release.2026-08-05T08-59-48Z.2496
teableio/teablerelease.2026-08-05T08-59-48Z.2496Aug 5, 2026by tea-artist
AI Summary
A refinement release addressing computed field bottlenecks, AI generation interruption handling, and maintenance workflows for legacy users.
Key Highlights
- Improved computed-field processing with bounded stages
- Graceful interruption handling for AI generation across all windows
- Admin maintenance check for finding and repairing legacy users
- Timezone standardization to prevent drift in non-UTC deployments
New Features
- AI generation interruption handling
- Admin maintenance check
- Computed field dependency fixes
- Timezone standardization
- Socket error fixes for saved views
Full Release Notes
## Feature Updates - Added targeted in-app announcements with banner, toast, modal, and sidebar formats, plus scheduling, localization, audience selection, links, and withdrawal controls. - Added read-only deleted-record snapshots to the recycle bin, with filters for resource type, operator, creator, and deletion or creation time. - Added Business-tier record archives with snapshot viewing, restore, permanent deletion, CSV export, and longer-term retention. Keyword search is no longer available in archive or trash snapshot views. - Added an admin maintenance check for finding and repairing legacy users without avatars, with bulk processing and progress feedback. - Published apps now have a more visible quick-open link, clearer unpublished-change indicators, and more reliable version history and rollback behavior. - Updated in-app changelog links for the App Builder Visual Editing release and GPT 5.6 pricing changes. ## Bug Fixes & Improvements - Stopping generation is now treated as a graceful interruption across chat, App Builder, and agent runs. Other windows no longer show server errors, and interrupted runs display the correct status. - Fixed BYODB compute backlogs that could block record creation, deletion, or primary-field updates with 503 errors. - Improved computed-field stability on large or highly connected bases by processing formula, lookup, and rollup updates in bounded stages while preserving dependency propagation and progress. - Fixed computed-field write failures involving field-to-field conditional comparisons, conditional rollups, circular dependencies, and complex dependency graphs. - Improved lookup read-after-write latency for bounded single-record updates. - Fixed schema changes that could become stuck when converting field types on tables with computed fields. - Standardized database timestamp handling on UTC, preventing timezone-dependent drift in scheduled tasks, metadata, billing periods, and other date-based workflows. - Fixed AI-created nodes being placed outside the source node’s folder and clarified folder selection for create commands. - Strengthened sharing permission enforcement across tables, views, forms, and access tokens. Unavailable actions are now disabled with explanatory tooltips. - Fixed view filter changes not appearing immediately, collaboration inconsistencies, filter deletion errors, and failures caused by legacy view metadata. - Improved reliability across view creation, updates, sharing, imports, duplication, concurrent edits, shared views, and realtime projections. - Improved audit logs by recording user identity more consistently, including field details in creation entries, preserving event times, and reducing missing entries during large bulk operations. - Fixed deletion of orphaned linked-record fields after their referenced tables have been removed. - Improved V2 record-write compatibility with V1, including empty-value normalization, required-field enforcement, duplicate record ID handling, link value normalization, and stale constraint cleanup after field conversions. - Empty text, multi-value, attachment, user, link, and unchecked checkbox values written through V2 are now stored as `null`; required fields reject these empty values consistently. - Fixed V2 record updates failing when user fields contained members without custom avatars. Affected legacy user data is repaired automatically while preserving initial-based avatar fallbacks. - Improved V2 behavior across formulas, lookups, rollups, filters, sorting, grouping, search, aggregation, field validation, base duplication, and table and field lifecycle operations. Many-to-one lookup results remain arrays in V2. - Fixed V2 date validation so strict writes reject invalid calendar dates and typecast writes store them as empty values instead of silently changing them. - Text-to-date and formula-to-date conversions now skip invalid calendar dates without blocking the entire field conversion or exposing raw database errors. - Fixed rating conversions producing invalid zero-star values. Decimal values are now rounded, and results below one star become empty. - Strengthened record and table permission enforcement across API versions, including disabled grants and field- or row-level access restrictions. - Improved record and grouped-read consistency for filters, sorting, search, attachments, shared views, masked fields, and collapsed groups. Unsupported queries against unreadable fields may now be rejected. - Fixed projected record reads omitting metadata required by grouped personal and grid views. - Improved self-hosted signup so new users reliably join all eligible auto-join spaces as Viewers, including when SSO and admin auto-join target the same space. - Fixed the unsubscribe page becoming unusable after a successful preference change when a subsequent refresh failed. - Enabled compression for realtime WebSocket traffic, reducing payload sizes and improving large-grid and record-update performance, especially on slower networks. - Fixed automation runs executing the same action chain twice, reducing duplicate messages, webhooks, and emails. Interrupted runs are now recovered more safely without replaying completed actions. - Automations watching formula fields now run when source-field updates change the formula result, while safeguards prevent duplicate triggers and loops. - Fixed large V2 deletion undo operations failing to restore records when triggered quickly, improving undo and redo reliability. - Improved large import and table duplication stability by reducing unnecessary history writes and database pressure. Imported and duplicated records no longer create initial null-to-value history entries. - Improved mobile attachment previews with wider portrait layouts, better spreadsheet scrolling, non-overlapping controls, and correct rotated-image fitting. - Improved base loading and navigation by reducing duplicate requests and opening previously visited, pinned, or first-entry tables and views directly when permitted. Stale destinations now fall back safely without redirect loops. - Improved localized record validation messages for required and unique fields, including clearer field-specific duplicate errors in bulk operations. - Improved generation checkpoint reliability during heavy database load, reducing failed chat and App Builder progress saves. - Future AppSumo activations no longer receive the previous one-time credit bonus. Existing credits remain valid, while credits from refunded or deactivated licenses may still be revoked. - Removed outdated in-app pricing and FAQ content; current pricing information is available from the official marketing pricing page.