About sembr
sembr is an open-source news intelligence system that synthesises multi-source signal (RSS / NewsAPI / RSSHub) into a daily, subscribable digest.
How it works
Two design choices do the heavy lifting:
- Semantic matching — multi-source signal (RSS / NewsAPI / RSSHub and more) is recalled by vector similarity and matched to the intent you define, by meaning rather than keyword hits. When the framing shifts, the stance flips, or entirely new wording is used, related content still surfaces as long as it is semantically relevant.
- A free-form analysis framework — behind every series is an analysis instruction you write yourself: which lens to break the topic down with, which variables to track, what structure to output — all defined by the operator. This site's macro series and its geopolitics series run two completely different frameworks on one and the same engine.
What it's for
- Read only the delta each day — digests persist as history, so today's analysis can read yesterday's conclusions and report the marginal change (new / ongoing / reversed) instead of re-telling old news.
- Distill a narrative from history — the daily archive can be re-synthesised: each series' "last 30 days" sidebar boils a month of daily analysis down to a few narrative threads.
Every digest carries clickable citation numbers, sources, a relevance score and a link to the original — conclusions stay traceable.
Open source & self-hostable
sembr's full source is released on GitHub under Apache-2.0. You can
docker compose your own instance and configure your own intents
and sources — everything you see on this site becomes yours once deployed.
About this site
sembr.live is the public showcase of the sembr project. Every series here is the real daily output of our own sembr production instance — generated and published automatically each morning. What you see is exactly what the engine produces day to day.
This site collects no user data: no analytics, no cookies, no email forms.
About us
sembr is open-sourced by Peakstone Labs — an AI quant lab made of one person and a swarm of agents. Follow our WeChat official account for the latest:
Disclaimer
For research reference only. Not investment advice.
Content auto-generated by sembr, curated by humans.