Open Chat Updates

Added a new Coding Standards and Banned Patterns section to clarify which React and TypeScript approaches are not allowed, and what to use instead. This helps teams avoid common sources of subtle UI bugs, like effect-driven state syncs, fetch-in-effect patterns, and hard-to-debug user action triggers.

Added a dedicated Cursor Cloud guidance section to the agent handbook to help cloud-based agents start the app and verify changes more reliably. This clarifies which services can be run during development, what to expect from each one, and common caveats that can block startup or testing.

This release tightens up chat cleanup and regeneration flows so streams and follow-up messages are handled more consistently. It also fixes a few edge cases across web and mobile, including safer typing requests, more reliable draft saving, and better mobile detection for responsive UI.

You can now generate a public share link for a chat and open a read-only shared view. Shared pages include dynamic preview metadata and a generated Open Graph image, making links look better when posted in messages or on social platforms.

Analytics loading is now more reliable by centralizing the OneDollarStats script URL and removing the outdated integrity check that could cause browsers to block the script. This helps ensure tracking continues to work as the vendor asset changes over time.

This release hardens the Sentry monitoring tunnel to better protect against oversized or malformed requests and reduce the impact of abuse. It also makes trace sampling configuration more resilient by falling back to safe defaults when values are invalid.

This release adds a dedicated monitoring endpoint to support in-app error reporting and performance tracing across both the browser and server. It includes safeguards like same-origin checks, payload size limits, and rate limiting options to help keep monitoring traffic reliable and controlled.

Starting the web app in local development is now more reliable, especially when the expected startup command has drifted over time. The dev launcher now corrects stale settings automatically and ensures the right hostname behavior is applied consistently when running through the standard workflow.

Starting the web app in development is now compatible with the current Portless command format, preventing startup failures caused by an unsupported flag. This keeps named host behavior intact, which helps avoid collisions when running multiple worktrees or instances.

Local development now avoids hostname collisions when you have multiple checkouts or worktrees. Each worktree gets its own stable local URL, and server-side origin checks now recognize that derived host so sign-in and other auth-related requests keep working.