDiscussionOrganization, completeDiscussionTranscriptSegment, correctDiscussionTranscriptSegment, createDiscussionCapture, createDiscussionOrganization, createDiscussionTranscriptSegment, deleteDiscussionSegmentAudio, failDiscussionOrganization, getDiscussionCapture, getDiscussionCaptureByClientRequestId, getDiscussionCaptureByNoteId, getDiscussionCaptureSettings, getDiscussionMetrics, getDiscussionTranscriptSegment, getLatestDiscussionOrganization, listDiscussionCaptures, listDiscussionTranscriptSegments, releaseDiscussionWorkLease, renewDiscussionWorkLease, retryOrFailDiscussionTranscriptSegment, updateDiscussionCapture };

package/dist/src/discussions/revisions.d.ts

@@ -0,0 +1,12 @@
+import type { DiscussionTranscriptSegment } from './types.js';
+export declare function saveTranscriptRevision(id: string): number;
+export declare function readTranscriptRevision(id: string, revision: number): DiscussionTranscriptSegment[] | null;
+export declare function replaceTranscriptSegments(id: string, segments: Array<{
+    text: string;
+    startedAtMs: number;
+    endedAtMs: number;
+    speakerLabel?: string;
+    correctedByUser?: boolean;
+    rawText?: string;
+}>): void;
+export declare function refreshCanonicalTranscript(id: string): void;

package/dist/src/discussions/revisions.js

@@ -0,0 +1,36 @@
+import { getSqliteDatabase, runSqliteWriteTransaction } from "../storage/sqlite/transaction.js";
+import { getDiscussionCapture, listDiscussionTranscriptSegments, updateDiscussionCapture } from "./repository.js";
+import { assembleDiscussionTranscript } from "./transcript.js";
+import { createHash } from "node:crypto";
+//#region src/discussions/revisions.ts
+function saveTranscriptRevision(id) {
+	const capture = getDiscussionCapture(id);
+	if (!capture) throw new Error("Discussion not found");
+	getSqliteDatabase().prepare("INSERT OR IGNORE INTO discussion_transcript_revisions (discussion_id, revision, segments_json, created_at) VALUES (?, ?, ?, ?)").run(id, capture.transcriptRevision, JSON.stringify(listDiscussionTranscriptSegments(id)), Date.now());
+	return capture.transcriptRevision;
+}
+function readTranscriptRevision(id, revision) {
+	const row = getSqliteDatabase().prepare("SELECT segments_json FROM discussion_transcript_revisions WHERE discussion_id=? AND revision=?").get(id, revision);
+	return row ? JSON.parse(row.segments_json) : null;
+}
+function replaceTranscriptSegments(id, segments) {
+	runSqliteWriteTransaction((db) => {
+		saveTranscriptRevision(id);
+		db.prepare("DELETE FROM discussion_transcript_segments WHERE discussion_id=?").run(id);
+		const insert = db.prepare(`INSERT INTO discussion_transcript_segments
+      (discussion_id, sequence, audio_sha256, started_at_ms, ended_at_ms, status, raw_text, display_text, speaker_label, revision, corrected_by_user, attempt_count, created_at, updated_at)
+      VALUES (?, ?, ?, ?, ?, 'confirmed', ?, ?, ?, 1, ?, 0, ?, ?)`);
+		segments.forEach((segment, index) => insert.run(id, index, createHash("sha256").update(segment.text).digest("hex"), segment.startedAtMs, segment.endedAtMs, segment.rawText ?? segment.text, segment.text, segment.speakerLabel ?? null, segment.correctedByUser ? 1 : 0, Date.now(), Date.now()));
+		db.prepare("UPDATE discussion_captures SET transcript_revision=transcript_revision+1 WHERE id=?").run(id);
+		saveTranscriptRevision(id);
+	});
+}
+function refreshCanonicalTranscript(id) {
+	const text = assembleDiscussionTranscript(listDiscussionTranscriptSegments(id));
+	updateDiscussionCapture(id, {
+		canonicalTranscript: text,
+		canonicalTranscriptSha256: createHash("sha256").update(text).digest("hex")
+	});
+}
+//#endregion
+export { readTranscriptRevision, refreshCanonicalTranscript, replaceTranscriptSegments, saveTranscriptRevision };

package/dist/src/discussions/sealer.d.ts

@@ -2,6 +2,7 @@ import type { Config } from '../config/schema.js';
 import type { NotesService } from '../notes/service.js';
 import type { DiscussionCapture } from './types.js';
 export interface DiscussionSealerDeps {
+    processRecordingJob?: () => Promise<void>;
     notes: NotesService;
     getConfig: () => Config;
     transcribeRecording?: (filePath: string, capture: DiscussionCapture, signal?: AbortSignal) => Promise<{

package/dist/src/discussions/sealer.js

@@ -1,12 +1,16 @@
 import { createLogger } from "../utils/logger/index.js";
 import { init_logger } from "../utils/logger.js";
+import { decodeWavToMonoFloat32 } from "../voice/local/wav.js";
 import { forEachNormalizedAudioSegment } from "../voice/audio/normalize.js";
-import { isSTTAvailable } from "../voice/stt/availability.js";
 import { transcribe } from "../voice/stt/transcribe-core.js";
+import { isSTTAvailable } from "../voice/stt/availability.js";
 import { mergeSttConfigFromAppConfig } from "../voice/stt/config.js";
 import "../voice/stt/index.js";
 import { getDiscussionCapture, listDiscussionCaptures, listDiscussionTranscriptSegments, updateDiscussionCapture } from "./repository.js";
-import { appendSequentialTranscript, assembleDiscussionTranscript } from "./transcript.js";
+import { assembleDiscussionTranscript } from "./transcript.js";
+import { replaceTranscriptSegments, saveTranscriptRevision } from "./revisions.js";
+import { missingAudioRanges, slicePcmWav } from "./audio-repair.js";
+import { getRecordingJob } from "./recordingJobs.js";
 import { createHash } from "node:crypto";
 //#region src/discussions/sealer.ts
 init_logger();
@@ -34,6 +38,7 @@ var DiscussionSealer = class {
 		if (this.running) return;
 		this.running = true;
 		try {
+			await this.deps.processRecordingJob?.();
 			const capture = [...listDiscussionCaptures({
 				status: "sealing",
 				limit: 1
@@ -50,6 +55,8 @@ var DiscussionSealer = class {
 	async sealWhenReady(capture, now) {
 		const ageMs = now - (capture.recordingStoppedAt ?? now);
 		if (!capture.audioAttachmentId) {
+			const job = getRecordingJob(capture.id);
+			if (job && (job.state === "queued" || job.state === "running")) return;
 			if (ageMs >= UPLOAD_GRACE_MS) this.fail(capture, "audio_upload", "recording_missing", "Original recording was not uploaded");
 			return;
 		}
@@ -65,6 +72,7 @@ var DiscussionSealer = class {
 			const confirmed = segments.filter((segment) => segment.status === "confirmed");
 			const result = expected >= 0 && missing.length === 0 && !pending && confirmed.length === expected + 1 ? { text: assembleDiscussionTranscript(confirmed) } : await this.transcribeFullRecording(sealing);
 			const text = result.text.trim();
+			saveTranscriptRevision(capture.id);
 			if (!text) throw new Error("Discussion transcription produced no text");
 			const updated = updateDiscussionCapture(capture.id, {
 				status: "organizing",
@@ -85,22 +93,64 @@ var DiscussionSealer = class {
 	async transcribeFullRecording(capture) {
 		const attachment = await this.deps.notes.getAttachmentPath(capture.noteId, capture.audioAttachmentId);
 		if (!attachment) throw new Error("Discussion audio file is missing");
-		if (this.deps.transcribeRecording) return this.deps.transcribeRecording(attachment.filePath, capture);
+		if (this.deps.transcribeRecording) {
+			const result = await this.deps.transcribeRecording(attachment.filePath, capture);
+			replaceTranscriptSegments(capture.id, [{
+				text: result.text,
+				startedAtMs: 0,
+				endedAtMs: capture.durationMs ?? 1e3
+			}]);
+			return result;
+		}
 		const config = this.deps.getConfig();
 		const sttConfig = mergeSttConfigFromAppConfig(config.tools?.media?.audio, config.tools?.media);
 		if (!isSTTAvailable(sttConfig)) throw new Error("STT is not configured");
-		let text = "";
+		let offset = 0;
+		const confirmed = listDiscussionTranscriptSegments(capture.id).filter((segment) => segment.status === "confirmed");
+		const segments = confirmed.map((segment) => ({
+			text: segment.displayText ?? segment.rawText ?? "",
+			rawText: segment.rawText,
+			startedAtMs: segment.startedAtMs,
+			endedAtMs: segment.endedAtMs,
+			speakerLabel: segment.speakerLabel,
+			correctedByUser: segment.correctedByUser
+		}));
 		let language;
-		await forEachNormalizedAudioSegment({ filePath: attachment.filePath }, async (buffer, index) => {
-			const result = await transcribe(buffer, sttConfig, {
-				mime: "audio/wav",
-				fileName: `discussion-recovery-${index}.wav`
-			});
-			text = appendSequentialTranscript(text, result.text);
-			language ??= result.language;
+		await forEachNormalizedAudioSegment({
+			filePath: attachment.filePath,
+			maxDurationSeconds: 7200
+		}, async (buffer, index) => {
+			const decoded = decodeWavToMonoFloat32(buffer);
+			const durationMs = decoded.durationSeconds * 1e3;
+			const gaps = missingAudioRanges(offset, offset + durationMs, confirmed);
+			for (const [gapIndex, gap] of gaps.entries()) {
+				const audio = gaps.length === 1 && gap.startedAtMs === offset && gap.endedAtMs === offset + durationMs ? buffer : slicePcmWav(decoded, gap.startedAtMs - offset, gap.endedAtMs - offset);
+				if (audio.length <= 44) continue;
+				const result = await transcribe(audio, sttConfig, {
+					mime: "audio/wav",
+					fileName: `discussion-recovery-${index}-${gapIndex}.wav`
+				});
+				if (result.segments?.length) for (const part of result.segments) {
+					if (!Number.isFinite(part.startMs) || !Number.isFinite(part.endMs) || part.startMs < 0 || part.endMs <= part.startMs || part.endMs > gap.endedAtMs - gap.startedAtMs + 100) throw new Error("Provider returned an invalid transcript time range");
+					segments.push({
+						text: part.text,
+						startedAtMs: gap.startedAtMs + part.startMs,
+						endedAtMs: Math.min(gap.endedAtMs, gap.startedAtMs + part.endMs),
+						...part.speaker ? { speakerLabel: `${index + 1}:${gapIndex + 1}:${part.speaker}` } : {}
+					});
+				}
+				else if (result.text.trim()) segments.push({
+					text: result.text,
+					...gap
+				});
+				language ??= result.language;
+			}
+			offset += durationMs;
 		});
+		segments.sort((a, b) => a.startedAtMs - b.startedAtMs);
+		replaceTranscriptSegments(capture.id, segments);
 		return {
-			text,
+			text: assembleDiscussionTranscript(listDiscussionTranscriptSegments(capture.id)),
 			...language ? { language } : {}
 		};
 	}

package/dist/src/discussions/service.d.ts

@@ -1,13 +1,10 @@
+import { type DiscussionRecordingManifest } from '@xopcai/gateway-contract';
 import type { NotesService } from '../notes/service.js';
 import type { ProjectService } from '../projects/project-service.js';
 import type { CreateDiscussionInput, DiscussionCaptureSettings, DiscussionDetail, DiscussionListResult, DiscussionMetrics, DiscussionTranscript, ListDiscussionsQuery } from './types.js';
-export declare const DISCUSSION_AUDIO_MAX_BYTES: number;
-export declare const DISCUSSION_SEGMENT_MAX_BYTES: number;
-export declare const DISCUSSION_MAX_DURATION_MS: number;
-export declare class DiscussionServiceError extends Error {
-    readonly code: 'invalid_input' | 'not_found' | 'conflict';
-    constructor(code: 'invalid_input' | 'not_found' | 'conflict', message: string);
-}
+import { editMeeting } from './edits.js';
+export { DiscussionServiceError } from './errors.js';
+export { DISCUSSION_MAX_DURATION_MS, DISCUSSION_AUDIO_MAX_BYTES, DISCUSSION_SEGMENT_MAX_BYTES } from '@xopcai/gateway-contract';
 export declare class DiscussionService {
     private readonly notes;
     private readonly projects;
@@ -15,6 +12,8 @@ export declare class DiscussionService {
     private readonly createsInFlight;
     private readonly mutationTails;
     private readonly objectLinks;
+    private readonly recordingOwner;
+    private readonly recordingControllers;
     constructor(notes: NotesService, projects: ProjectService, emit?: (type: string, payload: unknown) => void);
     settings(): DiscussionCaptureSettings;
     acknowledgeConsent(policyVersion: number): DiscussionCaptureSettings;
@@ -24,7 +23,7 @@ export declare class DiscussionService {
     getByNoteId(noteId: string): Promise<DiscussionDetail | null>;
     list(query?: ListDiscussionsQuery): DiscussionListResult;
     metrics(): DiscussionMetrics;
-    transcript(id: string): DiscussionTranscript | null;
+    transcript(id: string, revision?: number): DiscussionTranscript | null;
     uploadSegment(input: {
         discussionId: string;
         sequence: number;
@@ -36,13 +35,15 @@ export declare class DiscussionService {
         endedAtMs: number;
         sha256: string;
     }): DiscussionTranscript;
-    correctSegment(id: string, sequence: number, displayText: string, expectedRevision: number): DiscussionTranscript;
-    uploadRecording(id: string, file: {
-        name: string;
-        buffer: Buffer;
-        mimeType: string;
-    }, durationMs: number): Promise<DiscussionDetail | null>;
-    stop(id: string, lastSequence: number, durationMs: number): Promise<DiscussionDetail | null>;
+    correctSegment(id: string, sequence: number, displayText: string, expectedRevision: number, speakerLabel?: string, applyToSpeakerGroup?: boolean, expectedTranscriptRevision?: number): DiscussionTranscript;
+    editSummary(id: string, input: Parameters<typeof editMeeting>[1]): import("@xopcai/gateway-contract").DiscussionOrganizationRecord;
+    recordingChunks(id: string): import("@xopcai/gateway-contract").DiscussionRecordingChunk[];
+    reorganize(id: string, template?: string): Promise<DiscussionDetail>;
+    uploadRecordingChunk(id: string, sequence: number, sha256: string, body: ReadableStream<Uint8Array>): Promise<import("@xopcai/gateway-contract").DiscussionRecordingChunk>;
+    recordingJob(id: string): import("@xopcai/gateway-contract").DiscussionRecordingJob;
+    sealRecording(id: string, input: DiscussionRecordingManifest): Promise<import("@xopcai/gateway-contract").DiscussionRecordingJob>;
+    /** Runs inside the existing sealer lifecycle; unfinished leases survive process restart. */
+    processRecordingJob(): Promise<void>;
     retry(id: string): Promise<DiscussionDetail | null>;
     cancel(id: string): Promise<DiscussionDetail | null>;
     deleteAudio(id: string): Promise<DiscussionDetail | null>;