Build an AI chatbot where each conversation keeps its full history in memory for instant context without database queries.
Create a real-time collaborative document editor where each document is one actor broadcasting changes to all connected users.
Run background tasks with automatic retries and scheduling without setting up a separate job queue system.
Self-hosted production deployment requires Docker, Rivet Cloud option skips infrastructure setup entirely.
Rivet is a framework for building stateful server processes called actors. An actor is a lightweight, long-running program that holds its own data in memory and can communicate with clients in real time. The idea is that instead of a traditional server where every request is handled independently and state is stored in a separate database, each actor keeps its own state right next to where the code runs, making reads and writes much faster and simpler to reason about. The model is well-suited to use cases where each user, session, or conversation naturally has its own isolated context. For example, in an AI chatbot, each conversation would get its own actor that keeps the full message history in memory. When the user sends a new message, the actor processes it, streams the response back in real time, and saves the result. No external database call is needed for the most common operations. For a collaborative document editor, each document would be one actor broadcasting changes to everyone connected to it. Actors support WebSocket connections for real-time two-way communication, a built-in message queue for reliable background processing, scheduling for timed and recurring tasks, and multi-step workflows with automatic retries if a step fails. When an actor is idle, it hibernates and costs nothing to run. When needed again, it starts back up in roughly 20 milliseconds. The framework provides a JavaScript and TypeScript library for writing actors, and a Rust-based engine that runs them. There are three ways to deploy: running everything locally inside your own application process during development, self-hosting the engine on your own servers using Docker or a single binary, or using Rivet Cloud, the managed hosting option that handles scaling and distributes actors across a global network so they run close to your users. A built-in browser-based inspector lets developers watch actor state, browse the actor's local database, monitor events, and send commands directly to a running actor during development and in production.
← rivet-dev on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.