Review inbox

The review inbox is the one place suggestions land — whether they come from the AI reviewer or from a human collaborator. Both flow through the same queue and the same accept/discard controls.

One queue, two sources

• AI suggestions arrive from the AI reviewer: a manufacturability flag, and often a proposed parameter change to fix it.
• Human suggestions arrive from a collaborator who forked your shared model, made an edit, and submitted it back.

They sit side by side, each labeled by source. Nothing distinguishes them procedurally: an AI suggestion is reviewed the same way a teammate’s is.

What a suggestion shows

• The flag or intent — what the change addresses, framed against the actual model.
• The concrete change — for a parametric part, a typed parameter edit shown as old → new.
• A visual diff — for a human suggestion, the differences are highlighted: added, modified, removed, moved, renamed, or resized.

Click a flagged issue and the camera frames it in the viewport, so you see the problem before you decide.

Accept or discard

Every suggestion is resolved by you, with one click:

• Accept applies the change. For a parametric part, it lands as a single, reversible history entry — one line you can read and undo.
• Discard dismisses it without touching your model.

Nothing is applied automatically. The AI proposes; a human accepts. This is the discipline that makes an automated reviewer safe to keep on: it can only ever suggest, never edit.

Why one pipeline matters

Because AI and human review share the same rails, a suggestion behaves identically no matter who made it. There is no separate, less-careful path for machine edits, and adding review from people is a natural extension of the same mechanic rather than a second system to trust.