Show HN: A Tool to Summarize Kenya's Parliament with Rust, Whisper, and LLMs

6 days ago 1

Bunge Bits provides convenient summaries of Kenyan National Assembly and Senate proceedings, making legislative information more accessible and digestible.

The driving force behind Bunge Bits is to strengthen Kenya's democracy by making legislative processes more transparent and understandable to all citizens. The aim is to bridge the gap between complex government proceedings and the average Kenyan, fostering increased civic engagement and political awareness. By offering concise, easy-to-digest summaries of legislative sessions, I hope to empower citizens with the knowledge they need to participate more fully in their democracy, hold elected officials accountable, and engage in informed discussions about the issues that affect their lives.

Ultimately, Bunge Bits seeks to contribute to a more engaged, informed, and participatory democratic process in Kenya.

ytInitialData parser: Parsing logic that parses scraped data from youtube
stream_datastore: Sqlite Postgres database bindings for storing, retrieving and modifying stream data
yt-dlp bindings: Bindings to interact with the yt-dlp cli as well as some utilities to interact with video and vtt data
ffmpeg bindings: Bindings to complement the Ytdlp bindings for the purpose of processing audio
stream_pulse_cron: A cron job that periodically fetches and processes streams
stream_digest: Functions that make it possible to efficiently process transcripts in chunks
LLMs: A service that interacts with OpenAI's Whisper and ChatGPT 4o to handle transcribing and summarizing the downloaded audio
Manual publish flag – is_published field added to each stream for editorial control. Yes it sucks, but it's necessary for now since LLMs don't seem to get everything 100% right. Only a temporary step for now though.
Web App: The end user interface that will display the summarized content #3
Email Newsletter – Periodic delivery of new summaries to subscribers