Dibao is a self-hosted RSS reader that ranks articles only inside the feeds you choose. It turns RSS unread debt into an explainable personal reading queue, with data kept in SQLite on your own host. It avoids expensive LLM-as-feed-engine designs: recommendations can use a small 0.6B embedding model through a provider or locally, with baseline ranking fallback.
No reviews yetBe the first to leave a review for Dibao
Maker
📌
Hey Product Hunt!
I built Dibao because RSS gives me the control I want over sources, but once the list grows, a chronological inbox turns into unread debt.
Dibao keeps the RSS boundary intact: it only ranks articles from feeds you already chose. The idea is not to replace RSS with AI summaries or another infinite feed, but to make personal reading easier to keep up with while keeping the data local.
A design goal was to avoid making this another project where you connect an LLM API to your reading list and hope the bill is fine. Recommendation is mostly matching and ranking, not generation. Dibao can use a small 0.6B embedding model via a provider or locally; for normal personal use, provider cost should stay under $10/year, and local embeddings can make it effectively $0 apart from your own hardware.
It is early, but v0.1.0 is runnable with Docker/Compose. I would love feedback from people who still use RSS, self-host personal tools, or care about keeping their reading behavior on their own machine.