Floodlight

How Hemma is built — in the open.

All releases

2026.05.1

2026.05.1

Date: May 1, 2026

What's new

  • Mark up your plans — open any plan and annotate directly on the drawing. Sticky notes, shapes, callouts, tags — whatever helps you flag a question or capture a decision. Invite your architect with a private link and they get the same canvas: they can drop their own annotations, reply to yours, and resolve threads as you work them out. No accounts, no passwords — just a link that lets them in. A new sharing center keeps track of who has access to what, when they last opened it, and lets you revoke access in a click. Works the same on laptop and phone.

  • Trusted senders — Hemma now keeps a per-project ledger of who is allowed to email documents into your inbox. New senders land in a review queue until you say yes, so strangers can't push paperwork into your project on their own.

  • Try Hemma without signing up — six guided demos on the marketing site (find anything, know your budget, compare quotes, mark up your plans, track your progress, stay informed) walk visitors through what Hemma actually feels like, on real-looking data.

  • English and French sites — the marketing pages, help articles, and changelog are now available in English and Belgian French alongside Dutch.

What's better

  • VAT transparency — invoices and quotes now consistently surface VAT-inclusive amounts in budget linking, with a per-line-item VAT breakdown. The "incl. VAT" convention is spelled out everywhere it matters.

  • Credit notes line up correctly — credits and refunds now match against the right budget lines by magnitude, so a -€500 correction reconciles cleanly against a +€500 charge instead of failing silently.

  • More reliable document processing — the email-to-asset pipeline was rewritten as a durable state machine. Failed messages retry, duplicates are detected, stuck messages are reconciled, and when something does go wrong you see it explicitly on the message instead of guessing why nothing happened.

  • Sharper extraction reviews — the AI consensus check now normalises fields before comparing them, which means fewer documents end up flagged for review when the two providers actually agree.

  • Profile and Settings, separated — the avatar menu now splits cleanly into Profile (you) and Settings (your project), and the mobile project tile shows real milestone progress instead of placeholder data.

What's fixed

The short /sign-in URL works again. Negative amounts (credit notes, refunds) no longer break budget linking. Assets that failed processing now show a clear failure state with a retry option instead of sitting silent. Reply authors on plan annotations show the actual name, not the email address. Admin impersonation exposes the right project context, so support can debug as the user sees the app.