Dear @Community
This is the biggest release we have shipped since Real-Time LCA launched.
Projects are rebuilt around a single left rail. Every change you make is now a version you can restore. Milestones can be saved as named revisions and returned to. Reports have a proper builder. The AI Agent can do the mapping for you — and asks before it writes anything. And the AI can read an EPD PDF straight into the Library.
The entire user guide has been rewritten against this release. Let’s dig in.
A new project structure
Every project now shares one left rail — Dashboard, 3D model, Materials, Transport, Waste, History, Reports, Settings — so the same navigation follows you from an empty project through to a delivered report.
Creating a project is now a short dialog: country, name, building type, calculation type, and the models the project will hold. Everything else moved into Settings.
One Settings page
Project, building area, calculation, models, classification, transport (A4), installation (A5), energy use (B6), members, and integrations now live on one page with one Save.
- A Sections panel is both a table of contents and a status list — complete, edited but not saved, or needs attention before the configuration can be saved — with a plain-language count underneath (1 of 8 sections needs attention before this configuration can be saved).
- Your limit value shows the budget the project is judged against, and where it came from.
- Manage holds the project-level actions: integrations, members, clone, delete.
History — every change is a version
Changes to base information, materials, and mappings are now recorded as versions. Each entry carries its version number, what changed, who changed it, when, and where it came from — in-app, service, or an integration. Filter by change type, search the log, limit it by date, or export it.
Restore never overwrites your project. It builds a new project from the state at that version — materials and project information, transport and waste mappings, custom emission factors, and the history up to that point — and leaves the original exactly as it was.
Revisions — named save points
Any version can become a named revision: Initial Design, Detailed Design, Planning submission. The project is always on exactly one revision, and everything you do is saved to it — it never drifts away from the revision it is on.
- Preview opens a revision read-only, for any user. Continue from this moves the project’s working state onto that revision, and every open client follows live.
- Each revision shows who saved it, when, and the emission figure at that point — with the change from the previous revision beside it (0.13 kgCO₂-eq/m²/year, −0.1). The delta between revisions tells the story of the project on its own.
- The revision name in the breadcrumb is a switcher, available on every page of the project.
- Reports are pinned to the revision they were taken from. That is what makes a delivered number reproducible: months later you can preview the revision a report was issued from and see exactly the figures behind it, even though the live project has moved on.
Save a revision at every milestone you report on. It costs nothing.
Clone, rebuilt
Cloning now copies the project’s current state — materials, mappings, settings, team, and its history so far — into a brand-new project, leaving the original untouched. The dialog names the exact state being copied, and asks what should happen to the 3D model: a new Speckle project for the clone, or the same published models feeding both.
Pick a new project when the variant will diverge geometrically. Pick the same one when only the assessment differs — swapping EPDs, testing a different calculation type, trying alternative transport distances against one model.
The report builder
Reports in the project rail opens a builder: a page list, a live preview of exactly what will print, and the three controls that apply to the whole document — a mapping incomplete warning, a marker for the pages that take your input, and the report language.
- Cover — project name, headline result, your client logo.
- Overview — your project description, a slot for a building render or photo, the QA/QC fields, and which life-cycle modules the result covers.
- Results — the calculation method and every project parameter behind the number, then the total against the limit value.
- Additional Results — ranked impacts and the cumulative curve, each with a note box beside it. The chart shows what happened; your note says why it matters.
- Background Data — the active standard values and every EPD used, with its declaration number. The page an auditor turns to first.
Which reports a project offers depends on its country: every project gets the LCA Report PDF, plus the Excel reports its country’s scheme or authority expects. Anything you have generated before stays listed and reachable.
The AI Agent
The agent opens beside whatever you are working on and picks up the project you are in.
It can map for you. It resolves the materials, reports what is blocking them, finds candidate EPDs, and states what it is about to do — which EPD, for which materials, and why — then waits for a yes. Anything that changes data pauses for Approval needed, naming the exact operation, with Allow, Deny, and every tool call it has run visible above the prompt. Afterwards it summarises what changed: the EPD used, the lifetime it set, how units were converted, and how much mapping coverage went up.
[[IMAGE: images/user-guide/screens/AI Agent approval needed.png]]
Every one of those changes is an ordinary entry in History — and can be restored.
It answers product questions from the documentation. “How do I…” questions are answered from the guide, with a Sources list linking the pages it drew on.
Skills are reusable instructions you can attach to a message — hotspot analysis, comparing relevant EPDs, mapping a material. Curated skills are maintained by us; duplicate one to adapt it, or write your own from scratch.
Usage and History track tokens and sessions across the workspace, by month and by feature. Both export.
Extract From PDF
On the Create Material form, Extract From PDF reads an EPD document and fills the form for you — name, life time, owner, data source, dates, EPD type and specification, declared and mass values with their units, and the indicator tables. Roughly 15–60 seconds depending on the document, and anything it could not find stays empty and flagged, so the gaps are easy to spot before you save.
Across roughly 10,000 values from 35 test files, extraction returned 98% of the values correctly, with 11 individual entries wrong — the rest were missing or flagged for review. That comes from a neuro-symbolic approach: generative AI does the reading, symbolic rules reason over the output to catch hallucinations.
It is an accelerator, not a substitute for checking. A wrong indicator value is invisible once the material is in the Library — it just produces a wrong answer in every project that maps to it. Check the declared and mass units, Valid to, type and specification, and spot-check A1–A3 before you save.
AI Classification, reworked
From Mapping tools on the Materials page, pick the building components to classify — Only unclassified hides anything that already has a code — and review the proposals in a table: existing code, object class, proposed code, whether it is a change, an AI confidence score, and a per-row Apply tick. Object class and proposed code are both editable, so you can correct a proposal instead of rejecting the batch. Accept writes the ticked rows; Reject discards the lot. What you accept is recorded in History.
Mapping
- An empty project offers four ways to start: upload an IFC model, upload an Excel file, add a material or type by hand, or load a set of sample building elements to explore the workflow. The New menu in the inventory header offers the same routes at any time.
- Drag an EPD or construction onto a row, onto a group, or onto the strip at the top to drop it into a new group. Mapped rows turn green and show the EPD beneath the element name.
- The flag button limits the library to EPDs allowed for the project’s country; the funnel filters by data source, category, and validity.
- Auto Mapping now lands as a preview — rows tagged Auto Mapped, the header swapped to Accept or Reject Auto Mapping, nothing written until you accept. Drop a different EPD onto anything that looks wrong first.
- Missing a lifetime? The app asks for it there and then, with the lifetime table one click away.
Transport and 
Waste
Transport is one row per mapped material — how many items use it, total tonnage, emission per tonne, and the resulting total — with the vehicle type and distance you set. Defaults come from the country’s data, set once for the whole project in Settings → Transport (A4), so here you only adjust the exceptions.
Waste lists your scheme’s categories on the left and your mapped materials on the right, each with the emission it contributes, the wastage assumed on top, and the mass available to sort. Assign sends a material to a category and shows the fraction percentage it represents.
The Library
- New in the toolbar creates a material or a construction.
- Rows now carry EPD type, standard, dataset, declared unit, declaration number, and headline GWP — plus Valid to and an Expires soon warning when an EPD is close to its end date.
- Compare up to six materials side by side: an attribute overview, and a per-module Lifecycle Stages tab where you pick a calculation type (Full LCA (A–D)) or toggle individual stages. The comparison downloads.
- Materials show the collections they belong to as a chip on the row, so a saved material is recognisable while you scroll.
- Constructions get Edit, Clone, and Delete from the row menu, filters for units, owners, and organisations, and a dimensions dialog with a 3D preview, a λ-value for insulation, and the unit conversion spelled out.
Tutorials, inside the app
A library of short videos at the bottom of the left sidebar, reachable from any page. Cards show the category and duration, a tick marks the ones you have already watched, and chips filter by category.
The best part: open the player’s ⋮ menu and choose Picture in Picture. The video pops out into a small floating window that stays above the app, with its own play, skip, and progress controls — so you can do each step as you watch it. Playback speed is in the same menu.
Speckle and imports
- You now create your own Speckle account from Account menu → Speckle, choosing the password you will use in the connector. The dialog also holds Reset password, Sync projects, and the connector download — and the server address to sign the connector in to.
- A project’s connection is configured from Settings → Integrations: a new connection for its own stream, or a shared connection so two projects read the same published model. Once connected, the row reports the models in sync, when they were last published, and offers Sync now.
- Excel and IFC uploads moved to the inventory’s New menu. Imported rows are tagged with their source, the Datasource filter separates Excel from Speckle, and an imported item’s Quantity tab shows the current value alongside the original from the sheet — so an edit never hides what was uploaded.
- IFC imports run in the background: keep working, and the project updates itself when processing finishes.
Dashboard and Projects
- A filter bar above the dashboard applies to every number and chart on the page — models, life-cycle stages, indicator, and units.
- Any chart opens full screen with Absolute, Relative, and Stage breakdown views.
- The project breakdown table has a column manager, including material properties and a per-stage emission split.
- The projects page adds a denser sortable list view, and each card shows the revision it is on, its % USED bar against the limit value, and % mapped — so an emission figure is always read in context.
Help that answers first
Feedback in the top-right corner now offers the AI Agent, the user docs, and the community before the form. Issue and Idea still go straight to us, with the page you were on attached automatically.
The user guide, rewritten
Every page has been rewritten against this release, with annotated screenshots that render properly in both light and dark mode, and brand-new pages for the 3D model viewer, History, Revisions, Reports, Clone a project, Extract From PDF, and Tutorials.
This is the release we have been building toward since the platform launched: a project that remembers everything it has been, milestones you can name and return to, reports you can reproduce years later, and an assistant that does the work — and asks before it writes.
Go break it, then tell us what to sharpen next. 