AI chat

Conversations, steering text, and LLM endpoint administration on V3Client. Chat in the web app is AI Chat.

LLM administration (ai_config_*) is an operator surface: endpoints, models, failover order, and the identical-tool-call guard. ai_config_set_llm_auto_switch stores the failover flag. The client documents that flag as stored only; the platform does not enforce it yet.

Conversations

create_ai_conversation()

• Parameters:

  • create_conversation_request (CreateConversationRequest) – (required). Optional title (max 256 characters).
  • http_request_timeout_secs (int, optional)

• Return Type: AiConversationDto

list_ai_conversations()

• Parameters:

  • q (str, optional) – Search string.
  • include_archived (bool, optional)
  • cursor (str, optional)
  • size (int, optional)
  • include_total (bool, optional)
  • http_request_timeout_secs (int, optional)

• Return Type: CursorPageAiConversationDto

update_ai_conversation()

• Parameters:

  • conversation_id (str) – (required)
  • update_conversation_request (UpdateConversationRequest) – (required). Optional title (max 256 characters) and archived.
  • http_request_timeout_secs (int, optional)

• Return Type: AiConversationDto

list_ai_conversation_messages()

• Parameters:

  • conversation_id (str) – (required)
  • http_request_timeout_secs (int, optional)

• Return Type: list of AiMessageDto

append_ai_conversation_message()

• Parameters:

  • conversation_id (str) – (required)
  • append_message_request (AppendMessageRequest) – (required). role is user, assistant, or system. content is required.
  • http_request_timeout_secs (int, optional)

• Return Type: AiMessageDto

ai_chat()

Sends a chat turn to the configured LLM.

• Parameters:

  • chat_request (ChatRequest) – (required). messages is a non-empty list of {role, content}. system_prompt is optional.
  • http_request_timeout_secs (int, optional)

• Return Type: ChatResponse

ai_list_user_models()

Lists models the caller can select in chat.

• Parameters:

  • http_request_timeout_secs (int, optional)

• Return Type: UserSelectableModelsResponse

Steering and tenant AI config

get_ai_user_steering()

Returns the caller's steering text.

• Return Type: UserSteeringDto

update_ai_user_steering()

Sets the caller's steering text. steering_prompt null clears it (max 2000 characters).

• Parameters:

  • update_user_steering_request (UpdateUserSteeringRequest) – (required)
  • http_request_timeout_secs (int, optional)

• Return Type: UserSteeringDto

get_ai_config()

Returns the tenant AI configuration.

• Return Type: AiConfigDto

update_ai_config()

Patches the tenant AI configuration. A field left null is unchanged.

• Parameters:

  • ai_config_update_request (AiConfigUpdateRequest) – (required). Optional: steering_prompt, system_prompt_delivery (system_role or user_turn), tool_mode (native or prompt), user_steering_enabled.
  • http_request_timeout_secs (int, optional)

• Return Type: AiConfigDto

LLM endpoints and models

ai_config_list_available_hosts()

Lists host slots an endpoint can use.

• Return Type: AvailableHostsResponse

ai_config_list_llm_endpoints()

Lists endpoints and their model configurations.

• Return Type: LlmConfigResponse

ai_config_create_llm_endpoint()

• Parameters:

  • upsert_endpoint_request (UpsertEndpointRequest) – (required). name, slot, and enabled.
  • http_request_timeout_secs (int, optional)

• Return Type: LlmEndpointDto

ai_config_update_llm_endpoint()

• Parameters:

  • endpoint_id (str) – (required)
  • upsert_endpoint_request (UpsertEndpointRequest) – (required). name, slot, and enabled.
  • http_request_timeout_secs (int, optional)

• Return Type: LlmEndpointDto

ai_config_delete_llm_endpoint()

Deletes an endpoint and its model configurations.

• Parameters:

  • endpoint_id (str) – (required)
  • expected_revision (int) – (required)
  • http_request_timeout_secs (int, optional)

• Return Type: object

ai_config_create_llm_model()

Adds a model to an endpoint.

• Parameters:

  • endpoint_id (str) – (required)
  • upsert_model_request (UpsertModelRequest) – (required). Required: name, model_id, disable_system_prompt, disable_native_tools, enabled, user_selectable. Optional: path, max_iterations, carry_tool_context, return_thinking_blocks, carry_tool_max_calls, carry_tool_max_chars, carry_args_max_chars, extra_params.
  • http_request_timeout_secs (int, optional)

• Return Type: LlmModelDto

ai_config_update_llm_model()

• Parameters:

  • model_pk (str) – (required)
  • upsert_model_request (UpsertModelRequest) – (required). Same fields as create.
  • http_request_timeout_secs (int, optional)

• Return Type: LlmModelDto

ai_config_delete_llm_model()

• Parameters:

  • model_pk (str) – (required)
  • expected_revision (int) – (required)
  • http_request_timeout_secs (int, optional)

• Return Type: object

ai_config_reorder_llm_model()

Sets a model's place in the failover priority list.

• Parameters:

  • model_pk (str) – (required)
  • reorder_model_request (ReorderModelRequest) – (required). position and expected_revision.
  • http_request_timeout_secs (int, optional)

• Return Type: LlmModelDto

ai_config_set_llm_auto_switch()

Stores whether availability-based model failover is enabled. The value is stored only; the platform does not enforce it yet.

• Parameters:

  • set_auto_switch_request (SetAutoSwitchRequest) – (required). enabled.
  • http_request_timeout_secs (int, optional)

• Return Type: SetAutoSwitchRequest

ai_config_set_llm_max_identical_tool_calls()

Sets the cluster-wide cap on identical tool calls.

• Parameters:

  • set_max_identical_tool_calls_request (SetMaxIdenticalToolCallsRequest) – (required). max_identical_tool_calls.
  • http_request_timeout_secs (int, optional)

• Return Type: SetMaxIdenticalToolCallsRequest