February 28, 2026
Tiny DBs, giant feelings
Show HN: SQLite for Rivet Actors – one database per agent, tenant, or document
A tiny database for every bot — cheers and side‑eye erupt
TLDR: Rivet proposes giving each small app component its own tiny database for simplicity and speed. The community split fast: fans love the clean isolation, critics worry about too many databases, cross‑actor queries, and transaction pitfalls—while one user reports success moving from Cloudflare’s model. It could reshape how AI and real‑time apps are built
Rivet just lit up the timeline with a spicy pitch: give every "actor"—their mini app that powers a bot, chat room, or doc—its own tiny SQLite database. The idea promises simple isolation and fast, local storage. The crowd reaction? Half applause, half clutching pearls.
Fans like swaminarayan hail it as “clean isolation,” but immediately ask the million‑table question: how do you do cross‑actor queries without pain? Skeptics like nudpiedo argue it’s wasteful to spin up a whole database for each tiny thing when a big shared one could handle complex queries centrally. And siliconc0w throws a red flag on the play: don’t reinvent transactions, the guardrails that keep databases from mixing up your data.
Meanwhile, a real‑world win: fastball says they swapped their feature from Cloudflare’s "Durable Objects" (Cloudflare’s per‑object compute) to Rivet for better fit with their AWS/Vercel stack—and the Rivet team was quick on support. Others, like ZeroAurora, want a head‑to‑head comparison in the README with Cloudflare Durable Objects for clarity and marketing.
The memes flowed: “one tiny DB for every imaginary friend,” “Gotta shard ’em all,” and “my toaster gets a ledger now.” Love it or side‑eye it, Rivet just made the debate over how to build AI agents and real‑time apps way more interesting—and way more dramatic.
Key Points
- •Rivet Actors provide a serverless, stateful actor model with built-in state, storage, workflows, scheduling, and WebSockets.
- •Persistence options include SQLite or JSON, enabling per-actor databases for agents, tenants, or documents.
- •Use cases span AI agents, sandbox orchestration, multi-step workflows, collaborative documents, per-tenant databases, and chat.
- •Deployment options include self-hosting (Rust binary/Docker with Postgres, file system, or FoundationDB) and Rivet Cloud (managed, global edge) with integrations to Vercel, Railway, and AWS.
- •The project is open source (Apache 2.0) with SDKs (TypeScript, Rust, Python), integrations (Hono, Elysia, tRPC), and components like Rivet Engine, Pegboard, Gasoline, Guard, and Epoxy (EPaxos).