Routing & Budgets Explained

Two questions every AI request answers: which model does the work, and may this AI member spend any more today? Settings → Super Intelligence → Brains & Tools → Model Choices answers the first, and Settings → Super Intelligence → Safety & Oversight → Spending the second. Model Choices needs Manage AI Settings. Spending opens for anyone who may read SI usage (View AI Usage) or change the budgets (Manage AI Workforce); only Manage AI Workforce can change them.

How a model is chosen

Kamo never names an AI vendor in advance. Each request asks for a kind of work, such as text generation, tool calling, reading images, embeddings, speech to text or text to speech, and Kamo picks a model that can do it, in this order:

1. the models you list for that kind of work under Model Preferences, top first;
2. then your own providers before the ones Provided by Kamo (for speech to text and text to speech, Kamo's own speech service comes first);
3. then providers that are currently healthy, the quality the request asked for, and your provider order in the AI Provider Registry.

Models you don't list still serve as fallbacks. If the first choice is unavailable, Kamo moves on to the next one rather than failing.

An AI member's charter adds to this: its Brain section can ask for a lighter or heavier model for a kind of work, and can require that health data stays with the AI services Kamo runs itself. See The AI Charter Explained.

Set your preferred models

1. In Model Preferences, choose a Kind of work.
2. Click Add a model and choose one. Each model shows whether it is Your provider or Provided by Kamo, and whether it is Verified, Not verified or Unavailable.
3. Put them in order with Move up and Move down, and take one out with Remove.
4. Click Save order.

Use default order removes your list for that kind of work, and Kamo goes back to choosing by itself.

Verified models only, for AI members

AI members only use capabilities Kamo has verified with a live test. A model in your list that is Not verified is skipped for them, and used only by people and by Kamo's own background tasks. Verify it in the AI Provider Registry to make it available to AI members too; see AI Providers Explained.

Embeddings use exactly one model

Search indexes, such as the one behind your knowledge base, are built with one embeddings model. Kamo pins the first one it uses. Changing it means the existing indexes must be rebuilt before search works well again, so change it only on purpose.

AI member budgets

Every AI member can have a daily and a monthly spending limit in US dollars. Its charter's Limits section sets them for a role, department, job title or one AI member. AI member budgets on this tab sets the default for every AI member whose charter sets none:

1. Enter a Daily limit (USD), a Monthly limit (USD), or both. Leave one empty for no default limit.
2. Click Save limits.

Budgets reset at 00:00 UTC: the daily one every day, the monthly one on the first of the month.

What happens at the limit

• At 80% of a budget, its supervisor is told.
• At 100%, the AI member goes On hold (Daily budget reached or Monthly budget reached), and its supervisor is told again. It does no AI work until the budget resets, then carries on by itself. Work that arrived meanwhile waits for it.

Its workbench shows what it has spent Today and This month against each limit, and when each resets. SI Workforce → Results → Spend shows the cost per SI for any period; see The SI Workforce Page.

These budgets are for AI members. How much the people in your organization may use the AI Assistant is set per role with an access policy; see AI Security Models Explained.