This release includes 1 security fix for security teams reviewing exposed deployments.
Topics
+6 more
Affected surfaces
Summary
AI summaryAI Chat adds field name targeting, table context selection, draft persistence across pages, and session‑scoped OTP cooldown hardening.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Medium |
AI Chat supports adding field names from column headers. AI Chat supports adding field names from column headers. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
AI Chat supports table context with selected rows, columns, or cells. AI Chat supports table context with selected rows, columns, or cells. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
AI Chat selections are clearer with labels like “Row 1”. AI Chat selections are clearer with labels like “Row 1”. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
Improved mobile model selection using a clearer bottom sheet. Improved mobile model selection using a clearer bottom sheet. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
AI Chat saves drafts per session, restoring unsent content after refresh. AI Chat saves drafts per session, restoring unsent content after refresh. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
AI Chat input persists across pages (App Builder, table chat, node pages). AI Chat input persists across pages (App Builder, table chat, node pages). Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
AI Chat automatically refunds failed usage with records in usage logs. AI Chat automatically refunds failed usage with records in usage logs. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
Clearer long-chat compression prompts appear in the response stream. Clearer long-chat compression prompts appear in the response stream. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
AI field abnormal usage detection intercepts excessive Credits usage for confirmation. AI field abnormal usage detection intercepts excessive Credits usage for confirmation. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
More controllable AI field batch autofill with added batch limits and concurrency controls. More controllable AI field batch autofill with added batch limits and concurrency controls. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
Improved AI image generation reliability with enhanced retry handling and multimodal support. Improved AI image generation reliability with enhanced retry handling and multimodal support. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Feature | Medium |
Notification Center adds “All” category to view all messages by default. Notification Center adds “All” category to view all messages by default. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Feature | Medium |
Important notifications stay visible, requiring manual dismissal for high-priority alerts. Important notifications stay visible, requiring manual dismissal for high-priority alerts. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Feature | Medium |
Admins can send in-app notifications with longer, multi-line content. Admins can send in-app notifications with longer, multi-line content. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Feature | Medium |
Single select / Multi select option names support up to 1,000 characters (V2). Single select / Multi select option names support up to 1,000 characters (V2). Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Feature | Medium |
New table data safety limits configurable for fields, views, formulas, records, options, and bulk writes. New table data safety limits configurable for fields, views, formulas, records, options, and bulk writes. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Feature | Low |
Notification Center includes an “All” category showing every message by default. Notification Center includes an “All” category showing every message by default. Source: granite4.1:30b@2026-05-20-audit Confidence: low |
— |
| Feature | Low |
High‑priority notifications (billing, usage) stay visible until manually dismissed. High‑priority notifications (billing, usage) stay visible until manually dismissed. Source: granite4.1:30b@2026-05-20-audit Confidence: low |
— |
| Bugfix | Medium |
Improved login authentication config dialog with adjusted spacing, colors, and layout. Improved login authentication config dialog with adjusted spacing, colors, and layout. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Bugfix | Medium |
More stable Linked record updates; updating with only `id` no longer clears existing display `title`. More stable Linked record updates; updating with only `id` no longer clears existing display `title`. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Bugfix | Medium |
AI field auto-update deduplication merges simultaneous dependency changes into one generation. AI field auto-update deduplication merges simultaneous dependency changes into one generation. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Bugfix | Medium |
Fixed empty values in number formulas to prevent new record failures. Fixed empty values in number formulas to prevent new record failures. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Bugfix | Medium |
Fixed validation when deleting tables to allow moving V2 tables with duplicate linked field names to trash. Fixed validation when deleting tables to allow moving V2 tables with duplicate linked field names to trash. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high |
— |
| Bugfix | Medium |
Fixed Markdown editing in feedback forms; changes save correctly and deleted content stays deleted. Fixed Markdown editing in feedback forms; changes save correctly and deleted content stays deleted. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Bugfix | Medium |
Fixed drag fill in empty tables for more reliable operation without table data. Fixed drag fill in empty tables for more reliable operation without table data. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Bugfix | Medium |
More reliable AI field chained dependencies; downstream fields wait for upstream results. More reliable AI field chained dependencies; downstream fields wait for upstream results. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Bugfix | Medium |
Ask AI prompt dialog no longer blocks chat, allowing easier viewing of AI responses. Ask AI prompt dialog no longer blocks chat, allowing easier viewing of AI responses. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Bugfix | Medium |
Fixed nested filter result counts to show correct records and footer counts in views. Fixed nested filter result counts to show correct records and footer counts in views. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Bugfix | Medium |
Fixed AI field auto-update after table duplication to respond to dependency changes. Fixed AI field auto-update after table duplication to respond to dependency changes. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Bugfix | Medium |
Fixed user avatar display in V2 cards for creator and last editor avatars. Fixed user avatar display in V2 cards for creator and last editor avatars. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Bugfix | Medium |
Fixed `SWITCH()` number matching; numeric fields match equal numeric case values without `ROUND()`. Fixed `SWITCH()` number matching; numeric fields match equal numeric case values without `ROUND()`. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Bugfix | Medium |
Fixed some unusable AI-generated tables; missing physical columns are automatically repaired when possible. Fixed some unusable AI-generated tables; missing physical columns are automatically repaired when possible. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: low |
— |
| Bugfix | Medium |
Chained AI field dependencies wait for upstream results, reducing stale reads and duplicate generations. Chained AI field dependencies wait for upstream results, reducing stale reads and duplicate generations. Source: granite4.1:30b@2026-05-20-audit Confidence: low |
— |
| Bugfix | Low |
Markdown editing in feedback forms now saves changes correctly and retains deletions. Markdown editing in feedback forms now saves changes correctly and retains deletions. Source: granite4.1:30b@2026-05-20-audit Confidence: low |
— |
| Bugfix | Low |
Drag‑fill operations work reliably even in empty tables. Drag‑fill operations work reliably even in empty tables. Source: granite4.1:30b@2026-05-20-audit Confidence: low |
— |
Full changelog
Feature Updates
-
AI Chat supports adding field names: Add field names from column headers to specify target fields.
-
AI Chat supports table context: Add selected rows, columns, or cells from Grid view for better data understanding.
-
AI Chat selections are clearer: Single rows or columns show labels like “Row 1”; click to highlight them.
-
Improved mobile model selection: AI model selection now uses a clearer, easier-to-tap bottom sheet on mobile.
-
AI Chat saves drafts per session: Unsent content is restored after refresh, with drafts kept separate by session.
-
AI Chat input persists across pages: Unsent content remains when switching between App Builder, table chat, and node pages.
-
AI Chat automatically refunds failed usage: Credits are refunded for system errors or empty replies, with records shown in usage logs.
-
Clearer long-chat compression prompts: Context compression prompts appear in the response stream, then generation continues after compression.
-
AI field abnormal usage detection: Potentially excessive Credits usage in AI fields or automation tasks is intercepted for confirmation.
-
More controllable AI field batch autofill: Added batch limits and concurrency controls to reduce abnormal Credits usage risks.
-
Improved AI image generation reliability: Enhanced retry handling and support for multimodal image generation attachments.
-
Notification Center adds “All” category: View all messages by default to avoid missing notifications from previous category filters.
-
Important notifications stay visible: High-priority billing and usage alerts require manual dismissal to prevent missing key information.
-
Admins can send in-app notifications: Instance admins can send custom in-app notifications with longer, multi-line content.
-
Longer Single select / Multi select option names: V2 supports option names up to 1,000 characters.
-
New table data safety limits: Configurable limits for fields, views, formulas, records, options, and bulk writes improve stability.
Fixes & Improvements
-
Improved login authentication config dialog: Adjusted spacing, colors, and layout for a clearer configuration process.
-
Ask AI prompt dialog no longer blocks chat: AI responses are easier to view while asking questions.
-
Fixed Markdown editing in feedback forms: Changes in the expanded editor save correctly, and deleted content stays deleted.
-
More stable Linked record updates: Updating with only
idno longer clears the existing displaytitle. -
Fixed nested filter result counts: Fixed views showing correct records but 0 in UI or incorrect footer counts.
-
Fixed drag fill in empty tables: Drag fill and value clearing work more reliably even with no table data.
-
AI field auto-update deduplication: Simultaneous dependency changes merge into one generation, reducing duplicate tasks and Credits usage.
-
More reliable AI field chained dependencies: Downstream fields wait for upstream results, reducing stale reads and duplicate generation.
-
Fixed AI field auto-update after table duplication: Auto update AI fields still respond to dependency changes after copying tables.
-
Fixed user avatar display in V2 cards: Creator and last editor avatars now display correctly in edit cards.
-
Fixed empty values in number formulas: New records no longer fail when number formulas return empty values.
-
Fixed
SWITCH()number matching: Numeric fields now match equal numeric case values without requiringROUND(). -
Fixed validation when deleting tables: V2 tables can move to trash even with duplicate linked field names.
-
Fixed some unusable AI-generated tables: Missing physical columns from failed schema updates are automatically repaired when possible.
-
Expanded record cards can close on outside click: In Grid view, click outside an open record card to close it.
-
More consistent table safety limit validation: Table creation, view updates, CSV import, duplication, and schema import follow configured limits.
-
Clearer validation for overlong option names: Imports, field conversions, and API typecast writes return clear errors when option names exceed limits.
-
Improved session security: Sensitive session-authenticated operations now validate origins to reduce malicious webpage risks.
-
Overall performance and stability improvements: Optimized high-frequency data access paths to reduce slow requests and high-load timeouts.
Security Fixes
- Improved session security by validating origins for sensitive authenticated operations
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
Related context
Related tools
Beta — feedback welcome: [email protected]