Layerbase

Free/cheap, unmetered, branchable, agent-ready databases. Postgres, MySQL, MariaDB, Redis, Valkey, SQLite, libSQL, ClickHouse, DuckDB & more. No lock-in.

[Layerbase @LayerbaseDB]

In this demo:

- Migrate Atlas MongoDB to Layerbase DocumentDB
- Integrate with GH/Vercel w/ branching
- Convert MongoDB to Postgres
- Migrate from Supabase w/ Auth
- DuckDB
- Agent Servers and mobile voice mode

Full Video: https://t.co/W1c896aegy

[Layerbase @LayerbaseDB]

@rovert_snikle 2.4x the code for a 20x smaller footprint is a very Rust trade. how long did the port take you?

[Layerbase @LayerbaseDB]

tired of copying every column from SELECT into GROUP BY? duckdb does it for you

SELECT city, street, avg(income)
FROM addresses
GROUP BY ALL;

groups by every non-aggregate column. add a column to the SELECT and the grouping follows
https://t.co/8CEkxex9i3

[Layerbase @LayerbaseDB]

@haydxnqn @grok Cloud agents are fine for PR-sized stuff, but they start fresh every time. A box that just stays on with your sessions alive is nicer when you want to check in from your phone and pick up where the agent left off. We run those at https://t.co/a9VG2KUiaR if you want to compare

[Layerbase @LayerbaseDB]

@lanre_olat refetching every full job description on a timer is the sneakiest egress leak there is. list view gets title + company, body loads on click, and most of that 7.25GB just disappears

[Layerbase @LayerbaseDB]

throw Better Auth in the bake-off. your users live in your own Postgres, so leaving Supabase later doesn't mean a painful user export, and there's no per-user bill creeping up. Clerk still wins on raw time to first login though. we built our own auth on Better Auth for exactly the lock-in reason

[Layerbase @LayerbaseDB]

@todoza_io that gap is absurd. what did you land on for backups and failover on the dedicated box? that's usually the part that drags people back to managed

[Layerbase @LayerbaseDB]

@lukehebb @laravel @PlanetScale congrats on winning the boss over. before the bandwidth question even matters, make sure the PlanetScale region matches wherever Laravel Cloud puts the app. a Laravel page firing 30 queries across regions feels slow way before any bill line does

[Layerbase @LayerbaseDB]

@nadir_seen_fire @k081e @firatoezcan @ElectricSQL @neondatabase @databricks scale to zero and a WAL subscriber are basically sworn enemies, so yeah the Neon discount just evaporates. you thinking about self hosting Electric against a plain always-on Postgres instead?

[Layerbase @LayerbaseDB]

@jordienr As long as they keep hosting their databases on Layerbase, I'll allow it

[Layerbase @LayerbaseDB]

@sspaeti I tried to do a demo of DuckDB since we added serverless DuckDB + DuckLakes - there was was no way to put together a demo showing all the ways someone can use it

I just cross-queried MariaDB x Postgres + JSON + CSV files with a single query

It didn't do it justice

[Layerbase @LayerbaseDB]

all this talk about Lean got me tinkering and it tickled the same part of my brain as modeling schemas with @TypeDB_

making assumptions explicit about what should be allowed

they're very different tools but its a familiar way of thinking

https://t.co/bfIEsZi1Xy

[Layerbase @LayerbaseDB]

@_heyeddi_ @webdevcody It came from a bunch of places, refactoring and rethinking what needs to be tested and when, using more processing in some places, choosing different tests on different branches, auto-detecting which files had changed and selectively running suites, etc

[Layerbase @LayerbaseDB]

@Eng_ALIAK thanks it wasn't easy

the proxy doesn't parse the protocol, it doesn't care about COPY vs query

it only hands off a socket once both buffers are drained

(≤200 ms wait retried every 2 s)

mid-stream COPY stays on the old process until done, a query still executing moves fine

[Layerbase @LayerbaseDB]

⛴️ SHIPPED
zero-downtime proxy upgrade

live database connections now move between proxy processes via Unix socket fd passing (SCM_RIGHTS)

your open transactions including LISTENs and Redis SUBSCRIBEs carry on through a deploy with a ~15 ms pause with no reconnects

this means...

we can ship new features, bugfixes, and enhancements several times a day without maintaining legacy containers

new users won't penalized for being early to the party

[Layerbase @LayerbaseDB]

@chenzhangreal We are trying to encourage people to leave Atlas for Layerbase, a customer asked us if we could help them convert from Mongo to Postgres and it sounded like a fun challenge

Believe it or not, in the week since release, it has been used by actual customers

[Layerbase @LayerbaseDB]

In this demo:

- Migrate Atlas MongoDB to Layerbase DocumentDB
- Integrate with GH/Vercel w/ branching
- Convert MongoDB to Postgres
- Migrate from Supabase w/ Auth
- DuckDB
- Agent Servers and mobile voice mode

Full Video: https://t.co/W1c896aegy

[Layerbase @LayerbaseDB]

@CodesHub @Railway Well, if you have any databases hosted on Railway we can help you out there

https://t.co/U6iio1lQpN

but if it's app hosting, we don't offer that... yet

[Layerbase @LayerbaseDB]

clickhouse ReplacingMergeTree only dedupes during background merges, which run whenever. until then you see old and new rows

SELECT count() FROM users FINAL;

FINAL dedupes by the ORDER BY key at query time. skip OPTIMIZE, it rewrites a ton of data
https://t.co/ePaf5QR7hP

[Layerbase @LayerbaseDB]

@thejohnwjohnson real reads instead of a homepage ping is smart since the pause looks at database activity, not site traffic. just don't let the keep-alive bot become your most loyal user lol

[Layerbase @LayerbaseDB]

@drsergiotavio @supabase @grok PocketBase is great if you're happy owning the VPS and the backups too. if you'd rather keep real Postgres plus auth without babysitting a box, that's literally what we build, and the free tier is a fine place to kick the tires

[Layerbase @LayerbaseDB]

@jmvldz @shreyas1009 @pmilano1 classic. everyone optimizes build minutes and then the Postgres instance quietly eats the invoice. is it the instance tier or storage doing the damage?

[Layerbase @LayerbaseDB]

@mirceadesign cron and webhooks are always the ones that sneak up on you. what's the plan for the sync jobs, a separate worker service or a cron container poking an endpoint?