Build privacy controls for transcripts, audio, cache, data export, and deletion by charneykaye · Pull Request #124 · outrightmental/ConversationSimulator

charneykaye · 2026-07-08T03:02:53Z

Summary

This PR implements comprehensive privacy controls for ConversationSimulator by adding user preferences for transcript and data persistence, combined with bulk and per-session data management controls. It addresses issue #47 by building a full privacy-centric feature set.

Key Features

Privacy Preferences

Save transcripts: Toggle whether conversation transcripts are persisted (default: enabled)
TTS cache: Toggle whether generated speech is cached locally (default: enabled)
Raw audio: Advanced option to save unprocessed microphone recordings for debugging (default: disabled)
All preferences are persisted to localStorage and applied to new sessions

Backend Privacy Enforcement

When save_transcript=false, event records are skipped during session operation (start, turn, end, debrief)
Real-time events are still delivered in API responses using synthetic event IDs for immediate UI display
Export endpoint returns empty event arrays for sessions created with transcript saving disabled, enforcing the user's stated privacy preference

Data Management UI

Clear all local data: Single-click (with confirmation) deletion of all sessions, transcripts, and caches
Per-session export: Download any session as JSON for archival or analysis
Per-session delete: Individually delete specific sessions with confirmation
Data folder info: Display of the local data directory path for user transparency

API Additions

GET /api/privacy/data-folder — retrieves the path where all local data is stored
POST /api/privacy/clear — deletes all sessions and events from the database
GET /api/sessions/:session_id/export — exports a session and its events as JSON
DELETE /api/sessions/:session_id — deletes a specific session
Client API methods: getDataFolder(), clearLocalData(), exportSession(), deleteSession()

Implementation Details

Privacy preferences are read from localStorage on component mount with sensible defaults
The backend respects save_transcript at session creation time and consistently enforces it across all event-persisting endpoints (start, turn, end, debrief)
Session exports transparently include the save_transcript setting in the setup JSON so users can see whether their preference was honored
Comprehensive test coverage for all privacy endpoints and transcript enforcement

Testing

261 lines of API tests covering privacy routes (data folder, clear, export, delete)
454 lines of UI tests for Settings component (preferences, session management, error handling)
Tests verify that save_transcript=false sessions do not persist events but still deliver them in real-time responses
Tests confirm state machine correctness for clear-data operation (idle → confirming → clearing → done/error)
Tests validate error handling and recovery in session deletion and export flows

Closes #47

Implements issue #47: exposes privacy settings and backend actions so users can manage all locally-stored conversation data. Backend (apps/api): - Add GET /api/privacy/data-folder — returns the path where data is stored. - Add POST /api/privacy/clear — atomically deletes all sessions and events, returning the count of deleted sessions. - Add GET /api/sessions/:session_id/export — returns a full JSON export of a session (metadata, parsed setup/state-vars, and all events). - Wire setDataFolderPath() in index.ts so the live server reports the real data directory. - Add privacy.test.ts: 15 backend tests covering all three new endpoints, idempotency, cascade deletion, and export shape. Frontend (apps/web): - Extend api/client.ts: add deleteSession, exportSession, getDataFolder, and clearLocalData helpers; make post() body optional and add a del() helper. - Rewrite screens/Settings.tsx with a full Privacy Settings UI: - Prominent notice that no data is sent to external servers. - Save-transcripts toggle with plain-language explanation and live note (local-only / not-saved) that updates on toggle. - Cache-TTS-audio toggle. - Data-folder path display (loaded from API). - Two-step clear-all-local-data action: first click shows a confirmation alert; second click calls the API; result shows deleted count or error. - Advanced section (collapsed by default) revealing a raw-audio-saving toggle, which is off by default and shows a warning when enabled. - Add __tests__/Settings.test.tsx: 25 frontend tests covering all toggles, the local-only/not-saved notes, the data-folder display, the two-step clear flow (confirm, cancel, success, error, singular count), and the advanced raw-audio section. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

The Settings screen toggles (save transcripts, TTS cache, raw audio) were pure local React state — values reset on every navigation and the transcript toggle had no effect on the session setup form default. Extract privacyPrefs.ts with typed keys and read/write helpers guarded for environments where localStorage may be absent. Initialize each Settings toggle from localStorage and write back on change. Have ScenarioSetupPage read the persisted saveTranscripts preference as the initial save_transcript default. Add a localStorage stub in setupTests.ts (jsdom 25 dropped it) and clear it in Settings.test.tsx beforeEach to prevent inter-test state leakage. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

charneykaye · 2026-07-08T03:18:26Z