Changelog. What shipped, when.

v5.25 rounds out the colour picker into a proper hybrid — a 24-swatch palette, the spectrum wheel, a hex field, and recent colours in one popover — and swaps the grid on/off button to an eye / eye-off glyph so it reads distinctly from the graph↔isometric type switch in the zoom cluster.

v5.24 relocates the sketch grid on/off toggle into the bottom-left zoom cluster and adds a square↔isometric background switch beside it. The grid type is a property of the sheet (persisted), and the drawing snap follows whichever grid is active. On/off stays a per-view preference; existing sketches default to the square grid and are unchanged.

v5.23 adds a line-type picker (solid / dashed / dotted / dash-dot / center / phantom) that applies to any stroke-bearing mark and carries through to exports, and a Spline tool that smooth-fits your freehand trace into an open curve (rides the pen family; no new mark type). Both are web-only for now and leave every existing markup unchanged.

v5.22 replaces the three-pattern hatch row with a single chip that opens a labelled picker of ~12 industry-standard construction material patterns, each shown as a live swatch in the mark's own colour. Applies to any fillable shape (and the dedicated Hatch tool), in the viewer and in exports. The earlier generic diagonal / cross-hatch / dots patterns remain valid, so shapes hatched before this update are unchanged.

v5.21 adds multi-select to the markup editor: Shift/Ctrl-click to toggle marks in or out of a selection, or drag a marquee box to grab everything inside it. A group shows each mark outlined with a bounding box; you can drag it, recolour it, or delete it as one action, and undo/redo treats the whole group as a single step. Single selection and everything else in the editor are unchanged.

v5.20 adds hatch fill patterns (diagonal / cross-hatch / dots) to the Drawings & Sketch editor, in the mark's own colour, both as a Fill option on any fillable shape and as a dedicated Hatch tool. Patterns render in the viewer and carry through to the exported PDF/PNG. Existing fills and un-hatched shapes are unchanged.

v5.19 upgrades the markup colour control from seven fixed swatches to a full picker: a spectrum and a hex field for any colour, plus your recent colours remembered across sessions. Dragging the spectrum previews live on a selected mark and records a single, clean undo step when you commit — not one per shade.

v5.18 polishes the sketch tools: a first-class ellipse/circle tool (Shift constrains to a circle, with fill support like the box tool), a per-sheet grid show/hide toggle that also applies to the exported PDF/PNG, a real title block (name + scale + date) on sketch exports in place of the old edge caption, and the photo-pin tool suppressed on sketch sheets. All changes are scoped to sketches or additive — PDF plan sheets are unaffected.

v5.17 adds a proper undo/redo history to the web Drawings & Sketch markup editor (it previously had none — every mark saved to the server as you drew it). Undo replays the inverse of each change against the server: a draw is undone by deleting it, a delete by re-creating it, and a move/restyle/text edit by restoring the prior value. History is per-sheet, holds your last 60 changes, and each step is only recorded once the server confirms it. Linked pins that create tracker items (RFI/punch/task) are intentionally left off the stack so an undo can't silently reverse a tracker record.

v5.16 brings the Drawings markup editor in line with the photo editor: the tool rail is now a horizontal top bar that wraps instead of scrolling, with contextual style controls (only the active tool’s width/fill/text options show). A first-class Select tool joins Pan — click a mark to move, resize, restyle, or delete it — and text notes wrap onto multiple lines rather than overflowing the sheet (carried through to PDF exports too). It also fixes a boot-time reference error (_externalGateThen was defined inside one auth path but called from a second one) that surfaced the Error-Log affordance on load without affecting any data.

A follow-up to v5.14: three more dialogs were rendering behind the full-screen photo viewer or drawing sheet (the photo-delete confirmation, the drawing text/pin label editor, and the new dimension-label editor). All now float above the viewer, so they're visible and clickable. The photo-delete one mattered most — its confirmation was hidden, so a stray Enter could confirm a delete you couldn't see.

v5.14 adds a scaled sketching surface inside Drawings: a blank sheet carries a real drawing scale, snap-to-grid, dimension marks that convert to true dimensions, and a scale bar on export — so a hand sketch behaves like a measured drawing, takeoffs included. It also fixes three markup-editor issues where the confirmation and text dialogs were rendering behind the full-screen editor (or the photo lightbox underneath it): the Text/Callout tool no longer kicks you out, the trash clears markup, and hold-to-save presets work.

v5.13 brings the web photo-markup toolbar in line with the field app: the Width / Fill / Text / Background controls are now contextual to the selected tool (or the selected mark) instead of always showing, and the range sliders were restyled with a clear track and a high-contrast handle so they read as controls, not decoration.

v5.12 brings the RFI and Schedule sharing surfaces up to the same management bar as bid invitations: a status strip (sent / viewed / responded / expired / revoked) with Remind and Revoke on RFI review links, and a Share-progress-link control on the Schedule with a manage panel to copy or revoke each link. Revoking is immediate and can be done by any project owner.

v5.11 makes the annotation preset slots discoverable: a visible hint that they're customizable, and a hold-to-save gesture that previews the tool, color and width it will keep before you commit. On the platform and the field app, drawings and photos alike.

v5.10 puts the discussion where the decision gets made. Every mention rides the same notification pipeline as assignments, so a question directed at you reaches you the same day — and the whole thread stays on the record for the next person who opens it.

v5.9 cleans up the web markup editors: photo text entry renders above the editor, the color/fill/width controls apply to a selected mark (not just the next one you draw), and Shift-drag on a resize handle preserves aspect ratio. No saved markup was ever affected — these were input and display fixes only.

v5.8 rebuilds the drawing and photo markup feel around the jobsite: palm rejection, stroke stabilization, shape-snap, angle constraints, and quick undo — all on the same annotation format that renders identically on the app, the platform, and in PDF exports. Dimension labels reuse the measurement engine, so they stay true to scale.

v5.7 makes assignment real: whoever you hand work to now hears about it, in the app and by email, from every assignable tracker. Under the hood the fan-out runs at the database layer, so it catches every write path — web, field app, and AI-created records alike — and can never slow down or fail a save.

v5.6 closes the loop between planning and the field. Progress entered in the daily log flows to the schedule, its linked records, and the risk forecast in one motion; drawings downloads are explicit and honest about what's on the device; and anything a camera captures can become an owned, assigned record in two taps.

v5.5 brings the field and the office to the same picture. Photo and drawing annotations share one format across mobile and web, so a markup made on either surface renders the same on the other and in PDF exports. The schedule gains trade and CSI-division grouping end to end — Gantt, look-ahead report and CSV. And the field app picks up the edit actions that mattered most: RFI status and assignment, schedule ownership, and meeting action-item completion and reassignment.

v5.4 closes the loop on meeting accountability: assignees set during minutes review carry a real user identity, flow into promoted tasks, and show up in each person's My Work. Under the hood, permission checks across the API now resolve in the organization that owns the resource being changed.

v5.3 is the accountability release. Assignees are wired end to end — picker, database identity, and the My Work dashboard — across every work-item tracker, with Mine filters to match. The schedule closed its loop: the Monte Carlo engine's per-activity results persist to the plan, appear on the Gantt, and connect procurement to the activities it feeds. Escape-key dismissal now reaches every dialog on the platform and always routes through each dialog's own close logic, so cleanup behaves exactly as if you had clicked Cancel.

v5.2 makes the granular permission model for teams enforced end to end. Every Finance, Project and Sales capability is now checked on every action and every screen — a read-only viewer can browse but never modify, and the maker/checker controls on pay-app certification, payment approval, GL posting and lien-waiver verification hold across all paths. We also tightened the isolation between organizations so each company's data — including the documents and photos the AI assistant retrieves — stays strictly its own. Define a role once and trust that the platform enforces it everywhere.

v5.1 sharpens the meeting recorder. Extracted action items, decisions and risks are now anchored to the moment they were spoken, so you can replay the exact context with a tap; long meetings produce tighter, de-duplicated notes and finalize reliably end to end; and the recorder is now available as a native Mac app for Apple Silicon in addition to Windows. Housekeeping: recordings that capture no audio are cleaned up automatically instead of cluttering a project's Meetings list.

v5.0 makes the General Ledger we shipped in v4.5 actually tie out. Job costing is now accrual-based and reconciles to the books to the penny: deliveries book cost the day they land, deposits and partial shipments are handled correctly, and a one-click year-end close rolls the year into retained earnings. Vendor payments and pay-app reversals restate cleanly, the procurement lifecycle drives financial status automatically across all 22 trackers, and every financial surface now degrades gracefully instead of going blank. Next: two-way QuickBooks Online sync so the ledger Trueleveler keeps flows into the accounting system your bookkeeper already lives in.

v4.5 is our biggest financial release since launch. Trueleveler now keeps a real, double-entry General Ledger under every job — with statements, CSV/PDF exports, WIP↔GL reconciliation, and the maker/checker and closed-period controls that sureties and lenders expect. Alongside it, a dedicated Sales workspace turns the pipeline board into a full pursuit-to-award command center: bid room, estimate model, branded proposals, forecasting, CRM and an AI coach — with a clean hand-off that seeds the job budget the moment a deal is won. The release also closes a round of cross-tenant security gaps. Next milestone: making this ledger tie out under load — accrual job costing and year-end close (shipped in v5.0).

v4.0 is the marketing-and-app modernization release. The marketing site went from inconsistent (different headers, different brand styles, different CTAs across pages) to coherent — one nav system, one logo, one signup flow. The five P0/P1/P2 engine pages went from teaser-stubs to confident product pages with real mockups and FAQs. The app side closed three structural backend gaps (133/134/135) and fixed two annoying UX bugs (the Results header flash, the cryptic task source chips). The build process itself is now durable — the source files match the canonical taxonomy so future builds don't silently undo work. Next milestone (v4.1+): the Phase 1 financial-core sequence — Project Budget tracker, Job Cost rollup, Owner Pay App Drafter — plus the CO outbound input and T&M ticket tracker that round out the change-order story.

v3.10 is the design-handover closure release. Every PDF template from the Claude Design v2.0 batch is wired, every tracker has a standardized Export dropdown, and the live marketing site is reconciled to the canonical "26 engines + 15 trackers" taxonomy. Two Supabase migrations (131 + 132) close the historical_bids audit gap and fix a latent benchmark-RPC type mismatch. 13 new help articles ship in /help/ covering the new engines + workspace views. The remaining design-handover work (12 P1 engine-page refreshes, OG image generation, marketing video) is the v3.11 backlog.

v3.9 follows v3.8 on the same day — the engineering marathon kept rolling into a focused content + security batch after the major engine work landed. The resource-page additions complete 17 of the 18 pages on the original SEO content roadmap (the 18th, Modular/Prefab Procurement, sits in the backlog as a niche P3). Migration 131's historical_bids lockdown was a deferred item from the May 9 audit; the prior migration (129) noted that the benchmark RPC needed cross-tenant visibility but didn't recognize that SECURITY DEFINER already bypasses RLS — this release corrects that gap by tightening the policy to org-only while leaving the aggregate benchmark unchanged.

v3.8 is the largest single release since launch — 36 commits introducing the Negotiation Intelligence Module (the biggest revenue lever on the post-audit roadmap), modernizing Change Order Review into a server-side engine with full tracker integration, and finishing the migration of every engine result panel onto Claude Design's PDF template system. Five Supabase migrations (126: rfi_citations · 127: CO analysis columns · 128: negotiation schema · 129: org-aware RLS · 130: NUMERIC widening for mega-project COs) shipped alongside. Cross-engine reads (Bid vs Actual, Schedule Risk, Daily Log Synthesis, Pay App Review) all caught up to the live column names after a follow-up audit surfaced three silent failure modes. The Negotiation Module's "closed-loop" story spans Bid Leveling → new negotiation session pre-filled with the recommended vendor → AI coaching with vendor profile history → close-won feeds the next negotiation's coach.

The v3.7 release brings the app surface in line with the landing page identity. Every shell element — sidebar, header, modals, charts, breadcrumbs — now uses the Blueprint Amber tokens, mono section labels, and the unified focus ring system. Skeleton loaders smooth out the dashboard reload moment, and the AT A GLANCE stat grid is locked to a deterministic 2-column layout so wide values like savings totals never clip out of the box. A new long-form resource on subcontractor prequalification joins the guide library. SEO got a structured-data pass: rich-result markup landed on the Platform Atlas, sitemap orphans closed, and meta descriptions tightened to within Google's recommended length. Internally, the design audit's 5-phase token migration completed (legacy <code>--blue-*</code> aliases moved to <code>--accent-amber-*</code> across the codebase) so future contrast and theme adjustments have a single source of truth.

The v3.6 release focuses on closing user-reported gaps and hardening the platform. Beta testers asked for Schedule Impact in Change Orders and tracker-level PDF exports — both shipped. Procurement import friction, a hidden import button and a non-functional "Open in Sheets" button, is resolved. A large accessibility sweep brings the app within reach of WCAG 2.1 AA: every modal is announced by name, every sidebar button is labeled, the language picker is fully keyboard-accessible, and focus rings are visible on dark backgrounds. Under the hood, plan-limit enforcement was unified across three API endpoints, GDPR data export scales to 50,000 rows per table, and photo-heavy PDF reports now gzip and auto-downsample so they fit within serverless payload limits.

The v3.5 release makes Trueleveler team-ready. Team members share a unified view of all project data through organization-level access. The split dashboard gives executives a bird's-eye view while PMs get project-specific alerts with actionable next steps. Health scores now compute automatically across all project views. Mobile support covers all screen sizes from large monitors down to small phones.

The v3.4 release is focused on closeout workflows and project health visibility. The Punch List tracker was rebuilt with professional forms, photo management, and shareable reports. The Pulse health score was expanded to 10 dimensions that pull live data from all 5 trackers, and AI recommendations now generate specific, actionable alerts instead of generic summaries.

The v3.3 release focuses on speed, polish, and accessibility. Document analysis runs faster through smarter text processing. The progress bar provides smooth visual feedback during long analyses. Security was hardened across all endpoints, and accessibility was improved to meet WCAG 2.1 AA standards.

The Pay Application Review engine now renders a full financial controls panel alongside the line-item table. Retainage, cumulative billing, and lien waiver data were already being extracted by the AI but were not displayed — this release brings them to the surface as structured panels with colour-coded status indicators. Front-loading patterns, retainage discrepancies, and missing waivers all trigger prominent warnings. AIA compliance, SOV overbilling flags, stored materials claims, and prevailing wage requirements are surfaced as status badges above the red flags section.

This release adds five new post-award AI engines covering the entire construction project lifecycle from COI compliance through project closeout. Every scope check now produces a structured bid/no-bid recommendation with confidence score and itemized reasons. Trust signals were added to all engines: a confidence bar and document volume indicator. Bid Leveling was expanded to support up to 10 simultaneous vendor comparisons.

The v3.0 release focused on security, UX polish, and new tools. Document processing was hardened with input sanitization. The tariff calculator provides a free resource for estimating material import costs.

This release focused on production readiness. Security was hardened across all endpoints, translation coverage was expanded to 10 languages, and a full audit trail was added for compliance requirements.

The v2.0 release marks the feature-complete beta milestone with 26 AI engines covering the full construction project lifecycle. The submittal tracker manages a 7-phase workflow from extraction through approval. Procurement tracking supports CSI MasterFormat codes with purchase order and invoice matching.

This content release added 13 in-depth resource pages covering construction insurance, lien waivers, bid templates, incoterms, material price tracking, RFP/RFQ guides, cost estimating, contract types, supply chain management, vendor management, material takeoffs, and payment processes. Blog content was localized for the US, UK, Germany, France, Spain, Norway, Sweden, Denmark, Netherlands, Italy, and Poland. All pages include JSON-LD structured data (Article, FAQPage, BreadcrumbList schemas) and hreflang tags for multi-language discovery.

The initial release of Trueleveler launched with five core AI engines: Bid Leveling, Contract Review, RFQ Generator, Document Compare, and PM Contract Summary. Authentication is powered by Supabase with row-level security policies ensuring data isolation between organizations. The interface supports both dark and light modes with full responsive behavior across desktop, tablet, and mobile viewports. New users receive 5 free AI analyses to evaluate the platform before subscribing.