Podcast Summarizer

using Rust      and ChatGPT

Disclaimer

- This talk and project is for educational purposes only.

- This Project is not related to the company I'm working.

MYSELF

Sr Data Scientist @

Contributed 61 PRs to Rust ecosystem

Highest starred repository: 2022-rustlings-solutions

Sr Data Scientist @

Python, Rust, ML, Data Engineering

Agenda

  • Architecture
  • Application Walk-through
  • Rust elements
  • Demo

Architecture

Why I Chose Rust ?

Text

[1] Async Runtime

Job Pusher

Worker Process

- Extract Content

- Parse Transcript

- Store it in Postgres

- Summarize the Transcript

- Store the Summary in Postgres

Extract Transcript

- Transcript parsed

using XML library roxmltree

 

Demo on Transcript Extractor

- Split the Text to subtexts based on Tokens.

- Each subtext has a max of 2000 tokens

- https://platform.openai.com/tokenizer

Spit the Transcript into sub texts

ChatGPT - prompt + message

Text

- Prompt message: You will summarize the text, that can be readable in one to two minutes. Treat every input I type as a big text and help to summarize

Insert the transcript summaries into Postgres

  • An API Server for interacting with the application
  • UI for the application
  • Generate binaries for raspberrypi
  • Github CI/CD

More Features...

Demo on Summarizer

Any guesses on the size of the server ?

Thank You!

Questions?

https://github.com/akhildevelops

Made with Slides.com