FAQ / 07
CLEAR RULES.
CLEAR
BOUNDARIES.
TORI is designed to be understandable: what it knows, how it reaches an intent, which fact it retrieved, and where it needs a human to review.
Is TORI a large AI model?
The first Core is a deterministic JavaScript language-and-knowledge engine. It uses phrase families, normalization, entity matching, context scoring and explicit business facts instead of a heavyweight neural model.
Can I add multiple products?
Yes. Products, services, custom facts, keywords and learned phrases are separate knowledge entries. The current local environment has no fixed product-count slot.
Will TORI invent a business name if I forget to add one?
No. An unconfigured business name stays unknown. The same rule applies to price, stock, delivery, policies, contact details and links.
Will bad grammar make a message fail?
Not automatically. TORI normalizes common spelling errors, abbreviations and chat forms. Language quality and meaning confidence are treated as different things.
Can one message contain more than one question?
Yes. The engine keeps a stack of intent candidates, can combine compatible high-signal intents, and retrieves the facts needed for each part.
What makes a result amber?
Amber means the interpretation is plausible but needs review — for example, the intent is clear but a required fact is missing, or multiple products are similarly likely.
Does the website need a server?
No. The website and Core Lab are static. Knowledge Studio stores data in browser localStorage in this version. Account authentication is now handled through Supabase; Knowledge Studio data remains browser-local in this prototype.
What is the desktop app supposed to do?
Desktop is the first full environment for the shared Core, business knowledge, message testing and monitoring preview. Mobile can later reuse the same Core with an Android-specific input layer.