New Features

  • Improved transparency for AMA thought process with trace
    Start seeing agent’s thought process right away when you ask a question and follow along its thought process behind the scenes, so reviewers stay in flow and can verify how answers came together.

  • Extraction mode change tracking
    Audit history now includes when an attribute’s extraction mode changes between agentic, computed, spread and manual, giving you traceability for audits and faster root-cause analysis.

  • Improved Citations with filename and page number
    PDF citation now show the source file and page number on every citation to make verification one click and one glance.

  • Visible cells for spread citations
    Reveal the exact spreadsheet cells referenced by spread citations, making it simple to confirm the source of each value.

  • Parallelized storage import
    Import data in parallel to accelerate large migrations and reduce cutover windows for new teams.

  • Supabase package updates
    Upgrade Supabase dependencies for improved performance, security patches and better compatibility going forward.

🎨 UX Improvements

  • Dynamic HTML titles
    Set page titles dynamically across the app so browser tabs reflect the exact screen and entity you’re viewing, making multitasking and tab switching faster.

  • Draggable attribute type rows with expanded extraction modes
    Reorder attribute types with drag-and-drop and choose extraction modes inline, reducing clicks and making configuration more intuitive.

  • Entity approval states and Approved icon
    Clarified approval styling and refreshed the Approved icon so review status is recognizable at a glance across lists and detail views.

  • Spread list default grouping by target attribute
    Default to group by target attribute to surface context immediately, speed scanning and make bulk review more efficient.

  • Computed attribute errors surfaced in UI
    Surface calculation errors directly where you work, so you can spot issues instantly, fix inputs and quickly re-run.

  • Widget and dashboard style cleanup
    Polish spacing, alignment and typography in widgets and dashboards to improve readability and present data with greater clarity.

  • Improved Document page loading state
    Display a clear loading indicator on document pages to set expectations, reduce misclicks and make long fetches feel responsive.

  • Improved Spread citation counts and relocated extraction-mode icon
    Display citation counts for spread attributes and place the extraction-mode icon where it is easier to see, improving scannability and review speed.

🐞 Bug Quashes

  • Synthesis

    • Corrected approval percentage calculation so stats display accurately.

    • Included computed attributes in missing-document checks to flag absent sources.

    • Set empty or manual attributes to Queued status at the appropriate step.

    • Split attribute filling into agentic and computed workflows for greater reliability.

    • Updated max-duration settings to prevent premature timeouts or runaway jobs.

    • Enhanced the DB dump script to improve safety and predictability of migrations.

    • Refactored Excel entity extraction to improve correctness and performance.

  • Cortex

    • Enabled the SQL sync poller in staging to validate sync behavior before production.

  • Flow

    • Enabled the SQL sync poller in staging to validate sync behavior before production.

  • Misc

    • Fixed bun build compatibility for version 1.3.1.

Keep Reading

No posts found