Sohay does not resell model access and takes no cut. Your only running cost is what your AI provider charges, billed directly to you.
What you are charged for
Providers charge per token — roughly, per chunk of text in and out. A chat reply costs you:
- The visitor’s question.
- The recent conversation history, so the model can follow a thread.
- Any Knowledge Base articles retrieved to answer it.
- The reply itself.
That third item is the one people do not expect. A grounded answer carries the source articles along with it, so a support reply costs more per message than a bare chatbot’s would. It is also exactly what you are paying for — it is the difference between an answer from your documentation and an answer from the model’s imagination.
Indexing your Knowledge Base has its own small cost, paid once per article when it syncs rather than per conversation.
What it actually comes to
On a cheap model, a support reply is a fraction of a cent. A small support site typically stays in single-digit dollars per month.
The things that move that number, in order:
- How many conversations you get. This is the honest driver, and it is the one you want going up.
- Which model you chose. Flagship tiers cost a large multiple of the mini ones for the same work. See Picking a model.
- How long your articles are. Retrieval sends article text to the model, so very long articles cost more per answer than focused ones.
Keeping it predictable
Two things, and both are worth doing before you go live rather than after.
Set a daily token cap under Sohay → AI Settings → Spend protection. Once the day’s usage passes it, Sohay simply stops calling your provider until the ceiling resets at UTC midnight. A runaway loop or a hostile bot costs you the cap, not the weekend.
Watch Diagnostics. Sohay → Diagnostics shows today’s token usage. A number that jumps without a matching jump in conversations is the signal to look at — that is what a loop or an abusive client looks like before it becomes a bill.
Most providers also let you set a hard spending limit on the account itself. Doing both is not redundant: the provider’s limit protects the account, and Sohay’s protects the site.
Where to go next
Daily token caps and rate limits covers the ceilings in detail, including the per-visitor ones.
