Product

Changelog

What's new in DB AI Magic. We ship updates roughly every two weeks — features, performance work, fixes and security.

  1. v1.7.0

    • Feature

      AI Chat — connection picker dialog

      On a new chat, a modal now prompts you to select the connection and database. Selections flow into the header chip and remain editable until the first message is sent.

    • Feature

      Report widget delete confirmation

      Deleting a widget in edit mode now opens a confirmation dialog with the widget title, preventing accidental destructive clicks.

    • Improvement

      Faster schema fetch for MSSQL

      Switched to information_schema-only introspection on MSSQL connections — typical schema loads dropped from 2.4 s to 380 ms.

  2. v1.6.2

    • Fix

      MongoDB string ObjectId equality

      AI-generated find() filters now correctly cast 24-char hex strings into ObjectId where the schema indicates it.

    • Fix

      Dashboard editor — drag handle hit zone

      Resize handles on small widgets were being eaten by the grid gutter on Safari. Increased the hit slop to 8 px.

  3. v1.6.0

    • Feature

      Scheduled reports — Slack delivery

      Send report exports (CSV, XLSX, PDF) into a Slack channel on a cron schedule, with attachment fallback for files over 25 MB.

    • Feature

      Public dashboard links

      Share a read-only dashboard via signed URL. Tokens are revocable, scoped to a single dashboard, and rate-limited per IP.

    • Security

      Per-employee permission audit log

      Every permission change is now written to an immutable audit log table, exportable as CSV for compliance review.

  4. v1.5.3

    • Improvement

      Server-side pagination for AG Grid

      Large result sets no longer materialise client-side. Pages are streamed over the /query socket as the database returns rows.

    • Fix

      Refresh token rotation race

      Resolved a rare race where two concurrent tabs could each invalidate the other's refresh token within ~50 ms.

  5. v1.5.0

    • Feature

      AI auto-retry up to 5 attempts

      Failing AI queries are now repaired by a retry engine that feeds the database error back to the model with full schema context.

    • Feature

      CockroachDB & Redshift adapters

      Both engines now use the Postgres wire protocol path with adapter-specific reserved-word and EXPLAIN handling.

Subscribe to the changelog by following the RSS feed at /changelog.rss (coming soon), or check the public roadmap for what's next.