streaming-chat

A Preact demo of devframe's streaming-channel API combined with shared state for persistent chat history. The server emits synthesized "tokens" one at a time over a streaming channel, while the conversation log lives in shared state so it survives reloads, syncs across panels, and replays cleanly when a client re-joins mid-stream.

Package: streaming-chat-example · framework: Preact + Vite

What it shows

• A scoped context (ctx.scope('devframe-streaming-chat')) auto-namespaces every id.
• my.rpc.streaming.create('tokens', …) registers a streaming channel for low-latency token rendering.
• my.rpc.sharedState('history', …) keeps the message log on the server; each send appends a user + assistant pair atomically.
• The producer streams tokens live, then commits the joined content back to shared state when done — so refreshes and new clients see the finished message immediately.
• reader.cancel() aborts mid-stream; the assistant message is marked cancelled with whatever content accumulated.
• replayWindow lets a panel reopened mid-stream replay buffered tokens before resuming live.

To wire it to a real LLM, replace the fake token generator in src/devframe.ts with anything that yields strings — the stream's signal propagates cancellation from the browser all the way to the upstream request.

Run it

pnpm -C examples/streaming-chat run build
pnpm -C examples/streaming-chat run dev

Open the printed URL, type a prompt, watch tokens stream in, refresh mid-conversation, and cancel a long answer.

Source

examples/streaming-chat