Post Chat Completion | Fern Documentation

Authentication

AuthorizationBearer

Bearer authentication of the form Bearer <token>, where token is your auth token.

Path parameters

domainstringRequired

Request

This endpoint expects an object.

messageslist of objectsRequired

The messages to use for the chat completion

modelenum or nullOptional

The model to use for the chat completion

Allowed values:

max_tokensinteger or nullOptional100-3500Defaults to 3000

The maximum number of tokens to generate. Note: setting a token count lower than 2000 may result in incomplete responses. You can add a custom system prompt to control the verbosity of the response.

system_promptstring or nullOptional

The system prompt to use for the chat completion

rewrite_queryboolean or nullOptionalDefaults to false

Whether to rewrite the query using query decomposition

user_is_authedbooleanOptionalDefaults to false

Whether the requesting user is authenticated. When true, authed chunks are included in results.

allowed_roleslist of strings or nullOptional

Roles the authenticated user has. Used to filter chunks by role-based access control.

sourcestring or nullOptional

The origin of the request (e.g. 'web', 'slack', 'api').

basepathslist of strings or nullOptional

Restrict retrieval to pages under these URL path prefixes (e.g. ['/my-product']). Matches the scoping the Ask AI widget applies when embedded under a product path. Omit to search the whole site.

skip_save_querybooleanOptionalDefaults to false

Set to true for test or eval traffic. The request is answered (and billed) normally but is excluded from usage analytics.

Response

Successful Response

turnslist of objects

The conversation turns in the chat completion

citationslist of strings

List of citation strings

sourceslist of objectsOptional

Pages the answer was grounded in, one entry per cited page

Errors

422

Unprocessable Entity Error