Struct RuntimeAgent 

pub struct RuntimeAgent {
    pub system_prompt: String,
    pub model: String,
    pub tools: Vec<ToolDefinition>,
    pub max_iterations: usize,
    pub temperature: Option<f32>,
    pub max_tokens: Option<u32>,
    pub tool_search: Option<ToolSearchConfig>,
    pub prompt_cache: Option<PromptCacheConfig>,
    pub driver_options: HashMap<String, Value>,
    pub network_access: Option<NetworkAccessList>,
    pub parallel_tool_calls: Option<bool>,
    pub conversation_context: Option<String>,
}

Runtime configuration for the agent loop

Fields

system_prompt: String

System prompt that defines the agent’s behavior

model: String

Model identifier (e.g., “gpt-5.2”, “claude-opus-5”)

tools: Vec<ToolDefinition>

Available tools for the agent

max_iterations: usize

Maximum number of tool-calling iterations (prevents infinite loops)

temperature: Option<f32>

Temperature for LLM sampling (0.0 - 2.0)

max_tokens: Option<u32>

Maximum tokens to generate per response

tool_search: Option<ToolSearchConfig>

Tool search config (set by openai_tool_search capability)

prompt_cache: Option<PromptCacheConfig>

Prompt caching config (set by prompt_caching capability)

driver_options: HashMap<String, Value>

Driver-namespaced opaque per-call options ("<driver-id>/<option>"), e.g. provider-executed server tools contributed by the openrouter_server_tools capability. Shapes are owned by the respective driver crates.

network_access: Option<NetworkAccessList>

Merged network access list (harness ∩ agent ∩ session). Used by tools (web_fetch) to enforce URL access policy.

parallel_tool_calls: Option<bool>

Request-level parallel tool calling preference (EVE-598).

None (default) preserves provider defaults and the act scheduler’s class-aware concurrent schedule. Some(true) explicitly signals the provider that parallel tool calls are wanted; Some(false) asks the provider to emit at most one tool call per turn AND forces the act scheduler to serialize the batch (see ActInput.parallel_tool_calls).

conversation_context: Option<String>

User-visible conversation context (e.g. hierarchical AGENTS.md). Renders as the leading user-role message of every turn: model-visible and re-resolved alongside the system prompt, but never folded into the cached system prompt.

Implementations

impl RuntimeAgent

pub fn new(system_prompt: impl Into<String>, model: impl Into<String>) -> Self

Create a new runtime agent configuration with required fields only

Trait Implementations

impl Clone for RuntimeAgent

fn clone(&self) -> RuntimeAgent

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RuntimeAgent

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

Formats the value using the given formatter.

impl Default for RuntimeAgent

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for RuntimeAgent

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for RuntimeAgent

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

where __S: Serializer,

Serialize this value into the given Serde serializer.