hiPDF

Inspiration

With the advent of AI. I saw this as a opportunity to learn about the AI technologies while building something useful.

What it does

It's a web app powered with the latest AI tools like langchain and OpenAI. The user can upload any pdf and can ask his/her queries in a chat like manner.

How we built it

• I used streamlit as a quick means to build the application's frontend.
• Created OpenAI account and generated api key to get access to the gpt models
• Leveraged langchain's apis to build the make Q&A retrieval system ## Challenges we ran into I had planned to add TTS (text to speech) too. I even had the TTS part checked using eleven labs But ...
• Due to limitations of streamlit i found it very difficult to add a button or image (speaker icon for TTS) and to write the logic for it
• Even after generating the tts.mp file. I had a lot of trouble figuring out ways to make the mp3 file play automatically using #python ## Accomplishments that we're proud of Was able to at least complete the pdf chat app on time. ## What we learned
• limitations of streamlit and such tools
• when to use tools like streamlit vs using web frameworks ## What's next for hiPDF
• Switch from streamlit to a dedicated framework
• Add features like TTS
• Pdf Chat Histories
• Many more....

Built With

• langchain
• openai
• python
• streamlit