fluxy.config.ts

Use @fluxy-chat/config to declare per-room authorization, publish middleware, and client defaults — similar to a next.config for your chat deployment.

Example

import {
  defineConfig,
  allow,
  block,
  defineMiddleware,
  allowPublish,
  blockPublish,
} from "@fluxy-chat/config";

const moderate = defineMiddleware("publish", (ctx) => {
  if (ctx.capabilities.publish === false) {
    return blockPublish("You do not have permission to post.");
  }
  return allowPublish();
});

export default defineConfig({
  workerUrl: process.env.FLUXY_WORKER_URL,
  client: {
    readOn: "visible",
    wsCache: "on",
    historyLimit: 50,
  },
  rooms: {
    "support-*": {
      anonymous: false,
      authz: (ctx) =>
        ctx.anonymous ? block("Sign in to contact support.") : allow({ publish: true }),
      onPublish: [moderate.handler],
    },
    "room-*": {
      anonymous: true,
      authz: () => allow({ publish: true, invokeAgent: true }),
      onPublish: [moderate.handler],
    },
  },
});

Copy from packages/config/fluxy.config.example.ts in the monorepo.

Room patterns

Middleware

defineMiddleware("publish", handler) runs before messages are accepted. Return blockPublish(reason) to reject with a user-visible error, or allowPublish() to continue.

Worker integration

The Worker reads fluxy.config.js (compiled from TypeScript) on self-host.

Hosted overlay (no forked Worker): GET/PUT /admin/projects/:id/publish-config stores deny-substrings, guestCanPublish, iotAutoAgentId, and rooms (anonymous, per-room deny, capabilities, extension slots). pnpm fluxy:deploy uploads fluxy.hosted.json. The publish pipeline runs file onPublish first, then these D1 rules. Put an agent handle in iotAutoAgentId to invokeAgent on each iot.reading without an @mention.

Room extension snapshots: client.getRoomExtensions(roomId) / setRoomExtension(roomId, id, { kind: "kv", data }). Hosted slots are declared kinds (kv | counter), not eval.

Related

• Auth JWT
• Self-host one command