Multi-user AI Chat with RedwoodSDK RSC and Cloudflare Agents

This is an experimental project, looking at how to live-stream AI responses back to multiple connected clients. All implementations use Cloudflare durable objects and React Server Components (RSC) with RedwoodSDK.

This is a companion repository for a blog post, deployed at https://agents-chat.jldec.me/.

Implementations

1. RSC Chat - Uses RedwoodSDK realtime websockets
2. Agent Chat - Uses Cloudflare Agents websockets with separate durable object storage
3. Agent SDK Chat - Uses Cloudflare Agents AIChatAgent with the useAgentChat hook
4. TinyBase Chat - Uses TinyBase websockets
5. Agent Agent Chat - Advanced Cloudflare agent with subagents and MCP tool calling

Takeaways

RedwoodSDK RSC:

• Server components provide succinct way to populate JSX and keep clients updated.
• This makes using react with Cloudflar workers super easy, and simplifies async data loading.

Cloudflare Agents:

• AIChatAgent provides its own websocket protocol for multi-user real-time sync
• Agents also offer raw websockets give full control over payloads for optimizations
• Vercel's AI SDK abstracts tool calling and supports different LLMs

TinyBase:

• DB sync engines can improve UX with local-first client-side persistence
• The approach requires careful validation since database operations run on the client

Links

• Live Demo: https://agents-chat.jldec.me/
• Blog Post: https://jldec.me/blog/multi-user-ai-chat-with-redwoodsdk-rsc-and-cloudflare-agents
• RedwoodSDK Docs: https://docs.rwsdk.com/
• Cloudflare Workers Docs: https://developers.cloudflare.com/workers/
• Cloudflare Agents Docs: https://developers.cloudflare.com/agents/