KataPDF

• 

  Landing Page

• 

  Virtual Assistant Page

Inspiration

As a student ourself, we knew how much time meant for all of us, yet it's all wasted away flipping through pages, scrolling through lecture notes, desperately looking for some information we need. As a student, it's also critical for us to get accurate and reliable sources of information. But that's only the first step, we also need to extract the informations we need and filter out any unneccesary information. Despite the importance of those skills, based on the survey conducted by the Program for International Student Assessment (PISA), released by Organization for Economic Co-operation and Development (OECD) in 2019, Indonesia are among one of 10 worst country in terms of literacy. Alongside that a lot has evolved in the past couple of years, and maybe how we study should change too. Speaking from personal experience, we have numerous experience where we had trouble finding, processing, and summarizing a certain piece of information. So we came up with this idea where user can upload a pdf file and a virtual assistant will summarize the whole file just for you. All informations related to the file will be available within a single prompt. Not only that, it could also serves well in helping you study, might it be for an exam or just daily studies, it's really convenient to have someone alongside you that you could ask and explain to you a well formated, summarized, and accurate answer, well with KataPDF we're hoping to assist everyone by providing our own virtual knowledge assistant that could answer all of your questions in no time, fully customized to you and you're document of choice.

What it does

KataPDF is a tool built for students to efficiently extract informations within a pdf file, it could also serve as your own personal tutor and your learning buddy that you could ask any question you have regarding a subject and the document you provide it and it'll respond with a well formatted, simple, and easy to understand answers. With it's intuitive UI, users will have an amazing experience using it on a daily basis. Its main features are:

Any information within 1 prompt

Users can input any prompt related to the contents of the file, and their own personal virtual assistant will assist them in extracting, summarizing, and explaining the informations within the file.

Sharing knowledge bank

Users can also share their assistant and knowledge bank to other people all within a single link. Results may vary due to each users having their own virtual assistant, but it'll not vary too much.

How we built it

Frontend

• NextJS
• ChakraUI
• TailwindCSS

Backend

• Prisma
• Langchain
• Postgresql & Pgvector
• Vercel Blob

Challenges we ran into

While building this product, there's some challenges that we face, such as:

• It cannot be used for large files, as it results in timeout when parsing the file;
• We're using bleeding edge technology, thus lesser robust documentations;
• Deployment phase, as we need to sync between pgvector and the client, and any latency will cause a drop in performance;
• No budget for a better server, thus it hinders the performance of the product. Besides that, it's been smooth sailing

Accomplishments that we're proud of

We're proud that we're able to bring it from a simple idea into a working product, despite some of it's flaws we're still joyous to be able to make a MVP. We're excited that this product might solve a lot of common problems we face as a student, just like how it solved our problems.

What we learned

We learned a lot about how to develop our idea into a working product while working better as a team, we also learned how to overcome challanges and find other creative solutions to fulfill our goals.

What's next for KataPDF

First and foremost we would like to improve the capacity of our parser, so it could accomdate larger files and provide faster responses. Second of all we would like to polish it even more and hope we might include more features.

Built With

• chakraui
• langchain
• nextjs
• pgvector
• postgresql
• prisma
• tailwindcss
• vercel