Cabi.Chat Ignite — Models and AI usage
Every oracle reply is an LLM run, and you choose the model. The model is set per agenda, so one oracle can answer on a fast, cheap model while a campaign's first contact runs on a stronger one. Cabi.Chat Ignite — Agendas
Choosing
- Model — a model id, set on the agenda. Leave it empty and the platform default runs.
- Reasoning — off, low (default), medium or high: how hard the model thinks before it answers. Higher costs more tokens and time.
- When a conversation belongs to a campaign, the campaign's swappable agenda brings its own model and reasoning, and they win over the firmware agenda's; fields it leaves empty fall back to the firmware agenda.
Photos. When a message carries an image and the agenda sets no model, the run switches to a vision model by itself.
How much a run does
- A turn Cabi.Chat Ignite has a fixed step limit.
- The model sees the recent part of the conversation, up to a fixed budget, and pages back to older messages only when it needs them. Long chats do not grow the cost of every reply.
Paying for it. AI usage pricing is in the price list. link