Struct AgentConfig 

pub struct AgentConfig {

    pub system_prompt: Option<String>,
    pub system_prompt_file: Option<PathBuf>,
    pub tool_instructions: Option<String>,
    pub model: String,
    pub max_tokens_per_turn: u32,
    pub temperature: Option<f32>,
    pub budget_warning_threshold: f32,
    pub max_turns: Option<u32>,
    pub provider_params: Option<Value>,
    pub provider_tool_defaults: Option<Value>,
    pub output_schema: Option<OutputSchema>,
    pub structured_output_retries: u32,
    pub extraction_prompt: Option<String>,
}

Agent behavior configuration

Fields

system_prompt: Option<String>

System prompt to prepend

system_prompt_file: Option<PathBuf>

Path to system prompt file (alternative to inline system_prompt)

tool_instructions: Option<String>

Optional tool usage instructions appended to system prompt

model: String

Model identifier (provider-specific)

max_tokens_per_turn: u32

Maximum tokens to generate per turn

temperature: Option<f32>

Temperature for sampling

budget_warning_threshold: f32

Warning threshold for budget (0.0-1.0)

max_turns: Option<u32>

Maximum turns before forced stop

provider_params: Option<Value>

Provider-specific parameters (e.g., thinking config, reasoning effort)

This is a generic JSON bag that providers can extract provider-specific options from. Each provider implementation is responsible for reading and applying relevant parameters.

provider_tool_defaults: Option<Value>

Provider-native tool defaults resolved at factory build time.

Intentionally non-persisted (#[serde(skip)]): re-derived on every build (including resume) from Config.provider_tools + ModelProfile.supports_web_search. This means config changes (e.g., disabling web search) take effect immediately on resumed sessions without requiring session recreation. Explicit per-request overrides live in provider_params which IS persisted.

Merged with provider_params per-turn via RFC 7396 merge-patch.

output_schema: Option<OutputSchema>

Output schema for structured output extraction.

When set, the agent will perform an extraction turn after completing the agentic work, forcing the LLM to output validated JSON. The final response text becomes the schema-only JSON string.

structured_output_retries: u32

Maximum retries for structured output validation failures.

extraction_prompt: Option<String>

Custom prompt for the structured output extraction turn.

When output_schema is set, this prompt is sent as a user message after the agentic loop to elicit schema-valid JSON. Defaults to a built-in prompt if None.

Trait Implementations

impl Clone for AgentConfig

fn clone(&self) -> AgentConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for AgentConfig

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

Formats the value using the given formatter.

impl Default for AgentConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for AgentConfig

where AgentConfig: Default,

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for AgentConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.