Analytics - Hyprnote Documentation

What are we tracking?

We use PostHog for analytics in both the web and desktop applications. The web application also uses Outlit for pageview tracking (configured via the VITE_OUTLIT_PUBLIC_KEY environment variable). You can opt-out from the settings menu in the desktop app.

User Journey Events

1. Visiting the Website

Events tracked on the web app (apps/web):

Event	Description	Properties
`download_clicked`	User clicks a download button	`platform`, `spec`, `source`, `timestamp`
`reminder_requested`	User requests a mobile reminder	`platform`, `timestamp`, `email`
`os_waitlist_joined`	User joins waitlist for Linux/Windows	`platform`, `timestamp`, `email`

The web app also has PostHog autocapture and pageview tracking enabled.

2. Getting Started

Events tracked when users first launch the desktop app:

Event	Description	Properties
`show_main_window`	Main window is shown (fires on every app launch)	-
`account_skipped`	User proceeds without account (local mode)	-
`user_signed_in`	User signs in (triggers `$identify` to link anonymous ID)	-
`ai_provider_configured`	User configures an AI provider	`provider`
`data_imported`	User imports data from another app	`source`
`trial_started`	User starts a trial	`plan`

3. Before Meetings

Events tracked when preparing for meetings:

Event	Description	Properties
`note_created`	User creates a new note	`has_event_id` (whether linked to calendar)
`file_uploaded`	User uploads a file	`file_type` (audio/transcript), `token_count`

4. During Meetings

Events tracked during active sessions:

Event	Description	Properties
`session_started`	Listening session starts	`has_calendar_event`
`tab_opened`	User opens a tab	`view` (tab type)
`search_performed`	User performs a search	-

5. After Meetings

Events tracked when working with completed sessions:

Event	Description	Properties
`note_edited`	User edits a note	`has_content`
`note_enhanced`	AI enhancement is triggered	`is_auto`, `template_id`
`message_sent`	User sends a chat message	-
`session_exported`	User exports a session	`format` (pdf/vtt), `view_type`, `has_transcript`, `has_enhanced`, `has_memo`, `word_count`
`session_deleted`	User deletes a session/note	`includes_recording`
`recording_deleted`	User deletes a recording	-

6. Settings & Account Management

Events tracked when managing settings and account:

Event	Description	Properties
`settings_changed`	User changes settings	`autostart`, `notification_detect`, `save_recordings`, `telemetry_consent`
`upgrade_clicked`	User clicks upgrade button	`plan`
`user_signed_out`	User signs out	-

7. Notification Interactions

Events tracked when users interact with system notifications:

Event	Description	Properties
`collapsed_confirm`	User clicks collapsed notification to open app	-
`expanded_accept`	User accepts expanded notification	-
`dismiss`	User dismisses notification	-
`collapsed_timeout`	Collapsed notification times out	-

User Properties

Properties set to track user context and configuration:

Property	Description	Type
`is_local_mode`	Whether user is in local-only mode	set
`is_signed_up`	Whether user has signed up	set
`platform`	Operating system (macos, linux, windows)	set
`os_version`	Operating system version	set
`app_version`	Application version	set
`account_created_date`	When account was created	set_once
`telemetry_opt_out`	Whether user opted out of telemetry	set
`has_configured_ai`	Whether user has configured an AI provider	set
`plan`	Current subscription plan	set
`trial_end_date`	When trial ends	set

Analytics Commands

Available methods for tracking events and properties:

event(payload) - Track a custom event with optional properties
setProperties(payload) - Set user properties (supports set and set_once types)
identify(userId, payload) - Link anonymous device ID to authenticated user ID using PostHog's $identify event
setDisabled(disabled) - Enable or disable analytics tracking
isDisabled() - Check if analytics is currently disabled

User Identification

When a user signs in, the identify command is called to link their anonymous device ID (machine fingerprint) to their authenticated user ID. This uses PostHog's $identify event with the $anon_distinct_id property, enabling:

Attribution of pre-login activity to the authenticated user
Unified user profiles across anonymous and authenticated sessions
Accurate conversion funnel analysis from first visit to active usage

The identification flow works as follows: before sign-in, events are tracked using the device fingerprint as the distinct ID. Upon successful authentication, identify(userId, payload) is called, which sends a $identify event to PostHog with both the new user ID and the anonymous device ID, merging the two identities.

Conversion Funnel Tracking

To track the user journey from website visit to active usage, use PostHog's funnel analysis with these key events:

download_clicked (web) - User downloads the app
show_main_window (desktop) - User launches the app
user_signed_in (desktop) - User signs in (triggers identity merge)
note_created (desktop) - User creates their first note
note_enhanced (desktop) - User uses AI enhancement

PostHog's $identify event links anonymous device IDs to authenticated user IDs, enabling accurate conversion analysis across the entire user journey.

Feature Flags

Hyprnote uses PostHog feature flags to control feature rollouts. The system is split into two layers:

crates/flag - A standalone Rust client for evaluating PostHog feature flags via the /flags API. It returns typed FlagValue results with support for boolean flags, multivariate variants, and JSON payloads.
plugins/flag - A Tauri plugin that wraps the flag client with caching and provides a FlagPluginExt trait. Each feature is defined in the Feature enum with a FlagStrategy:
- Debug - enabled only in debug builds
- Posthog(key) - evaluated via PostHog API with response caching
- Hardcoded(bool) - always returns the given value

To check a flag from Rust:

use hypr_plugin_flag::FlagPluginExt;

let enabled = app.flag().is_enabled(Feature::Chat).await;

Crate Architecture

The analytics stack is split across several crates:

crates/posthog - Low-level PostHog HTTP client for sending events, setting properties, and identifying users.
crates/analytics - Higher-level client that composes PostHog and Outlit clients via AnalyticsClientBuilder. Used by both the desktop app and the AI server.
crates/flag - PostHog feature flag evaluation client (see above).

Server-Side Analytics

The apps/ai server also reports analytics for proxy requests:

LLM proxy (crates/llm-proxy) - Reports $ai_generation events with model, token counts, latency, and cost.
STT proxy (crates/transcribe-proxy) - Reports $stt_request events with provider and duration.

Both proxies attribute events to users via device fingerprint and optional user_id.

How to make changes?

See .cursor/commands/add-analytics.md for detailed instructions.