# Limits and quotas

## Per message

| Limit                  | Value           |
| ---------------------- | --------------- |
| Question length        | 2000 characters |
| Turns per conversation | 20              |
| Answer length          | 800 tokens      |

Longer questions fail with `bad_request`. A conversation past its turn limit fails with `conversation_too_long`.

## Rate limits

Public chat requests are limited per visitor and per project over a one-minute window. Higher plans allow more questions per minute. When a limit is reached the API answers `429 rate_limited` with a `Retry-After` header, in seconds.

## Usage

Projects have a conversation allowance. A conversation counts once, the first time the model is called for it; follow-up questions in the same conversation are free, and questions declined for lack of evidence or for safety do not count. When the allowance is used up, chat requests fail with `402 quota_exceeded` and the widget tells the visitor the assistant is unavailable.

How many conversations your plan includes is shown in the dashboard.

## Indexing

| Limit                        | Value                   |
| ---------------------------- | ----------------------- |
| Pages per source             | 1 to 2000 (default 500) |
| Include and exclude patterns | up to 20 each           |
| Docs sources per project     | set by your plan        |
| Projects per account         | set by your plan        |

Docora fetches only public pages on the site you add, follows redirects only within that site, and refuses private and internal network addresses.