Teable's View-Aware AI Chat gives you precise control over every piece of data, every row and column, every table, and every view.

1. What You See Is What AI Gets

AI Chat now perceives your active table view — including filters, sorts, and groupings. When you ask it to update data, it operates within your current view context, not the entire table.

2. Precise CRUD via Chat

Create, read, update, and delete data using natural language. Ask AI to "mark row 123 as High Value and row 456 as Follow-up," and it updates them instantly.

3. Improved AI Field Output

AI-generated field content is now more polished and accurate, with better formatting and fewer edge cases in output quality.

Additional Improvements

AI & Chat Fixes

• Fixed the AI settings reset button restoring the last saved model instead of the actual default.
• Fixed AI Chat attachment example upload failing during the onboarding flow.
• Fixed the Trash cleanup processor repeatedly failing due to a 404 error on deleted chat resources.
• Fixed the Created Time column stopped auto-updating for new records.

V2 Database Engine Upgrade — Invite-only Beta

• Fixed a stack overflow error in V2 computed field polling.
• Fixed database duplication failing with errors whether records were included or excluded.
• Fixed newly created tables not appearing in the sidebar navigation until a manual refresh.
• Optimized V2 computed field recalculation to avoid unnecessary full-table recomputation.
• Added Sentry tags to distinguish V1/V2 environments for faster issue triage.
• Rebuilt Create Table and Restore Table operations under the V2 architecture for improved reliability.
• Fixed Formula fields not displaying values in real-time after creation in V2 — previously required a page refresh.
• Fixed V2 trigger chains not properly propagating updates to dependent AI fields and automations.
• Fixed the API view filter returning all records instead of the filtered subset in V2.
• Integrated the V2 authorization module into the plugin architecture for unified permission handling.
• Introduced a plugin system with Hook and constraint injection for safe, flexible extensibility.
• Enhanced observability for Formula, Rollup, and Lookup computation tasks with better tracing and monitoring.
• Improved monitoring and resource usage for better overall stability.
• Improved real-time update reliability with recoverable and concurrency-safe outbox processing.
• Added a fast-path optimization for conditional aggregation to reduce redundant computation.