Omni

What Omni is set to in this release: whether it is on, which models it asks and in what order, what it may spend, and the questions it answers. How it works is on the Omni page.

These are the shipped defaults. What you can change lives in a config file you own once TESSERACT is installed, so your copy may differ, and should, once you tune it.

The switch

Omni ships off. It is the omni role’s mode in config/roles.yaml; /omni on and /omni off set it from the app or any channel.

Who it asks

Order	Model	Asked as
1	api.jev.jev_1_13_0	its own question format
2	api.openai.gpt6_luna	the omni card
3	api.google.gemini_38_flash	the omni card

Each is tried in order. One that cannot answer at all (no key, no reply in time, its budget spent) passes the question to the next.

What it may spend

Up to $0.50 a day, the omni line of config/budgets.yaml, whichever model answers.

The questions

Question	Shape	Answers	Floor	Leaves the machine
memory_type	choice	feedback, user, project, reference	0.6	yes

A floor is how sure an answer must be before the app acts on it. They are in config/decisions.yaml.

Floors for other places that hold a confidence: observer_suggestion 0.0.