Every answer uses the AI requests of your Opensolr plan. The Limits tab decides how much of that one visitor, or one conversation, can spend.
The longest question a visitor can send. A longer one is not sent: Your message is too long: 1,250 characters, at most 1,000.
The longest text the /translate command takes, counted on the text alone, without the command and the language codes.
How many questions one visitor may ask in a time window, counted per IP address, commands included. When they are used up: You have asked many questions in a short time. Please try again a little later. The visitor can ask again as the oldest questions leave the window.
How many questions one conversation may have. Commands do not count. When they are used up: This conversation has reached 20 questions. Start a new chat with the New chat button. A new chat starts the count again, while the visitor’s own limit above keeps counting.
A visitor gets one answer at a time, across all their tabs: a second question sent while the first is still being answered gets Please wait for the answer to your last question. This is not a setting; it is always on.
- By IP address. The address itself is never stored: the chat keeps a keyed hash of it, and deletes the counts once they are older than the time window.
- Behind a proxy, a load balancer or a CDN, every visitor arrives with the proxy’s address unless you set
trusted_proxiesin the front controller. Without it, all your visitors share one limit and reach it together. See Installing. - Visitors behind one shared address, such as an office or a school network, share the limit of that address.
How many AI requests your plan has, and how many are used this month, is explained on API Quota.