Interface●Since v1.2

ModelRequest

Configuration for modifying a model call at runtime. All fields are optional and only provided fields will override defaults.

interface ModelRequest

Properties

property

messages: BaseMessage<MessageStructure<MessageToolSet>, MessageType>[]

property

model: AgentLanguageModelLike

property

modelSettings: Record<string, unknown>

Additional settings to bind to the model when preparing it for invocation. These settings are applied via bindTools() and can include parameters like headers, container, etc. The model is re-bound on each request, so these settings can vary per invocation.

modelSettings: {
  headers: { "anthropic-beta": "code-execution-2025-08-25" },
  container: "container_abc123"
}

property

responseFormat: TResponseFormat

The tool response format.

If "content" then the output of the tool is interpreted as the contents of a ToolMessage. If "content_and_artifact" then the output is expected to be a two-tuple corresponding to the (content, artifact) of a ToolMessage.

property

runtime: Runtime<TContext>

The runtime context containing metadata, signal, writer, interrupt, etc.

property

state: TState & AgentBuiltInState

The current agent state (includes both middleware state and built-in state).

property

systemMessage: SystemMessage

The system message for this step. If no systemPrompt was provided, when createAgent was initialized, an empty system message will be used.

wrapModelCall: async (request, handler) => {
  return handler({
    ...request,
    systemMessage: request.systemMessage.concat("something")
  });
}

wrapModelCall: async (request, handler) => {
  return handler({
    ...request,
    systemMessage: request.systemMessage.concat(
      new SystemMessage({
        content: [{
          type: "text",
          text: "something",
          cache_control: { type: "ephemeral", ttl: "1m" }
        }],
      })
    ),
  });
}

property

toolChoice: "required" | "auto" | "none" | __type

Tool choice configuration (model-specific format). Can be one of:

"auto": means the model can pick between generating a message or calling one or more tools.
"none": means the model will not call any tool and instead generates a message.
"required": means the model must call one or more tools.
{ type: "function", function: { name: string } }: The model will use the specified function.

property

tools: ClientTool | ServerTool[]

deprecatedproperty

systemPrompt: string

The system message string for this step.

View source on GitHub

Properties

property

messages: BaseMessage<MessageStructure<MessageToolSet>, MessageType>[]

property

model: AgentLanguageModelLike

property

modelSettings: Record<string, unknown>

modelSettings: {
  headers: { "anthropic-beta": "code-execution-2025-08-25" },
  container: "container_abc123"
}

property

responseFormat: TResponseFormat

The tool response format.

property

runtime: Runtime<TContext>

The runtime context containing metadata, signal, writer, interrupt, etc.

property

state: TState & AgentBuiltInState

The current agent state (includes both middleware state and built-in state).

property

systemMessage: SystemMessage

The system message for this step. If no systemPrompt was provided, when createAgent was initialized, an empty system message will be used.

wrapModelCall: async (request, handler) => {
  return handler({
    ...request,
    systemMessage: request.systemMessage.concat("something")
  });
}

wrapModelCall: async (request, handler) => {
  return handler({
    ...request,
    systemMessage: request.systemMessage.concat(
      new SystemMessage({
        content: [{
          type: "text",
          text: "something",
          cache_control: { type: "ephemeral", ttl: "1m" }
        }],
      })
    ),
  });
}

property

toolChoice: "required" | "auto" | "none" | __type

Tool choice configuration (model-specific format). Can be one of:

"auto": means the model can pick between generating a message or calling one or more tools.
"none": means the model will not call any tool and instead generates a message.
"required": means the model must call one or more tools.
{ type: "function", function: { name: string } }: The model will use the specified function.

property

tools: ClientTool | ServerTool[]

deprecatedproperty

systemPrompt: string

The system message string for this step.

ModelRequest

Properties

LangChain Assistant

Menu

ModelRequest

Properties