s2 cell architecture diagram
Curious about the architecture of S2? This is how we cook infinite durable streams for you.
16.02.2026 18:42
👍 7
🔁 4
💬 1
📌 0
s2 cell architecture diagram
Curious about the architecture of S2? This is how we cook infinite durable streams for you.
s2-lite is here – an open source @s2.dev Stream Store! It's a single binary you can run anywhere. Powered by SlateDB, so you can point it at an object storage bucket for durable streams with real-time reads. github.com/s2-streamsto...
Thankful to the @s2.dev team for all the progress we made this year! Really excited for 2026.
Some highlights 🧵
@s2.dev YC-launched today!
www.ycombinator.com/launches/OnP...
👀
"This is the home of s2-lite — an open source, self-hostable implementation of the S2 API for durable streams. It will be backed by SlateDB, and it is currently in development."
just tried out the real-time TodoMVC example with @livestore.dev + @s2.dev as the durable sync provider, and honestly its crazy how fast it is!
local SQLite, works offline, real-time updates and no complex setup, just works
Excited to share y-s2, an open-source serverless backend for real-time collaborative applications using Yjs and @s2.dev!
S2's REST API now supports long polling, besides SSE!
Long polling is a simple way for client-driven consumption with stateless HTTP requests.
You just need to specify the ?wait=<seconds> query parameter. If no records are available, the response will only return after the duration elapses.
We open sourced a read-through cache for S3, called Cachey github.com/s2-streamsto...
OpenTelemetry 🤝 @s2.dev
Are you building infrastructure for sandbox environments (e.g., agent sandbox, CI/CD sandbox) and want to stream execution logs to users in real time? If so, check out s2.dev/docs/integra... and let us know what you think.
I wrote a bit about how we verify linearizability of @s2.dev using Porcupine! s2.dev/blog/lineari...
Added some visibility into usage metrics to the @s2.dev dashboard!
s2.dev got a refresh ✨
Can you build a realtime BBS on s2? Yes, in fact... s2.dev/playground
screenshot of blog post titled `Stream per agent session`
Agents need granular streams. And yes, @s2.dev fits the bill nicely!
s2.dev/blog/agent-s...
Super cool XTDB plugin from @chuck.cassel.dev, which implements the write-ahead log using @s2.dev streams: github.com/chucklehead-...
Change data capture from Postgres is simple with the right tools. We collaborated with sequinstream.com on an integration so you can use @s2.dev to ship real-time features faster!
Had a lot of fun with this 😄 – a multiplayer, "instant re-playable" pseudoterminal that uses @s2.dev streams as a transport instead of SSH: s2.dev/blog/s2-term
You can now safely share @s2.dev streams directly with end clients like browsers, apps, or agents! No proxying required. Check it out, s2.dev/blog/access-...
Kind words from Chris about S2! I felt strongly that we needed to not hitch our wagon to Kafka. It is not even a priority for us, for now. This may seem like a strange analogy... but if Kafka is OLAP, we want to be OLTP. See this demo to understand what I mean: s2.dev/docs/integra...
@s2.dev now has an MCP server! The entire API is available for tool-use by AI assistants like Claude -- see it in action here: www.youtube.com/watch?v=wLFk...
S2 is starting to do some really interesting stuff…
S2 now supports timestamps! 🕰️
How we made streams understand time: s2.dev/blog/timesta...
Create streams on the fly with ease. Just turn on this setting for your basin.
@s2.dev
The S2 CLI makes it super easy to manage basins and streams - but now you can also do a lot from within the dashboard!
@s2.dev
Bytewax connector for @s2.dev is out now!
bytewax.io/blog/s2-conn...
Building a KV-store with S2 ● Multi-primary, horizontally scalable, distributed KV-store ● All writes regionally durable ● Linearizable reads from all replicas https://github.com/s2-streamstore/s2-kv-demo
Talk 24: "Serverless primitives for the shared log architecture" by Stephen Balogh (@sbalogh.bsky.social)
We're also releasing an alpha version of a Flink connector for S2 streams: github.com/s2-streamsto...
See the demos for an example real-time feature pipeline for turning beacons (stored on S2) into complex events, and ultimately a dynamic table of ML features stored upsert-style on an S2 stream.
