Skip to content

User Interface

Assistant

Everything about how the Assistant works, in one page: the models it uses for each kind of job, how much it may spend in a day, the company standards it applies when it builds, and two advanced switches. The API keys it runs on stay in Integrations.

Every value is read on the next message, so nothing here needs a restart.

Models

The Assistant uses a model for each kind of job, so a quick question doesn’t pay for the strongest model and a hard build doesn’t settle for a fast one.

SettingWhat it doesDefault
AI modelAnswers and acts: the model behind most of the conversation.Claude Opus 4.8
Router modelReads each message first and decides what kind of request it is — a question, a change, a how-to, a picture — so the Assistant brings out only the tools that request needs. A short job, so a small, fast model is enough.The latest Claude Haiku
Vision modelLooks at the images you attach.Same as the AI model
Reasoning modelPlans multi-step builds — “a ring with a halo and pavé on the shank” — where thinking ahead pays off.Same as the AI model
Image modelMakes a picture from another one: a render from your viewport, a try-on, a variation of a photo you attached.Nano Banana Pro
Text-to-image modelMakes a picture from a description alone.Nano Banana 2

The first four run on your OpenRouter key, the two image models on your fal.ai key. Which model to pick, and why: Choosing models.

Costs

  • Daily limit (US$) — the most the Assistant’s AI calls may cost in a day, as OpenRouter bills them. Once today’s calls reach it, the Assistant stops and says so in the chat; the call that crosses the limit still finishes. 0, the default, means no limit.
  • Spent today — what the Assistant has spent so far today, across conversations and Rhino sessions.

The Assistant panel shows what the current conversation has cost next to the Auto button. Images generated with fal.ai are billed by fal.ai and aren’t counted here. More in Costs and Spending.

Company standards

The rules your company always follows, in plain language — minimum sizes per metal, default metals, production steps — which the Assistant applies whenever it builds or changes a design. Write them in the editor, or use Open in text editor and Open folder to work on the file directly. How to write them, and what the Assistant does with them: Company standards.

Advanced

  • Show answers as they’re written — on by default. The answer appears word by word in the panel while the model writes it, then settles into its final formatting. Turned off, it appears whole when it is done.
  • Keep each file’s conversation on this computer — on by default. Opening a saved file again brings back its Assistant chat (the text only). The conversation is stored in your user folder, never inside the .3dm, and Clear History deletes it. See The Assistant panel.
  • Route requests by capability (recommended) — on by default. Each message is read first by the router model, and the Assistant offers only the tools that kind of request needs: faster, cheaper and more focused. Turned off, every message gets every tool.
  • Write a debug trace — writes one line per turn (models used, tools run, timings) to the Logs folder, for troubleshooting with support. Off by default.