Changelog
What changed.
Audits are versioned: a re-upload after an engine change runs the new checks rather than serving the cached result.
6.3 · Pro and Max deliverables
2026-09-21- Report: a client-ready PDF with your logo and colours — verdict, score, fix-first list, findings by area, model facts, verified rewrites with the reviewer's verdict, accepted risks. Printed by your browser, nothing uploaded.
- Compare: what changed since the previous audit of the same model — fixed, improved, new and regressed findings per object, score and sub-score trend.
- Standards: switch checks off, change severities, add naming conventions and required descriptions/folders/formats, set a quality gate; every audit is judged against your standard and the report says so.
- Waivers: accept a finding as a documented risk; it leaves the score and is listed in the report.
- Documentation export: data dictionary as Excel or Markdown, schema diagram as SVG or PNG.
- Max: auto-documentation (descriptions, folder taxonomy and format strings written for the whole model, applied to your TMDL), impact analysis, cleanup export of unused objects, performance section from a .vpax, refactoring roadmap in phases with effort, regression test pack in DAX.
- Workspace: dark mode, command palette (Ctrl+K), triage keys in Findings, word-level diff on rewrites, audits grouped by model with a score trend. The Saved audits list loads in a second instead of downloading every report.
- Parser: /// doc comments are read as descriptions, so documented models are no longer flagged as undocumented.
- Pricing: Free · single audit 9 € · pack of 5 39 € · Pro 79 €/month · Max 199 €/month, excl. VAT.
- Bring your own Gemini key (Pro and Max): AI audits run on your Google project, unlimited within fair use, billed by Google to you; the key is stored encrypted and removable.
6.2 · Calibrated score, cheaper AI review, phone layout
2026-09-21- Health score recalibrated: one critical costs about 5 points, repeated findings of the same family are damped, and a model with only hygiene findings stays above 85. Scores from earlier audits are not recomputed.
- AI review reuses the model digest across its batches through server-side caching: same output, lower cost per audit.
- Every workspace view works at phone width.
- Guides section and this changelog.
6.1 · Model view rebuilt
2026-09-17- Schema diagram with a tiered star layout: facts in the middle, dimensions beside them, snowflake outriggers next to their parent, relationship issues drawn in red.
- Tables, Relationships and Dictionary tabs: table roles, sources, key columns, findings per table; measures with visual usage, dependencies, folder, format and description.
- Corrected export: selected rewrites are spliced back into your TMDL; manual tasks ship as a JSON plan.
- Overview gains model shape, data sources, report health and audit coverage sections.
6.0 · 363 static checks and an expert-level AI review
2026-09-16- Static checks rewritten as a declarative rule engine covering DAX, model, Power Query, formatting, report and security; every check is listed on the site.
- Semantic checks derived from a manual expert audit: text arithmetic, variant return types, fact tables on the one side, share denominators on fact columns, overlapping period windows, hardcoded years in keys.
- AI review in three passes: every shortlisted object is reviewed against an expert checklist, then a cross-object pass finds root causes and plans rewrites, then each rewrite is verified against the model and challenged by an independent reviewer.
- Asynchronous audits with stage progress; audits keep running if you close the tab.
- Pricing: static audit free and unlimited, one AI audit for 9 €, Pro subscription.
- Legal notice, terms, privacy and data pages for a French micro-entreprise.