Mi Jornada

Goal

Log working days and automatically calculate the time worked and the amount earned for each entry.

Problem it solves

Tracking hours and pay by hand (paper, notes or spreadsheets) is slow and error-prone, especially with split shifts, lunch breaks or several workplaces.

Architecture diagram

Features

• —Email login; the session persists across reloads.
• —Clock in and out with the exact time, with a lunch break that can be discarded; shifts that cross midnight are flagged for review.
• —Entries with date, start, end, place, description and lunch break; duration and amount are calculated automatically, keeping the hourly wage of each entry.
• —Time validation (the end must be after the start).
• —Short form with Today/Yesterday, suggested places, repeat last day and duplicate a day.
• —List filterable by month and place, with CSV export compatible with Spanish-locale Excel.
• —Monthly calendar with each day's hours.
• —Monthly stats: hours, amount, daily average, chart by day and hours by place.
• —AI "Type it or dictate it" assistant that turns sentences like "yesterday from 8 to 2 and from 4 to 7 at the site" into entries; nothing is saved until the user confirms.
• —Password recovery by email.
• —Light and dark mode; layout for mobile and desktop.
• —Admin panel with each user's hourly wage.
• —Public demo account with fictional data that regenerates itself.

Technical challenges

1. 01Django and Vue on VercelThe frontend is built and Django serves it as static files inside a Vercel Function.
2. 02Exhausted connections in productionThe Supabase pooler allows 15 clients and persistent connections filled it up; it was fixed by closing the connection on each request and releasing it while the AI responds.
3. 03Reliable clock-inMadrid time zone, at most one active shift per user and handling of shifts that cross midnight.
4. 04Useful AI on a free planStructured output validated on the server, mandatory confirmation, fallback models when the main one is overloaded and a daily usage limit.

Current status

In production. In September 2026 it was migrated from PythonAnywhere to Vercel and Supabase. The backend has 22 automated tests. There is no public sign-up, but it can be tried with a demo account with fictional data that resets itself.

Future improvements

• —Public sign-up so anyone can create their own account.
• —Let each user change their hourly wage from their profile.