DesktopModel review

It tells you what is wrong before you ask.

Connect Scope3D Desktop to an open Revit document and the Needs panel fills in before you have typed anything: thirteen kinds of gap, Revit’s own warnings diagnosed rather than counted, duplicated elements, and the storeys with a piece of them open to the sky.

This is counting, not judging. Every line is read straight out of the document that is open, so the whole panel costs nothing to produce — and each finding prints the count it came from underneath itself.

NEEDS

What this model still needs

Read the moment it connected. No prompt written, no model called, nothing spent.

The review

Two questions Revit does not answer

Sixty-two warnings is one fault, not sixty-two

Revit’s warning list says whether what you built is
well-formed. It does not say what to do about it. Each warning
kind is matched against a rule written in Revit’s own wording
and comes back as a cause and a correction, so a pile of the
same warning reads as one problem repeated. Five of a kind is
where it starts calling it a pile.

  • Two stairs a fifth of a foot apart raise no Revit warning — that one wants exact coincidence
  • This does: same size and position to within an inch, and how many to delete of the total
How the checking works
62×

“That felt slow” becomes “step 14 took ninety seconds”

Every run, phase, tool event, notice, usage record and setup
step is appended to %LOCALAPPDATA%\Scope3D\run-log.jsonl as
it happens, one JSON object per line. A drawer reads the last
300 back, newest first: kind, verb, status, ms, summary, model
and the error verbatim. It is the file to attach to a bug report.

  • Terminal facts only — streaming deltas are never written, so the file stays readable
  • Every write is fire-and-forget, so a logging failure can never be why a run fails
Reporting a problem
RUN LOG

Thirteen kinds of gap, and what each cannot see.

Each kind knows three things: how it is counted in Revit, when a building of this sort should have it, and which stage builds it. A kind it cannot count produces nothing at all — not a card, not a line in the report — because a button that cannot be honoured is how you stop reading the panel.

The catalogue names fourteen. Foundations is the fourteenth and is deliberately never offered: footings sit below the lowest level and appear on no plan sheet, so it is counted and never raised.

Four more findings come from outside the list: a stair that ends in mid-air where no slab was built above it; a storey with a piece open to the sky, reported only past both 40 sf and 5% of the storey; two of the same element in the same place; and solid clashes, when you ask for the scan.

Floor slabs

“Nothing is standing on anything.” One slab per storey, counted per level rather than as a total, so one floor across a two-storey house is still a gap. Not checked: whether the slab reaches the whole storey — that is the separate open-to-the-sky finding.

Stairs

“No way between the storeys.” Raised only where the document holds more than one level. Not checked: a single storey is silent, and it counts levels — it does not know a mezzanine from a floor.

Stairwell opening

“The stair runs into the slab above it.” Measured as a hole, not as a shaft element, because an opening cut into the slab’s own sketch is the better modelling and is the same thing to anyone using the building. Not checked: how many a building needs. Once one hole exists it stops asserting a second — that answer needs to know which stair passes through which slab, and the census carries totals, not geometry.

Posts and columns

“Posts are drawn and not modelled” only when a sheet actually drew some; otherwise “No posts or columns”. A commercial floor plate is expected to sit on a column grid. Not checked: a house with nothing drawn is never nagged, because a house may legitimately have none.

Beams and headers

Offered once posts exist, and only ever presented as missing when a sheet draws them. Not checked: whether a span needs a beam. That is a structural question about unsupported openings and nothing here computes it — one post at a deck corner is not evidence of a header.

Roof

“The building is open to the sky.” Not checked: the shape. A floor plan does not state a pitch, so the stage behind the button asks you for one rather than picking it.

Doors

“Every room is sealed”, or “Fewer doors than the sheet draws” with the count beside it — nine of eleven is the opening detector’s known shortfall, and reporting it as done would hide exactly that. Not checked: with no sheet loaded the expectation is one way into each room and nothing more precise.

Windows

Expected at the number measured off the plan. Not checked: with none measured it is offered rather than demanded — a wall with no window is not a fault the app can assert.

Ceilings

“No ceilings”, once rooms are placed, and always as worth doing rather than missing. Not checked: height. A ceiling nobody has chosen a height for is not a gap.

Railings

“An open edge with no guard”, once a stair exists — a code item rather than decoration. Not checked: it counts guards. It cannot see a handrail height.

Curtain wall

“No glazed frontage”, on a commercial building, where the ground floor is mostly glazing. Not checked: on a house it says nothing at all rather than offering it. Offering a bungalow a storefront is the panel telling you it does not know what it is looking at.

Plumbing fixtures

Expected at the number drawn on the plan; otherwise “Bathrooms with nothing in them”, from rooms named bath, WC, toilet, powder or ensuite. Not checked: a wet room named something else is not found, and nothing here reads a fixture off the model.

Casework

“The kitchen reads as an empty room”, from rooms named kitchen, pantry or utility. Not checked: the same blind spot — it reads room names, not what is in them.

It publishes what it did not look at.

Every review ends with a “Not checked” list. It is load-bearing rather than decoration: a review that goes quiet about its own blind spots gets trusted for more than it can carry. Its three entries are the first three below, and two of them carry the button that would go and check. The last two are the standing limits behind them.

Revit’s own warnings

If the warning read did not come back, that is a line in Not checked. It is never reported as a zero.

Solid clashes

Not scanned unless you ask — it is the slow read. “Scan for clashes” is a button on the panel, and a scan that hits its pair limit reports a floor rather than a total.

Nothing compared against a drawing

Drop a sheet and “Check it against the sheet” appears. It is the one action here that calls the model, and the panel says “this one costs” before it runs, not after.

Anything outside the fourteen kinds

It cannot see a handrail height, or a door that opens into a wall. Those are what the review wizards are for.

Wrong-building guard

Each conversation records the Revit document it was working on, keyed by file path rather than by title, because two files can share a title. Reopen it against a different model and work stops: “This chat was working on X. Revit currently has Y open.” The check comes from Revit rather than from what the chat remembers — a transcript reads identically whichever project is loaded, so nothing else in the app can catch this.

Every card shows its working.

The headline is the problem in a person’s words — “Nothing is standing on anything”, not “0 floors”. The count stays underneath it. The sentence replaces the headline, not the evidence.

Findings sort missing first, then wrong, then worth doing, then what is fine — and what is fine is said out loud, because a review that only ever lists problems stops being believed.

A severity, as a colour

Missing is expected and absent. Wrong is something built that is not right. Worth doing is available, not urgent. Sound is nothing to do, stated rather than implied.

A line saying how it was counted

“none on L1, L2 · every storey needs a slab, and this building has 2 levels”. Small, and under every finding, so you check the claim rather than take it.

“seen 3×”, from the second sighting

How many separate reviews of this model raised the same thing. “No roof for four runs” says something “no roof” does not. One sighting is not shown, because one is noise.

What will exist afterwards, then the button

“Build: a slab on L1 and L2, from the envelope. Cut: a shaft through the slab above, where the stair lands.” Pressing it routes into the stage that fixes it and sends the finding’s own words — what is wrong, how it was counted, and what was promised on screen.

Not this one, or Later

Set one aside for this model and it stays aside, counted at the foot of the panel and restorable in one click. Later hides it for twelve hours. And where a run has already gone and looked and found nothing, the card stays — the thing really is absent — but the button becomes “Look again”, so you cannot buy the same investigation twice.

See it work on your model

Bring a floor plan and an open Revit session. We will walk you through a trace end to end.

Book a demo Models and cost