Mail-minders: a zero-UI task assistant in your inbox

This is a submission for the Postmark Challenge: Inbox Innovators.

What I Built

Mail-minders is a fully email-based task reminder system — no apps, no dashboards, just your inbox. Users interact with Mail-minders entirely through email:

• Send an email with subject START to get onboarding instructions.
• Compose an email with the subject ADD, using the format provided in the onboarding email, to submit your task reminders.
• Mail-minders automatically reminds users of pending tasks.
• Users can mark tasks as complete or add new ones by replying to reminder emails.
• Send an email with the subject LIST to get a snapshot of all your current tasks.
• Send ANALYZE to receive a visual summary (bar and pie charts) of your categorized tasks.

TL;DR High Level Flow:

Demo

Try it out:

1. Just shoot out an email to [e-mail] with the subject START and wait for your instructions.
2. Follow the instructions in the onboarding email.
3. Submit tasks, wait for reminders, and reply to mark them done.

No UI. No login. Just email.

Sample Screenshots:

1. Inbox showing onboarding email:

1. Task Analysis Email (Bar & Pie Charts) — Sent when you email with subject ANALYZE:

Code Repository

Here is the link to the Github repository.

How I Built It

Mail-minders is built with:

1. Next.js — Handles incoming webhook requests from Postmark.
2. Supabase — Stores user's email-ID and tasks along with their categories.
3. Postmark — Powers all email communication (inbound + outbound)
4. Postmark templates — For the emails themselves
5. Python (FastAPI) — Generates bar and pie chart images for task summaries.
6. Render— Hosts the backend services, including the webhook and the cron job for daily reminders. (The cron job is implemented here but not configured to run because of Postmark's limit of 100 emails with the free trial)
7. Hugging Face Spaces — Hosts the task categorization fast-API, which uses the all-MiniLM-L6-v2 model from sentence-transformers to generate embeddings and classify tasks by similarity.

Postmark’s reliable delivery, easy to use email templates, and efficient inbound processing made it possible to build a complete experience using just email. Embedding-based classification added a layer of intelligence, allowing users to get structured insights.

Future Possibilities

Along with charts that show task summaries by categories, support for the following statistics may be added:

Basic Task Completion Metrics

• Total tasks created
• Total tasks completed
• Task completion rate

Time-Based Analysis

• Average time to complete a task
• Median time to complete a task
• Fastest completed task (shortest duration)
• Longest completed task (longest duration)

Daily & Weekly Patterns

• Most productive day of the week
• Least productive day
• Average number of tasks completed per day

Achievement Highlights

• Fastest streak (most tasks completed back-to-back in under an hour)
• Best day (day with most tasks completed)

This was a solo submission, built from scratch during the challenge.

Thanks to

• Dev.to for hosting this hackathon
• Postmark for sponsoring it.
• AI tools for helping with the coding, building and learning along the way.

Comments

[Jean-Michel 🕵🏻‍♂️ Fayard]

Congrats for your first post

What do you want to write about next?

[Ananya Kallankudlu]

Thanks a lot, got inspired after reading your post asking more people to start writing :)

[Nathan Tarbert]

Pretty cool seeing everything just run through email, way less clutter. I kinda wish I built something this simple before.

[Ananya Kallankudlu]

Thanks. I learned a lot, right from configuring parameters in a custom domain to text classification using vector embeddings. It was amazing. Looking forward to more building :)