如果您正在开发AI相关项目，推荐3个轻量级嵌入式向量数据库，不需要单独安装服务，直接引入官方SDK就能用。 1. LanceDB：适合Python、TypeScript、Rust 2. libSQL：由于SQLite 派生而来的开源数据库，目前已经支持向量数据。适合TypeScript、Python、Golang、Rust、Ruby 3. Ruvector：适合TypeScript、Rust Golang的选择比较少，TypeScript选择最多，libSQL支持的语言最全。这三个向量数据库都非常适合中小型AI项目，希望能有参考作用。 #向量数据库 #LanceDB #libSQL

Storage & infrastructure → ThreadState domain enables durable, agent-agnostic task tools → Postgres vNext observability domain for scalable OLAP logs → ClickHouse adds replicated tables for multi-node clusters → Durable Harness sessions across backends including LibSQL.

I've published a new package for TypeScript: a unified key-value storage interface. It provides throwable and `Result<T, E>` error handling. Supports custom codecs: JSON, SuperJSON, binary, etc. Observability with events pubsub. It currently works with node:fs, in-memory store, MySQL, PostgreSQL, libSQL, Redis and Cloudflare KV (worker bindings & rest api). NPM: npmjs.com/package/@temelj/st… JSR: jsr.io/@temelj/storage GitHub: github.com/tinrab/temelj/tre…

Cloudflare Workers Tursoのアプリをストレスかけずにローカル開発する為にlibsql-serverのコンテナ立ててるんだけど正解なのかどうか分かってない github.com/tursodatabase/lib…

完全にMac OS用の仕組みになるんのね Swiftネイティブでデスクトップ固定の金融チャート。Core Graphicsでまず実装し、Rendererを抽象化して後からMetal GPUへ差し替える設計。データはローカルDB（libsql ）で管理。リアルタイム描画

SQLiteとlibSQLとTursoは何が違うのか「おもちゃDB」が世界を飲み込むまで youtu.be/FO0O8OhVV68?si=PxDC… @YouTubeより

layerbase.com has soft-launched It is a one of a kind database solution, from managed cloud hosting to self-hosting tools, and desktop IDE that also hosts and deploys databases 21 engines supported Free tier for DuckDB, MySQL, Postgres, LibSQL and more

4️⃣Security and Tech Stack How secure is it? → Funds held in audited smart contracts. → Client cannot cancel after freelancer accepts. → Freelancer cannot touch funds until client approves. → Auto-release if client ignores for 7 days. → No admin can steal funds only resolve disputes Your USDC. Your contract. Your rules. For the builders reading this 🛠 → Smart contracts: Solidity 0.8.24 OpenZeppelin. → Chain: Arc Testnet (Chain ID: 5042002). → Frontend: React Vite Ethers.js v6. → Wallet: RainbowKit Wagmi. → Database: Turso (libSQL edge DB). → Backend: Node.js Express. → Gas token: USDC (no ETH needed)

I actually really liked what you guys did with libsql, that was the major limitation lol

@penberg already vibing at @tursodatabase with libsql and rust

При чому, як я зрозумів, це не працює для d1 через HTTP API calls без preserver connection state. Це саме має відбуватися й з libsql/Turso. Навіть на локальній емуляції з better-sqlite3

I forgot to mention all of SQL-99 with extensions for PostgreSQL and SQLite/libSQL dialects. I forgot all of openCypher 9 with extensions for Apache AGE and Neo4j. I have been busy the last month.

Turso（libSQL）が面白い。 SQLite互換のまま、 ・ベクトル検索 ・複数ライター ・WASMでブラウザ実行 ・AIエージェントごとの個別DB まで対応してる。 次の個人開発に利用してみるのはアリ。 github.com/tursodatabase/lib…

That’s the real take right there 👌 You can make SQLite sing if you’re willing to architect around it. But most of us would rather spend that brainpower on features, not tuning `PRAGMA` flags at 2am. LibSQL/Turso is kinda the best of both worlds SQLite simplicity with some of that “it just scales” peace of mind. Low key tempted to try it on the next side project.

Spent almost a week chasing a random 500 on Turso with Cloudflare Workers. SQL was correct. Credentials fine. Migrations clean. Still just “Failed query” with zero useful info. Turned out to be libsql-client issue (#339). A newer version switched internals to cross-fetch, which tries to use XMLHttpRequest under the hood. That works in Node or browsers, but in edge runtimes like Workers there is no XMLHttpRequest, only fetch. So the query builds fine, the client initializes fine, and then it quietly fails at runtime when the request is actually made. No clear error, just a generic “Failed query”. Downgraded and it instantly worked. AI actually helped a lot while skimming through the issue and connecting the dots. @tursodatabase this kind of silent runtime mismatch is brutal to debug. If you’re seeing “Failed query” for no reason, check your client version before you lose a week like I did. #webdev #debugging #turso #database #cloudflare #buildinpublic

I hate all the discussions around NVMe vs S3, because they miss all the nuances. The extremely obvious thing: the read or write latency from NVMe is going to be way smaller than S3 (or S3 Express). But... most discussions talk about database running in a single machine attached to NVMe. That comparison isn't fair, and here's why: When you write to S3, it is replicated to multiple machines. But your data is poof if your NVMe is gone. S3 Standard writes to multiple AZs. S3 Express One Zone writes to multiple machines in a single AZ. Only after the data is secured and durable is it acknowledged as successful. You can't get the same durability guarantees with NVMe attached to a single machine. If you implement your own quorum (within an AZ) to do the same, it will be around 2–3ms. I implemented a disaggregated storage engine for libSQL using FoundationDB. IIRC, the latency numbers were around 5–6ms with a 3 node FDB cluster in the same AZ. This was an out of the box FDB setup, without any config changes or tuning. I am really handwaving the numbers here, but the point is, when you write to a quorum the latency gonna be way higher than NVMe. The truth is somewhere in between, this is where DB vendors are fooling you: NVMe isn't as fast as they claim, but still faster than S3. So yeah, if you are comparing NVMe vs S3, include durability considerations. I'd love to see such benchmarks.

this is me Bun TypeScript Hono backend React 19 Vite frontend TanStack Router/Query Tailwind v4 Radix UI Drizzle ORM SQLite/libSQL Zod Baileys for WhatsApp Attio API OAuth Docker Biome

Kinariさんに教えてもらったデータベースが無料枠が結構あって、DBは100個まで・合計5GBまで無料で作れるということでかなり良かった。 AIエージェントアプリ作りたい人に刺さりそうなサービス。SQLiteみたいな感じで動くlibSQL。