Data & Persistence

Session Data Model

The core data type is TranscriptSession:

Field	Type	Description
id	string	Unique identifier
schemaVersion	number	Currently 3
title	string	Display title
date / time	string	YYYY-MM-DD / HH:mm
createdAt / updatedAt	number	Unix milliseconds
transcript	string	Plain text transcript
translatedTranscript	TranscriptTranslationData	Optional translated text
duration	number	Duration in ms
tokens	TranscriptTokenData[]	Timestamped tokens
speakers	TranscriptSpeaker[]	Speaker labels
segments	TranscriptSegment[]	Timestamped segments
postProcess	TranscriptPostProcess	AI briefing output
askHistory	TranscriptAskTurn[]	Q&A conversation turns
mindMap	TranscriptMindMap	Mind map data
topicId	string	Associated topic
tagIds	string[]	Associated tags
providerId	string	ASR provider used
status	TranscriptSessionStatus	recording, interrupted, completed

Storage Architecture

┌─────────────────────────────────────────┐
│  sessionStore (Zustand)                  │
│  Runtime state + sessions array          │
│  ↕ read/write                            │
├──────────────────────────────────────────┤
│  sessionRepository (in-memory cache)     │
│  cachedSessions: TranscriptSession[]     │
│  ↕ async persist                         │
├──────────────────────────────────────────┤
│  sessionStorage (IndexedDB)              │
│  DB: 'delive-app', store: 'sessions'     │
│  Index: 'updatedAt' (sorted retrieval)   │
└──────────────────────────────────────────┘

IndexedDB Schema

Database delive-app at version 3 with four object stores:

Store	Key Path	Indexes	Contents
sessions	id	updatedAt	Transcript sessions
meta	key	—	Migration flags
settings	id	—	Settings mirror
tags	id	—	Tags mirror

localStorage Keys

Key	Contents
desktoplive_settings	App settings (authoritative source)
desktoplive_tags	Tags
desktoplive_topics	Topics
language	Interface language
theme	Light/dark mode preference

Settings and tags are mirrored to IndexedDB for redundancy. On startup, if localStorage is empty but IndexedDB has data, the values are restored from IDB.

INFO

Topics are stored only in localStorage and are not included in backup/export.

Session Lifecycle

Creation

sessionStore.startNewSession() creates a draft via sessionRepository.createDraft() with status recording, empty transcript, and metadata (timestamp, provider, source info).

Autosave

Every transcript event triggers scheduleCurrentSessionAutosave() with a 1200ms debounce. The autosave only persists when the snapshot has meaningful content (transcript, tokens, translated text, or post-process data).

Completion

sessionStore.endCurrentSession() calls sessionRepository.completeSession() which sets status: 'completed'. If the session has no content, the draft is deleted instead.

Recovery

On loadSessions(), any session with status: 'recording' is marked as interrupted and wasInterrupted: true. The first interrupted session with content becomes the recoverySession. Users can restore (load back into recording state) or dismiss it.

Settings

AppSettings includes:

Field	Type	Description
currentVendor	string	Selected ASR provider
providerConfigs	Record<string, ProviderConfigData>	Per-provider credentials
captionStyle	CaptionStyle	Caption overlay appearance
colorTheme	string	Accent palette (cyan, violet, etc.)
aiPostProcess	AiPostProcessConfig	AI endpoint configuration
openApi	OpenApiConfig	API enable/token settings
autoCheckUpdate	boolean	Auto-update preference

Backup & Import

backupStorage.ts handles data portability:

• Export: exportAllData() → JSON file with sessions, tags, and settings (version 2.0, schema version 2)
• Import overwrite: replaces all data (preserves existing API key if set)
• Import merge: adds only new sessions and tags by ID
• Normalization: validates and repairs data structure on import