ShipSF Workshops - Build an agent that plays a game for you

하이라이트

• Vercel launched Python general availability, daemons for long-running workloads, sandboxes, queues, crons, workflows, and fluid compute for automatic scaling.

• Vercel Services project presets enable multiple applications to deploy together in a single project with shared preview environments and private networking.

• A fully collaborative Jupyter notebook application with AI chat, secure virtual machine sandboxes, and WebSocket event fanning was built in three and a half hours.

• Workshop attendees can use backends.today to access setup instructions and claim one hundred dollars in gateway credits.

• The workshop game server runs on a Vercel daemon with a Python FastAPI control plane and a Vercel queue.

타임라인

1. 00:00 - 04:35

   Backend Platform Announcements and Architecture Primitives

   The team introduces new platform capabilities designed to support complex back-end architectures alongside front-end code. Fluid compute handles automatic scaling for API workloads, while daemons manage long-running tasks. Additional primitives like queues, cron schedules, workflows, and secure sandboxes provide the infrastructure needed to build robust applications. Project presets combine these components to enable shared preview environments and direct communication between services without complex routing rules.

   • Vercel announced Python general availability, support for Rust, Go, and OCaml, and new daemons for long-running queue consumers.
   • Fluid compute scales workloads automatically and bills only for active CPU execution time.
   • Vercel Services allows multiple applications to deploy within a single project for atomic deployments and private networking.
2. 04:35 - 13:10

   Building a Collaborative AI-Powered Python Notebook Application

   An application was developed to demonstrate the new Python capabilities on the platform. The architecture uses a FastAPI backend on fluid compute connected to a PostgreSQL database, while Vercel blob storage serves rendered notebooks via a CDN. Code execution occurs inside disposable Vercel sandbox virtual machines, and durable AI workflows power conversational agent assistants. Real-time updates for notebook creation and deletion are distributed to users through Vercel queues and WebSockets.

   • A collaborative Jupyter notebook application integrating FastAPI, PostgreSQL, and AI chat agents was built in three and a half hours.
   • Vercel sandboxes act as disposable virtual machines to execute untrusted Python code securely.
   • Vercel queues and WebSockets fan out real-time UI update events across connected browser sessions.
3. 13:10 - 28:43

   Interactive Workshop and Game Agent Setup

   The hands-on portion of the workshop guides attendees through building a custom agent to participate in a shared game simulation. The game server architecture utilizes a FastAPI control plane on fluid compute that relays commands through a Vercel queue to a game engine running on a Vercel daemon. Attendees clone the project template, configure a coding agent using Python or other supported languages, and deploy their applications to join the shared virtual environment.

   • Workshop participants build custom agents that connect to a multiplayer game server hosted on Vercel.
   • The game infrastructure relies on a FastAPI control plane communicating via Vercel queues to a game engine running inside a Vercel daemon.
   • Developers can access instructions and gateway credits at backends.today to deploy their own agents.