Taufinity Studio

Appearance

Release Notes ​

Updates to Taufinity Studio are released regularly. This page lists what changed in each release, focusing on what it means for your workflow.

July 14, 2026 ​

New ​

  • Secure Embedding — Enterprise organizations can now embed Taufinity Studio's full portal directly inside their own product — dashboards, playbooks, content, and more — without sending users through a separate login screen. Server-to-server authenticated via your own API key, with a single-use claim token and no shared passwords or sessions. Contact us to enable it for your product. See Secure Embedding.

July 12, 2026 ​

New ​

  • Enterprise SSO — Enterprise organizations can now let their team sign in with their own identity provider (Microsoft Entra ID, with support for any OIDC-compliant provider) instead of a Studio password. New sign-ins are automatically provisioned and go through an admin-approval step, with admins notified by email as soon as someone's waiting. Contact us to enable it for your organization. See Enterprise SSO.

June 2, 2026 ​

Improved ​

  • Smarter query suggestions — Ask Data now auto-corrects unknown column names and returns clearer error messages when a question can't be answered with the available data. Fewer dead-ends, more useful feedback.
  • Always-warm service — Studio is now kept warm with a minimum of one running instance, eliminating the occasional cold-start delay on the first request of the day.

June 1, 2026 ​

New ​

  • Vision describe step — Playbooks can now automatically describe images and attachments using AI vision. Attach a photo or document and the playbook generates a description that flows into subsequent form fields — useful for on-site inspections, damage reports, and media-rich intake forms.
  • Sanity review step — A dedicated safety pass runs a second, independent AI model over playbook output before it leaves the pipeline. Catches factual contradictions, hallucinations, and format errors the primary model might miss.
  • Knowledge-driven rules engine — Playbooks can look up facts from your knowledge base and evaluate business rules against them. Rules are resolved in dependency order so complex multi-step logic works correctly.
  • Source provenance chips — Every AI-filled field now shows a small chip indicating whether the value came from a verified fact, a derived calculation, or an AI inference. Hover to see the exact source fragment, so you always know why a field was filled the way it was.
  • Formula step citations — When the formula engine resolves prices or numerical values, it emits a citation linking the result back to its source fact in the knowledge base.
  • Per-attachment AI status — While a playbook reads an attachment, each file card shows an indicator so you can track processing progress at a glance. Toggle individual attachments to control whether AI extraction runs for that file.

Improved ​

  • Playbook layout — The Run tab now shows notes and attachments side-by-side in a hero layout, with the form below. Multi-fill animations are staggered across field groups, and a sticky banner appears when AI fills happen below the visible area so you never miss a change.
  • Group progress chips — The Run tab shows progress chips per field group instead of a raw field count, making it easier to see where you are in a long form.
  • Accessibility — Fill animations respect reduced-motion preferences. Timer cleanups and focus management improved throughout the playbook UI.
  • Per-user and per-org rate limits — Rate limiting is now enforced at both the user and organisation level for playbook pipeline runs.

May 29–31, 2026 ​

New ​

  • Live auto-fill from notes — Type or paste free-form notes in the playbook Run tab and watch the form fill itself in real time. The AI extracts structured values as you type and merges them into open fields — no submit required.
  • File and audio attachments on playbook runs — Drop images directly onto the Run tab or record audio from within the browser. Attachments are uploaded, transcribed (audio), and fed into the playbook pipeline alongside your typed notes.
  • Glow animation on AI-filled fields — Fields that were just filled by AI pulse briefly with a glow so you can see exactly what changed. Only fields that actually received new values animate — unchanged fields stay still.

Improved ​

  • Signup page — The self-service signup page now uses the same branded portal layout as the login page, giving new users a consistent first-time experience.
  • Version history readability — Version diffs in playbooks, widgets, and routers are now formatted and pretty-printed, making AI-generated change summaries easier to read and compare.

May 27, 2026 ​

New ​

  • Customizable navigation — Reorder and pin sidebar items to match your workflow. Admins can define navigation profiles with role-based defaults; individual users can override pins. Open the new Navigation settings page under Settings → Navigation.
  • Published article URL — The article API now returns the live published_url so integrations can link directly to the published version without constructing the URL manually.
  • Render from preview branch — Articles can now be rendered using a template from a preview or feature branch, letting you validate new templates against real content before merging.
  • Dashboard data freshness — Dashboards now show when their data was last updated ("laatst bijgewerkt"), so you always know how current the figures are before acting on them.

Fixed ​

  • Dashboard filter chips — Filter definitions with display aliases now merge correctly into filter chip labels.

May 26, 2026 ​

New ​

  • Approval workflow notifications — When a playbook run requires approval, the responsible coach receives an email notification and an in-app alert. A "Send to client" button in the document editor triggers the approval flow directly from the run view.
  • Client group portal identity — Portal users are now properly identified by their client group context throughout the session, fixing permission edge cases in multi-tenant portals.

Fixed ​

  • API key org context — API keys sent as Bearer tokens now correctly inherit the organisation context, fixing authentication failures for cross-org integrations.
  • Impersonation auto-expiry — Lingering admin impersonation sessions now expire automatically instead of blocking normal login for affected users.

May 19–22, 2026 ​

New ​

  • Encrypted share links — Sharing a widget or playbook run now generates a short, opaque link. The API key and endpoint URL are never exposed in the browser address bar or server logs.
  • Form-driven playbook Run tab — Playbook intake forms now render with full field grouping and collapsible sections. Required fields and progressive disclosure make long forms manageable.
  • Notes-fill step — A new step reads free-form notes from the intake and fills structured fields using AI, without requiring you to map inputs manually.
  • AI model fallback — When the primary AI provider is overloaded, Studio automatically retries with a fallback model instead of failing the request. Scheduled jobs are rescheduled rather than dropped.
  • Image classification step — A new playbook step uses AI vision to classify and tag images — useful for document type detection, damage severity scoring, and content moderation.
  • Knowledge-file lifecycle events — Creating or updating a knowledge file now emits events that can trigger downstream playbooks automatically, enabling reactive pipelines without manual scheduling.
  • SSH deploy keys — Private Git repositories can now be connected via SSH deploy key, removing the need for personal access tokens or shared credentials.
  • Widget source routers — Widgets can now automatically route conversations to different knowledge bases or playbooks depending on the source of the request, configured without hardcoding IDs.

May 13–18, 2026 ​

New ​

  • Version history — Playbooks, widgets, routers, and dashboards now have a version history. Open the History panel to see a list of past versions, preview the diff, and revert to any earlier version with a single click.
  • AI cost ledger — Studio tracks AI token costs per organisation in real time. Admins can view cost attribution per org from the admin panel.
  • Provision tool — Customer Studio configuration (BQ providers, site pipelines, portal branding, KPI definitions, secure render) is now managed via the provision CLI tool. Config lives in version-controlled YAML files and is applied idempotently.

Improved ​

  • Audit trail — Tool audit events now include the user and API key that triggered each action, and requests are classified by client type (browser, CLI, MCP client) for better usage analytics.
  • Provision: knowledge base sync — The provision tool now verifies each knowledge file after upload and supports --strict mode (fail fast on any error) alongside the default skip-and-warn behaviour.

Fixed ​

  • Security gate for pending organisations — Organisations pending approval or that have been rejected can no longer trigger cost-incurring operations until approved.

May 1–12, 2026 ​

New ​

  • Admin user impersonation — System administrators can now switch into any user's session for support and debugging. Sessions are time-limited and fully audited.
  • Pinned dashboard filters — Filters you set on one dashboard stay pinned as you navigate between dashboards in the same session. A pin icon appears next to each filter so you can see and clear what's pinned.
  • Share link wizard — A guided 5-step wizard in Widget Settings walks you through creating a share link: choose permissions, select a playbook, set an expiry, and copy the link — no manual URL construction needed.
  • Orchestrator tool loop — Widgets configured as orchestrators can now dispatch sub-agents in two phases (plan then execute), enabling more structured multi-step agentic workflows.
  • Webhook response storage — The webhook_action playbook step can now store a specific JSON field from the webhook response in a variable, making it easy to chain webhook calls and use returned data in later steps.

Improved ​

  • Query insights suggestions — Ask Data now returns result-grounded follow-up suggestions based on what the data actually showed, and maintains conversation context within a session so follow-up questions narrow rather than restart.
  • Tab-conditional filter visibility — Dashboard filters can now be shown or hidden depending on the active tab, reducing clutter on multi-tab dashboards.

April 20–30, 2026 ​

New ​

  • Line, scatter, and pie/donut charts — Dashboards now support line charts (trends over time), scatter/bubble charts (two-axis comparison with optional bubble size), and pie/donut charts (share breakdown). Chart type is configured per tile.
  • KPI comparison cards — KPI cards can now show a comparison value (prior period, target, or benchmark) with a directional arrow and sparkline trend. Click a card to filter the dashboard table to matching rows.
  • Dashboard PDF export — Export any dashboard to PDF directly from the UI, producing a static document suitable for sharing or filing.
  • Grand totals row — Dashboard tables now show a grand totals row at the bottom, with section-level sort support for badge tables.
  • Tabbed dashboard layout with shared filters — Multiple dashboard tabs can share a single filter bar. Switching tabs preserves filter state; filters not applicable to a tab are hidden automatically.
  • Article slug transliteration — Accented characters in article titles are now transliterated to ASCII when generating URL slugs (é→e, ö→o), producing cleaner, more portable URLs.
  • Directory-style deployment paths — Deployment settings now support path prefix and directory-style options, enabling /article/ style paths instead of flat /article.html paths.
  • Faster AI responses — The default AI model for all widgets has been upgraded to Gemini 2.5-flash, delivering faster and more capable responses across content generation, chat, and insights.

Improved ​

  • OpenAI-compat passthrough system messages — The OpenAI-compatible endpoint now accepts passthrough system messages, giving integrators more control over AI persona when wrapping Taufinity widgets in their own tooling.

April 14–19, 2026 ​

New ​

  • Daily usage quotas — Organisations now have configurable daily caps on playbook runs, job runs, and total job runtime. When the daily limit is reached, tools and integrations are told it's a quota limit — not a temporary error — so they don't keep retrying automatically. A self-service quota page at Settings → Quotas shows current usage and blocks. Admins can set per-org overrides.
  • Playbook issues column — The playbooks overview now shows a green/yellow/red issues badge per playbook. Hover for a tooltip with the top recent failures or quota blocks, so problems are visible without opening each playbook.
  • Interactive dimension selector — Dashboards can now include a dimension selector that lets you switch the grouping interactively — for example, switching a table between "by team" and "by region" without changing pages.

Improved ​

  • Widget preview audio — The in-Studio chat preview and embed preview pages now support audio recording, matching the full widget experience.

March 31, 2026 ​

New ​

  • Email domain restriction — Administrators can now restrict which email domains are allowed to log into their organisation. Useful for organisations that want to ensure only company email addresses can access their Studio.

March 26, 2026 ​

New ​

  • Reports & Dashboards — Interactive dashboards with KPI cards, stacked bar charts, and data tables powered by BigQuery. Click chart segments to drill down into filtered table views. AI-generated insights analyze your data automatically using playbooks — cached by content so repeated views load instantly.
  • Whitelabel Portal — Access Taufinity Studio on your own branded domain with custom logo, colors, title, and favicon. Google SSO works seamlessly across portal domains with secure token-based session handoff. Available on *.taufinity.io subdomains; custom domains coming soon.
  • Playbook Lock — Lock playbooks to prevent accidental edits while they're running in production. Unlocking requires explicit confirmation.

Improved ​

  • Dashboard drill-down — Charts and tables are linked: click a bar chart segment to filter the table below. Filter state is preserved in the URL for sharing and browser navigation.
  • Playbook run summaries — Playbook runs now show created task counts in the summary view, making it easier to see what each run accomplished.

March 18, 2026 ​

New ​

  • Ask Data (text-to-SQL) — Ask natural language questions about your data directly in the widget chat. The AI translates your question into SQL, executes it against your BigQuery provider, and returns the results. Your data is always filtered to your organisation — cross-tenant access is not possible. Enable the ask_data tool in your widget's Agent settings.

March 13, 2026 ​

New ​

  • OpenAI-compatible API — Use any OpenAI SDK or tool to interact with your widgets. Point your existing code at /v1/chat/completions with model widget:<uuid> and it works out of the box, including streaming support.
  • Scheduled playbooks — Run playbooks on a recurring schedule (hourly, daily, weekly). Set a cron expression, pick a timezone, and playbooks execute automatically. Auto-pauses after repeated failures to prevent runaway costs.

Fixed ​

  • Chat conversation compaction — Fixed an issue where conversation compaction could fail with a "not found" error.