Currently building

TrackHrs
Architecture

A distributed time tracking and activity intelligence platform — desktop agent, web apps, and event-driven microservices at scale.

User Interface Layer
Desktop Agent
Admin Portal
Product Services
Activity Engine
HR & Identity
Project Intel
Billing
Intelligence & Data
Event Bus
Cache
Persistence
The Challenge

Product Overview

TrackHrs is a full-stack time tracking and activity intelligence platform designed for distributed teams. It combines a desktop agent, web applications, and event-driven microservices to capture work activity, classify it, and present actionable analytics for organizations.

Accurate time tracking

Time tracked across tasks and projects with reliable capture and aggregation.

Activity intelligence

Activity monitoring with automated categorization for clearer operational insight.

Operational visibility

Dashboards and reporting for HR, managers, and operations teams.

Product Scope

Service Map

tracking/

Desktop Agent

Cross-platform capture

Activity capture, local runtime, and offline-friendly sync.

activity/

Activity Engine

Core Processing

Ingestion and classification orchestration.

activity-provider/

Analytics Provider

Data Delivery

Read APIs and statistics for dashboards and reporting.

activity_classifier/

AI Classifier

Intelligence Layer

ML/rule-based classification service.

user-service/

Core Domain Service

Identity & HR

Auth, users, organizations, leave, payroll core domain.

projects-manager/

Project Intelligence

Resource Tracking

Project/task time duration aggregation.

plans/

Billing & Subscriptions

Revenue Management

Subscription and billing domain.

notifications/

Notification Engine

Communication

Email and notification processing.

storage-producer/

Object Pipeline

Storage Layer

Object upload pipeline (S3-compatible).

auto-update/

Distribution Server

Release Management

Desktop release and update distribution.

trackhrs-front/

Marketing Platform

Public Presence

Public marketing and lead capture.

trackhrs-portal/

Operations Portal

Admin Experience

Authenticated operations and admin experience.

Data Flow

Activity Intelligence Pipeline

Step 1

Desktop activity event

Captured by the Tauri agent and sent securely to the platform.

Step 2

Kafka topic

High-throughput ingestion decoupled from downstream consumers.

Step 3

Activity consumer

Processes streams with partition-aware scaling.

Step 4

Redis dedup

O(1) deduplication for multi-replica-safe processing.

Step 5

Batch service

Batches of up to 100 items or 5s windows — fewer immediate writes.

Step 6

MongoDB bulk upsert

Document model fits evolving activity payloads and org structures.

Step 7

BullMQ classification

Background jobs separate latency-sensitive writes from heavier work.

Step 8

Classifier API

Bearer-key FastAPI service; ML iteration isolated from Node services.

Step 9

Classification update

Categories persisted; failures use circuit breaker, retry, DLQ.

Step 10

Cache invalidation

Fresh dashboard stats on next read without TTL-only dependency.

Rationale

Technology Decisions

Decouples ingestion from downstream processing; supports throughput and replay-friendly analytics workloads.
Classification and delayed jobs stay off the hot path; latency-sensitive writes stay fast.
Cache acceleration, deduplication, queue backing, and distributed locks for multi-replica-safe cron.
Flexible documents for activity payloads and evolving domains; pairs with batched upserts and indexed reporting.
Native-level performance and low overhead; better control for system-level tracking than browser-only approaches.
ML lifecycle and dependencies stay independent from TypeScript microservices.
Outcomes

Achievement Highlights

Production-style distributed platform

Rust desktop + modern web + microservices end-to-end.

Up to 100× fewer immediate DB writes

Batching on the ingest path reduces write amplification.

High-throughput async processing

Kafka and BullMQ absorb spikes and keep pipelines resilient.

Multi-replica-safe coordination

Redis dedup keys and distributed locks for cron and idempotent work.

Isolated ML service

Dedicated Python classifier; core platform stays strongly typed in TypeScript.

Full product surface

Desktop app, admin portal, marketing site, update server, and backend ecosystem.

Ready to build your next platform?

We specialize in high-scale distributed systems and AI-integrated workflows. Let's discuss your vision.

Get in Touch
View RTL