~/projects / yovta
Yovta
A personal content reader. YouTube transcripts and summaries, RSS feeds and a read-later list in one place, and the good parts go into a second brain.
what it is
Following a few dozen thought leaders means hours of video and a pile of articles every week. Yovta pulls them in, transcribes and summarises videos, scores every item for relevance and surfaces what is worth reading. Whatever is interesting gets exported to an Obsidian vault as Markdown.
The next step is research: collect sources around one topic, then chat with them or let Yovta write a synthesis.
$ yovta nightly ✓ fetch youtube channels, rss feeds, read-later ✓ transcribe new videos ✓ summarise per item, with a template per source ✓ score relevance 0-10, keep >= 7 ✓ embed full-text + embeddings for chat → export selected items to the vault
how it is built
A single-user Django app, installed as a PWA on phone and tablet. Postgres is the source of truth; the vault is a one-way export. Summaries and chat run on configurable LLM providers, and an MCP server lets AI agents read from it and, through a guarded write gate, add to it.
stack
- Django
- PostgreSQL
- Embeddings + full-text search
- LLM providers
- MCP server
- PWA