TaskKeepr

UI Dashboard
Microsoft Teams Personal Update Bot
Microsoft Teams Threads Bot

Inspiration

In the world of remote-work, keeping track of your co-workers availability - and informing them of your own plans - is a constant headache for many. TaskKeepr seeks to streamline this process by monitoring MS Teams chats and providing summaries of chats and messages that your co-workers want you to see but don't have a good option for doing so. The goal of TaskKeepr is to take the meaning of the @General tag back to its roots, namely mission-critical updates instead of unscheduled appointments with the dentist.

What it does

When a Team uses TaskKeepr, a MS Teams bot will sit-in on their existing channels in order to find and summarize messages related to important updates and scheduled - or unscheduled - time off taken by your coworkers. By utilizing the Cohere LLM, we can quickly and efficiently summarize and categorize text without complex classification algorithms.

How we built it

TaskKeepr utilizes a simple React frontend to display info about your teams schedules, but the critical mass of the app lies below the surface. A Cohere LLM takes care of text summarization, sentiment analysis, and classification using modern prompt engineering. A Flask backend API communicates between the Cohere API and the CockroachDB SQL database while also sending information on events, times, and user calendars to the React/Tailwind frontend.

Challenges we ran into

Learning to leverage tools such as CockroachDB and Cohere were clear challenges when we set out on this project, but thanks to solid documentation, good workshops, and some expert help, we were able to create a fairly robust and comprehensive backend in only 36 hours. The frontend was challenging for its own reasons, mostly due to the complexity of the backend technology. However, React is a very bootstrap-able framework, and so the end result is still good at displaying the required info.

Accomplishments that we're proud of

Building a complete full-stack application that utilizes new technologies and a suite of ML technologies is (in our opinion) an accomplishment in-and-of-itself. However, the team is particularly proud of our ability to leverage technologies that were completely new to us before this event. Cohere, CockroachDB, and the MS Teams API, AKA the majority of our backend tech stack, were previously untouched before this weekend, so being able to successfully implement and use these tools is a useful skill to have.

What we learned

As the vast majority of Hackathon projects live and die by the clock, the whole team learned plenty about time management and the importance of efficient grinding throughout a short period of time. Taking short/medium breaks throughout the event proved to be an extremely helpful boon to our productivity. In addition, everyone on the team honed their ability to quickly pick-up and adapt to new technologies, an undoubtedly useful skill.