Weirling
Stop skimming 50 feeds: one daily digest, ranked by what you actually finish reading.
Self-hostable reader for RSS, newsletters and creators. Learns from scroll depth and reading time, and says why each item ranked.
Hosted beta · Documentation · Quickstart · Releases
Watch a 40-second narrated tour of the digest, why each item ranks and what it learns.
What it does
- Brings your reading into one place: RSS and Atom feeds, newsletters
forwarded to a private address, and creators on Substack, YouTube, Medium,
Bluesky, Mastodon, GitHub and podcasts.
- Ranks by behaviour, not keywords. The in-app reader records how far
you scroll and how long you actually read. Eight signals, weighted for you,
give each item a score; the biggest contributions become the reason shown
next to it (how ranking works).
- Sends one digest by email or in the app, at a time it learns from when
you open it.
- Shows what it learned: your interests are a topic graph you can browse,
rename, merge, boost or mute.
- Runs on your machine. Embeddings are computed locally. An LLM key is
optional; without one, summaries are extracts and nothing leaves your
server.
Compared with other readers
| What decides importance | Self-host |
|---|
| Weirling | Ranking learned from how you read (scroll depth, time, ratings), explained per item | yes (MIT) |
| Feedly | AI priorities you train with topics, keywords and examples (Pro+) | no |
| Inoreader | Rules and filters; AI summaries and tags, no ranking by reading | no |
| NewsBlur | A classifier you train on keywords, authors and tags | yes |
| Readwise Reader | A read-later queue; not ranked | no |
Prices, AI features and sources: docs/COMPETITORS.md.
When to use something else
- You follow a handful of feeds and read them all: a plain RSS reader is simpler.
- You want X, LinkedIn or Reddit timelines: they have no usable public feeds,
and Weirling marks them unsupported rather than scraping them.
- You need a team or enterprise product with SSO and an SLA: Weirling is an
early beta run by one maintainer.
Quickstart
Needs Docker with Compose v2 and about 2.5 GB of RAM (1 GB for the lite image).
git clone https://github.com/mohitmishra786/weirling.git && cd weirling
./scripts/init_env.sh # writes .env with fresh secrets
docker compose up -d # 8 services; migrations run on start
Open http://localhost:3001 and create your account: the first one owns the
instance. Ports 8000 and 3001 must be free; Install & run
covers changing them, the lite and full profiles, AI summaries, email,
backups and upgrades.
Status
Early beta, v0.3.x. Self-hosting works today; the hosted service at
weirling.com is invite-only. Expect breaking changes
between 0.x releases: every release's notes say what to do, and migrations
run automatically (upgrading).
Documentation
| |
|---|
| Install & run | Profiles, first-boot setup, AI summaries, email, monitoring, backups |
| Configuration | Every setting you are likely to change |
| Deploy to Vercel + Railway | The setup behind weirling.com |
| Architecture | Components, ingestion, how a digest is built |
| How ranking works | The eight signals and how your weights change |
| Development | Running from source, code layout, scheduled tasks |
| Privacy · Terms · Security | What is collected, the rules, reporting a vulnerability |
Contributing
Issues and pull requests are welcome; CONTRIBUTING.md has
the setup, the checks to run and how to add starter packs. Security reports go
to SECURITY.md, not public issues.
License
MIT. The ranking engine is also published on its own as
weirling-rank.
