TimeTracker/docs/telemetry-architecture.md

Telemetry Architecture

This document describes the privacy-aware, two-layer telemetry system: base telemetry (always-on, minimal) and detailed analytics (opt-in only).

Overview

Layer	When	Purpose	Events / Data
Base telemetry	Always (when OTLP sink is configured)	Install footprint, version/platform distribution, active installs	base_telemetry.first_seen, base_telemetry.heartbeat
Detailed analytics	Only when user opts in	Feature usage, funnels, errors, retention	All product events (e.g. auth.login, timer.started)

• Consent: Stored in installation.json (telemetry_enabled) and synced to settings.allow_analytics. Source of truth: installation_config.get_telemetry_preference() / is_telemetry_enabled().
• Identifiers: One install_id (random UUID in installation config) used for base telemetry and, when opt-in, sent with product events. Product events use internal user_id identity.

Base Telemetry (Always-On)

• Schema (no PII): install_id, app_version, platform, os_version, architecture, locale, timezone, first_seen_at, last_seen_at, heartbeat_at, release_channel, deployment_type.
• Events: base_telemetry.first_seen (once per install), base_telemetry.heartbeat (e.g. daily via scheduler).
• Sink: Grafana Cloud OTLP with identity = install_id. No user-level linkage.
• Trigger: First-seen sent at app startup (idempotent). Heartbeat via scheduled task (e.g. 03:00 daily).
• Retention: Configure in Grafana backend (e.g. 12 months for base). No raw IP storage.

Detailed Analytics (Opt-In Only)

• Gated by: is_telemetry_enabled() / allow_analytics. No product events sent without opt-in.
• Events: Existing names (e.g. auth.login, timer.started, project.created). Support funnel events use the support.* prefix (e.g. support.modal_opened); see all_tracked_events.md. Optional prefix analytics.* in future.
• Properties: Include install_id, app_version, deployment, request context (path, browser, device) only when opted in.
• Sink: Grafana Cloud OTLP (identity = user_id for events).
• Retention: Per Grafana retention policy. Document in privacy policy.

Consent Behavior

• Opt-in: Setup wizard or Admin → Settings (Privacy & Analytics) or Admin → Telemetry. Enabling triggers one opt-in install ping (check_and_send_telemetry()).
• Opt-out: Same toggles. Detailed analytics stop immediately; base telemetry continues (minimal footprint).
• Data minimization: Base layer is fixed schema. Detailed layer only when user agrees.

Event Naming

• Reserved: base_telemetry.* for base layer. Do not use for product events.
• Product events: Keep current names (e.g. timer.started) or use analytics.*; all gated by opt-in.

Implementation

• Service: app/telemetry/service.py — send_base_first_seen(), send_base_heartbeat(), send_analytics_event(), is_detailed_analytics_enabled().
• App entry points: app/__init__.py — track_event, track_page_view, identify_user delegate to telemetry service (consent-aware).
• Scheduler: app/utils/scheduled_tasks.py — job send_base_telemetry_heartbeat (daily).
• Startup: In create_app, after scheduler start, call send_base_first_seen() once per install.

Sink Configuration

• Base and detailed telemetry are emitted through the same OTLP sender in app/telemetry/service.py.
• Required configuration:
  • GRAFANA_OTLP_ENDPOINT
  • GRAFANA_OTLP_TOKEN