Skip to main content

assistant

Assistant Model​

Defines the structure and configuration of an assistant, including behavioral logic, prompts, and retrieval integrations.

id (integer, ReadOnly)​

Unique identifier of the assistant. Automatically assigned by the system.

generation_config (string) Required​

Name of a valid generation configuration preset to be used in the main model call. Must match a predefined configuration.

generation_config_pretools (string) Required​

Name of a valid generation configuration preset to apply before tool execution. Acts as a preprocessing directive. Must match a predefined configuration.

description (string) Required​

A short summary describing the assistant’s purpose or scope. Minimum 1 character.

system_prompt (string) Required​

Message that helps set the assistant's behavior, task and tone. Must be between 1 and 11400 characters.

temperature (number) Required​

Controls randomness in model responses:

  • 0: Deterministic
  • 2: Very creative
    Must be between 0 and 2.

max_responses (integer, Optional)​

Maximum number of user messages the assistant can process. Use -1 for unlimited.

max_msg_length (integer, Optional)​

Maximum length of each message in characters. Must be greater than 0.

messages_window_to_generator (integer, Optional)​

Number of user interaction windows sent to the generator. Use -1 to send all messages. max_responses and messages_window_to_generator cannot both be -1.

initial_message (string, Optional)​

The message with which the assistant starts the conversation.

end_message (string, Optional)​

The message the assistant sends when it is forced to end the conversation for safety reasons.

add_to_user_message (string, Optional)​

Additional text appended to user messages for specific purposes.

max_consecutive_tool_calls (integer, Optional)​

Maximum consecutive tool calls allowed for the assistant. Must be greater than 0.

created_at (string, date-time, ReadOnly)​

Timestamp when the assistant was created.

updated_at (string, date-time, ReadOnly)​

Timestamp of the last assistant update.

retrieval_faq (integer, ReadOnly, Nullable)​

Reference to the associated FAQ data source used by the assistant.

retrieval_fixed_faq (integer, ReadOnly, Nullable)​

Reference to the fixed FAQ response set.

retrieval_lessons (integer, ReadOnly, Nullable)​

Reference to a lesson-based knowledge source.

matrix_mode_available (boolean, Optional)​

Indicates whether matrix reasoning mode is enabled for the assistant.

not_info_message (string, Optional)​

Message that the assistant sends when it lacks information to answer a query.

strategy_to_optimize_tokens (string, Optional)​

Method to reduce the number of tokens in search messages from previous interactions. Must be at least 1 character if provided.

faq_available (boolean, Optional)​

Whether FAQ functionality is enabled.

fixed_available (boolean, Optional)​

Whether fixed predefined messages are supported.

lessons_available (boolean, Optional)​

Whether the assistant can use lesson-based content.

realtime_available (boolean, Optional)​

Enables real-time interaction features if set to true.

async_available (boolean, Optional)​

Enables asynchronous chat creation and processing if set to true.

info (string, Optional)​

Additional assistant info or metadata string. Minimum 1 character if provided.

colors (object, Optional)​

Object defining UI color configuration for the assistant. May be null.

logo (string, URI, Optional)​

URL pointing to the logo representing the assistant. Maximum 500 characters.

streaming_available (boolean, Optional)​

Indicates if streaming responses are supported for the assistant.

web_search_enabled (boolean, Optional)​

Enables web search for the assistant.

web_search_allowed_domains (array[string] or null, Optional)​

Allowed domains for web search. Values are normalized and validated by the backend.

web_search_location (object or null, Optional)​

Optional location data used to localize web search.

web_search_mode (string, Optional)​

Web search mode.

  • Enum: agentic, deep_research

reasoning_enabled (boolean, Optional)​

Enables reasoning settings for supported generation configurations.

reasoning_effort (string, Optional)​

Reasoning effort level.

  • Enum: low, medium, high, xhigh, none

reasoning_summary (string, Optional)​

Reasoning summary level.

  • Enum: auto, concise, detailed, none

attachment_config (object or null, Optional)​

Per-assistant file handling configuration, keyed by content type. Each content type can define:

  • accepted (boolean): whether files of this type are accepted.
  • send_to_model (boolean): whether accepted files are sent to the model.
  • pdf_mode (string, Optional): for PDFs, either image or text.

tools ([integer], Optional)​

List of tool IDs that are available to the assistant. Values must be unique.

pretools ([integer], Optional)​

List of pre-execution tool IDs triggered before user interaction. Values must be unique.

Example​

{
  "generation_config": "string",
  "generation_config_pretools": "string",
  "description": "string",
  "system_prompt": "string",
  "temperature": 2,
  "max_responses": 2147483647,
  "max_msg_length": 2147483647,
  "messages_window_to_generator": -1,
  "initial_message": "string",
  "end_message": "string",
  "add_to_user_message": "string",
  "max_consecutive_tool_calls": 2147483647,
  "matrix_mode_available": true,
  "not_info_message": "string",
  "strategy_to_optimize_tokens": "string",
  "faq_available": true,
  "fixed_available": true,
  "lessons_available": true,
  "realtime_available": true,
  "async_available": true,
  "info": "string",
  "colors": {},
  "logo": "string",
  "streaming_available": true,
  "web_search_enabled": true,
  "web_search_allowed_domains": [
    "example.com"
  ],
  "web_search_location": {
    "country": "UY"
  },
  "web_search_mode": "agentic",
  "reasoning_enabled": false,
  "reasoning_effort": "none",
  "reasoning_summary": "none",
  "attachment_config": {
    "application/pdf": {
      "accepted": true,
      "send_to_model": true,
      "pdf_mode": "text"
    }
  },
  "tools": [
    0
  ],
  "pretools": [
    0
  ]
}