Struct ChatOptions 

pub struct ChatOptions {

    pub model: Option<String>,
    pub allow_multiple_tool_calls: Option<bool>,
    pub conversation_id: Option<String>,
    pub frequency_penalty: Option<f32>,
    pub instructions: Option<String>,
    pub logit_bias: Option<HashMap<String, f32>>,
    pub max_tokens: Option<u32>,
    pub metadata: Option<HashMap<String, String>>,
    pub presence_penalty: Option<f32>,
    pub response_format: Option<ResponseFormat>,
    pub seed: Option<i64>,
    pub stop: Option<Vec<String>>,
    pub store: Option<bool>,
    pub temperature: Option<f32>,
    pub tool_choice: Option<ToolMode>,
    pub tools: Vec<ToolDefinition>,
    pub top_p: Option<f32>,
    pub user: Option<String>,
    pub additional_properties: HashMap<String, Value>,
    pub session: Option<AgentSession>,
}

Common per-request settings for a chat/AI service.

The ChatOptions::merge method implements the Python & operator: the right-hand side wins per scalar field, list/map fields are combined, and instructions are newline-concatenated.

Fields

model: Option<String>

allow_multiple_tool_calls: Option<bool>

conversation_id: Option<String>

frequency_penalty: Option<f32>

instructions: Option<String>

logit_bias: Option<HashMap<String, f32>>

max_tokens: Option<u32>

metadata: Option<HashMap<String, String>>

presence_penalty: Option<f32>

response_format: Option<ResponseFormat>

The requested structured-output format, if any.

seed: Option<i64>

stop: Option<Vec<String>>

store: Option<bool>

temperature: Option<f32>

tool_choice: Option<ToolMode>

tools: Vec<ToolDefinition>

top_p: Option<f32>

user: Option<String>

additional_properties: HashMap<String, Value>

session: Option<AgentSession>

The AgentSession of the agent run this request belongs to, if any.

This is a framework-internal side channel, not a wire field: the agent sets it in prepare_request, and the function-invocation loop pops it (before the inner provider client ever sees the options) and hands it to invoked tools via FunctionInvocationContext::session — which is what lets Agent::as_tool with propagate_session forward the parent’s session to a sub-agent. Mirrors upstream, where the agent passes session= through the chat-client kwargs and the function-invocation layer pops it (_middleware.py:1188). Provider converters must ignore it.

Implementations

impl ChatOptions

pub fn new() -> ChatOptions

pub fn with_instructions(self, instructions: impl Into<String>) -> ChatOptions

Builder: set the instructions (system prompt).

pub fn with_model(self, model: impl Into<String>) -> ChatOptions

Builder: set the model id.

pub fn with_temperature(self, temperature: f32) -> ChatOptions

Builder: set the temperature.

pub fn with_max_tokens(self, max_tokens: u32) -> ChatOptions

Builder: set the max output tokens.

pub fn with_tool(self, tool: ToolDefinition) -> ChatOptions

Builder: add a tool.

pub fn with_tool_choice(self, mode: ToolMode) -> ChatOptions

Builder: set the tool choice.

pub fn with_response_format(self, format: ResponseFormat) -> ChatOptions

Builder: set the structured-output response format.

pub fn merge(self, other: ChatOptions) -> ChatOptions

Merge other into self, with other taking precedence per the Python ChatOptions.__and__ semantics.

Trait Implementations

impl Clone for ChatOptions

fn clone(&self) -> ChatOptions

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ChatOptions

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl Default for ChatOptions

fn default() -> ChatOptions

Returns the “default value” for a type.