Async prompt runs

Major core upgrade delivering faster runs, higher reliability, and shared context.

What’s new

• Async runs — LLM API calls now execute asynchronously in a dedicated Cloud Run worker
• Realtime streaming — LLM response chunks stream to the web app over WebSockets for near-instant feedback
• Run history (foundation) — Every prompt run and result is persisted; latest run is visible now, full history browser coming soon

Why it matters

• Faster perceived responses, fewer timeouts, more stability
• Reproducible LLM outputs with shared context

How to use

• Run a prompt as usual
• Watch tokens stream live
• Refresh or switch file tabs anytime — execution logs persist