Workly - Personal AI Assistant banner

Workly - Personal AI Assistant

Created by: Xyon
3 devlogs
19h 48m 4s

Workly is a personal AI assistant with a modern web interface, deployed in production with automated tests and real-time monitoring.

Features:

  • FastAPI backend + LLM integration (GitHub Models)
  • JWT authentication (register, login, protecte…

Workly is a personal AI assistant with a modern web interface, deployed in production with automated tests and real-time monitoring.

Features:

  • FastAPI backend + LLM integration (GitHub Models)
  • JWT authentication (register, login, protected routes)
  • PostgreSQL cloud database (Supabase)
  • Modern interface with dark mode and animations
  • Automated tests (pytest + Selenium)
  • CI/CD pipeline (GitHub Actions)
  • Production deployment (Render + GitHub Pages)
  • UptimeRobot monitoring with live status badges

Tech Stack: Python, FastAPI, HTML/CSS/JS Vanilla, PostgreSQL, Selenium

This project uses AI

Development assisted by GitHub Copilot :

  • Architecture planning and task breakdown
  • Educational explanations of concepts
  • Debugging assistance
  • Code review and improvement suggestions
Demo Repository

README

Loading README...
Xyon
Xyon shipped Workly - Personal AI Assistant
about 1 month ago

Shipped this project!

Hours: 19.8
Cookies: 🍪 150
Multiplier: 7.58 cookies/hr

I built Workly, a personal AI assistant with a full authentication system — register, login, JWT tokens, and protected routes — all connected to a real PostgreSQL database in production.

The hardest part was understanding the full auth flow end-to-end: hashing passwords with bcrypt, signing JWT tokens, sending them in HTTP headers from the frontend, and verifying them on every protected request. Each layer depends on the previous one, so one mistake breaks everything.

I figured it out by building it step by step and writing tests at each stage — 8 automated tests (pytest + Selenium) that run on every push via GitHub Actions.

AI is something I’m genuinely passionate about — I love working at the intersection of building real software and making it intelligent. Workly is my way of exploring that hands-on.

Really proud of the fact that it’s not a tutorial project anymore. Real users can create accounts, log in, and chat with an AI — all secured, all deployed, all monitored.

Xyon
Xyon worked on Workly - Personal AI Assistant
about 1 month ago
4h 22m logged

Workly - Development Log #3
Date: March 2, 2026
Developer: Xyon
Status: v1.2.0-stable Released

What’s New

This update introduces a complete authentication system, a full frontend redesign, and a polished README — transforming Workly from a public demo into a real authenticated web application.

Key Features Added

JWT Authentication - Full register/login flow :

  • /register — Creates a user account with bcrypt-hashed password, unique username & email (PostgreSQL constraints)
  • /login — Returns a signed JWT token (access_token, token_type, username)
  • /me — Token verification endpoint used on every page load
  • /chat is now protected — requires a valid Bearer token, auto-redirect to login if expired

Frontend Redesign - Complete UI overhaul built around the auth flow:

  • New login and register pages
  • JWT stored in localStorage, sent as Authorization header on every request
  • Sidebar redesign with clean URL routing
  • Auto session verification on load with redirect logic

Email Validation - EmailStr enforced via Pydantic on /register with French-friendly error messages

GitHub Pages Fix - Added .nojekyll to prevent Jekyll from breaking frontend routing after the redesign

The Result

Workly is now a fully authenticated web application. Users can create an account, log in, and chat with the AI ​​— all secured with industry-standard JWT tokens and bcrypt password hashing. The codebase went from a public demo to a production-ready authenticated service.

Demo: https://xyon15.github.io/os-assistant/
GitHub: https://github.com/Xyon15/os-assistant
Release: https://github.com/Xyon15/os-assistant/releases/tag/v1.2.0-stable
License: PolyForm Noncommercial License 1.0.0
Version: 1.2.0-stable

Attachment
Attachment
0
Xyon
Xyon worked on Workly - Personal AI Assistant
2 months ago
3h 32m logged

OS Assistant - Development Log #2
Date: February 5, 2026
Developer: Xyon
Status: v1.1.0 Released

What’s New

This update focuses on production-grade monitoring, cloud database persistence, and CI/CD reliability improvements following the initial v1.0.0 release.

Key Features Added

Monitoring System - Complete observability with dedicated endpoints:

  • /health - Real-time health checks with uptime tracking
  • /metrics - Request counters and human-readable uptime
  • /stats - Database statistics (average response times, error rates)
  • Production logs persisted in cloud database with UTC+1 timestamps

PostgreSQL Migration - Moved from SQLite to cloud database:

  • Supabase PostgreSQL with Session Pooler (IPv4 compatibility)
  • Persistent logs survive backend redeployments
  • UptimeRobot monitoring keeps database and backend active 24/7

Testing Improvements - Enhanced CI/CD reliability:

  • Implemented unittest.mock for database-free testing in GitHub Actions
  • All 5 backend tests pass without DATABASE_URL environment variable
  • Fixed workflow to use requirements.txt for consistent dependencies

What I Learned

Database Challenges - Discovered ephemeral filesystems on Render delete SQLite files on every redeploy. Solution: migrate to cloud Postgres with always-on free tier (Supabase).

The Result

Production application with professional monitoring, persistent cloud storage, and robust CI/CD pipeline. Every feature built with full understanding of dependencies, testing strategies, and deployment considerations.

Demo: https://xyon15.github.io/os-assistant/
GitHub: https://github.com/Xyon15/os-assistant
Release: https://github.com/Xyon15/os-assistant/releases/tag/v1.1.0
License: GPL v3
Version: 1.1.0

Attachment
0
Xyon
Xyon worked on Workly - Personal AI Assistant
2 months ago
11h 53m logged

OS Assistant - Development Log #1

Date: February 3, 2026
Developer: Xyon
Status: v1.0.0-stable Released

The Journey

This project demonstrates building a complete AI assistant from scratch with full understanding of every component and deployment to production with professional development practices.

What I Built

OS Assistant is a personal AI chat application with a modern web interface. Users can have conversations with LLM through a clean, responsive UI with dark mode support. The entire system is deployed in production with automated testing and real-time monitoring.

The long-term goal is to evolve this into a desktop application that serves as an intelligent productivity companion, while maintaining simplicity and extensibility.

Tech Stack

Backend: Python 3.10+, FastAPI, Uvicorn, SQLite, GitHub Models API (GPT-4o)
Frontend: HTML5, CSS3 (Flexbox, animations, CSS variables), Vanilla JavaScript
Testing: pytest (backend), Selenium + ChromeDriver (UI tests), GitHub Actions (CI/CD)
Deployment: Render (backend), GitHub Pages (frontend), UptimeRobot (monitoring)

The Result

From zero knowledge to a production-ready application with professional development practices. I learned more building this than months of tutorials ever taught me. The project is live, tested, monitored, and ready for users.

Demo: https://xyon15.github.io/os-assistant/
GitHub: https://github.com/Xyon15/os-assistant
Release: https://github.com/Xyon15/os-assistant/releases/tag/v1.0.0-stable
License: GPL v3
Version: 1.0.0

Attachment
2

Comments

Foxreef
Foxreef 2 months ago

“Complete AI assistant from scratch”
Looks inside: ChatGPT wrapper

Xyon
Xyon 2 months ago

To clarify, the “from scratch” part refers to building the entire assistant system myself: backend architecture, frontend, API design, testing, CI/CD, deployment, and monitoring.

For the AI capabilities, the project currently relies on GPT-4o via GitHub Models, which is a deliberate choice for this stage of development in order to focus on shipping a stable, production-ready application and learning professional workflows.

A longer-term goal of the project is to experiment with custom or fine-tuned models trained for specific use cases, but that is outside the scope of the current version.