TimeTracker/ARCHITECTURE.md

TimeTracker Architecture

This document gives a high-level overview of the TimeTracker system for contributors and maintainers. For folder-level detail, see Project Structure. For migrating code to the service layer, see Architecture Migration Guide.

System Overview

TimeTracker is a self-hosted web application for time tracking, project management, invoicing, and reporting. The core is a Flask app serving both HTML (server-rendered) and a REST API. Optional components include background jobs (APScheduler), real-time updates (WebSocket via Flask-SocketIO), and monitoring (Prometheus, Sentry, PostHog). Deployment is typically Docker with Nginx as reverse proxy and PostgreSQL as the primary database.

flowchart LR
  subgraph client [Client]
    Browser[Browser]
    Mobile[Mobile App]
    Desktop[Desktop App]
  end
  subgraph server [Server]
    Nginx[Nginx]
    App[Flask App]
    DB[(PostgreSQL)]
  end
  Browser --> Nginx
  Mobile --> Nginx
  Desktop --> Nginx
  Nginx --> App
  App --> DB

Main Modules

Layer	Location	Role
Entry point	app.py	Creates Flask app, loads config, registers blueprints via blueprint_registry, starts server (and optional SocketIO/scheduler).
Blueprint registry	app/blueprint_registry.py	Single place that imports and registers all route blueprints so app/__init__.py stays manageable.
Routes	app/routes/	HTTP handlers: auth, main (dashboard), projects, timer, reports, admin, api, api_v1 (plus api_v1_* sub-blueprints), tasks, issues, invoices, clients, etc.
Services	app/services/	Business logic; routes call services instead of putting logic in view code.
Repositories	app/repositories/	Data access layer; services and routes use repositories for queries and eager loading.
Models	app/models/	SQLAlchemy ORM models (users, projects, time entries, tasks, clients, etc.).
Schemas	app/schemas/	Marshmallow schemas for API request/response validation and serialization.
Templates	app/templates/	Jinja2 HTML templates for server-rendered pages.
Utils	app/utils/	Helpers: timezone, validation, API responses, auth, setup_logging, legacy_migrations.
Config	app/config.py	Application configuration (env-based).
Desktop	desktop/	Electron-style desktop app (esbuild bundle) that talks to the API.
Mobile	mobile/	Flutter mobile app (iOS/Android) using the REST API.
Docker	docker/, root Dockerfile	Container build and runtime; optional Nginx, DB init scripts.
Tests	tests/	Pytest-based test suite (test_routes, test_services, test_models, test_utils, test_integration).

flowchart TB
  subgraph app [app/]
    Routes[routes/]
    Services[services/]
    Models[models/]
    Templates[templates/]
    Utils[utils/]
  end
  Routes --> Services
  Services --> Models
  Routes --> Templates
  Routes --> Utils
  Models --> DB[(Database)]

Data Flow

• Web request: User or browser → Nginx (if used) → Flask → blueprint in app/routes/ → optional service in app/services/ → repositories / models and DB → response (HTML or JSON).
• API request: Same path; API blueprints return JSON and use token auth. Request → route → service (or repository) → model/DB → api_responses helpers → JSON.
• Real-time: Flask-SocketIO is used for live timer updates; clients connect over WebSocket and receive events from the server.
• Background: APScheduler runs periodic tasks (e.g. scheduled reports, weekly summaries, remind-to-log end-of-day emails, reminders, cleanup) inside the app process. Report exports include time-entries PDF and summary-report PDF (app/utils/summary_report_pdf.py).

API endpoints are versioned under /api/v1/. Authentication is session-based for the web UI and API-token (Bearer or X-API-Key) for the API.

API Structure

• Base URL: /api/v1/
• Auth: API token in header Authorization: Bearer <token> or X-API-Key: <token>. Tokens are created in Admin → Api-tokens and have scopes (e.g. read:projects, write:time_entries).
• Sub-blueprints (all under /api/v1/): api_v1 (info, health, auth/login), api_v1_time_entries, api_v1_projects, api_v1_tasks, api_v1_clients, api_v1_invoices, api_v1_expenses, api_v1_payments, api_v1_mileage, api_v1_deals, api_v1_leads, api_v1_contacts, plus remaining routes in api_v1 (time-entry-approvals, per-diems, budget-alerts, calendar, kanban, saved-filters, etc.).
• Full reference: REST API.

Backend vs Frontend

• Backend: Flask (Python), Jinja2, SQLAlchemy, Flask-Migrate, Flask-Login, Authlib (OIDC), Flask-SocketIO, APScheduler. Configuration via environment variables (see env.example).
• Frontend: Server-rendered HTML from Jinja2, styled with Tailwind CSS. JavaScript is used for interactivity (e.g. Chart.js, command palette, forms). The app can be used as a PWA (offline and installable). There is no separate SPA; the main UI is server-rendered with JS enhancements.
• UI layer: The base layout is app/templates/base.html (sidebar, header, main content area). Styling uses Tailwind and design tokens in app/static/src/input.css. Reusable UI is built from component macros in app/templates/components/ui.html and app/templates/components/cards.html (page headers, stat cards, empty states, modals, buttons). Layout uses a max-width content container and consistent grid and spacing. See UI Guidelines and Project Structure for templates and static assets.
• Native clients: The desktop (Electron) and mobile (Flutter) apps are separate codebases that consume the REST API.

Design Decisions

• Service layer: Business logic lives in app/services/ so routes stay thin and logic is reusable and testable. See Service Layer and Base CRUD and the Architecture Migration Guide.
• API v1 split: Core resources (projects, tasks, clients, invoices, expenses, payments, mileage, deals, leads, contacts) are in separate sub-blueprints (api_v1_*.py) under /api/v1/ for maintainability; the main api_v1 module keeps info, health, auth, and remaining endpoints.
• Bootstrap: Logging is configured in app/utils/setup_logging.py; legacy migration helpers (task management, issues tables) are in app/utils/legacy_migrations.py. app/__init__.py creates the app and wires extensions.
• Blueprint registry: All blueprints are registered from app/blueprint_registry.py to keep registration in one place and simplify adding new modules.
• Database: PostgreSQL is recommended for production; SQLite is supported for development and testing (e.g. docker-compose.local-test.yml).
• API auth: The REST API uses API tokens (created in Admin → Api-tokens) with scopes; no session cookies for API access.
• Single codebase for web UI: No separate frontend repo; templates and static assets live in the main repo under app/templates/ and app/static/.

Further Reading

• Project Structure — Folder layout and file roles
• Architecture Migration Guide — Moving routes to the service layer
• REST API — API reference and authentication