Chat with any AI model available through OpenRouter in one interface.
Resume an AI reply exactly where it left off after a dropped connection.
Study a reference implementation of durable, resumable AI streaming.
Attach images or audio and get spoken replies in a chat conversation.
| prisma/open-chat | adrianhajdin/job_pilot | cneuralnetwork/pracdago | |
|---|---|---|---|
| Stars | 20 | 20 | 20 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | hard | moderate |
| Complexity | 3/5 | 4/5 | 2/5 |
| Audience | developer | developer | vibe coder |
Figures from each repo's GitHub metadata at analysis time.
Runs as a single Bun process with no separate build step, but production deploy targets Prisma Compute.
Open Chat is a multi-model AI chat application built by Prisma, a company known for its database tools. The app is publicly available at oss.chat and requires no sign-up to try. It works with any AI model available through a service called OpenRouter, which means you can switch between different AI models (text, vision, and image generation) within the same conversation. The most distinctive feature is what the README calls durable, resumable streams. When an AI model is generating a reply, every token it produces is saved to a durable log before it appears in your browser. If you close the tab or lose your internet connection while a reply is loading, reopening the chat picks up exactly where the answer left off. Nothing is lost to a dropped connection. This is the main behavior the codebase is designed to demonstrate. The application includes a credit system: new users get $2.00 of free credit on signup, and can add more via Stripe in amounts from $5 to $100 with a 10% fee. You can also attach images or audio to messages, dictate voice notes, and receive spoken replies. A public statistics page shows anonymous usage totals. The code is written in TypeScript using Bun (a JavaScript runtime) and React, and the whole application runs as a single process with no separate build step required during development. Prisma built this app specifically to show how their Prisma Streams, Prisma Postgres, and Prisma Next products work together in a real project. Deployment instructions target Prisma's own hosting platform, Prisma Compute. The repository is about 5,500 lines total. The README walks through the core files in reading order for developers who want to understand how durable streaming works. The README does not mention a license.
A multi-model AI chat app that never loses your reply mid-stream, even if you close the tab or lose your connection.
Mainly TypeScript. The stack also includes TypeScript, Bun, React.
Setup difficulty is rated moderate, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.