Struct TaskAgentConfig 

#[non_exhaustive]
pub struct TaskAgentConfig {

    pub max_iterations: u32,
    pub system_prompt: Option<String>,
    pub temperature: f32,
    pub max_tokens: u32,
    pub validation_config: Option<ValidationConfig>,
    pub loop_detection: Option<LoopDetectionConfig>,
    pub goal_revalidation_interval: Option<u32>,
    pub max_replan_attempts: u32,
    pub max_total_tokens: Option<u64>,
    pub max_cost_usd: Option<f64>,
    pub timeout_secs: Option<u64>,
    pub allowed_files: Option<Vec<PathBuf>>,
    pub plan_budget: Option<PlanBudget>,
    pub context_budget_tokens: Option<u64>,
}

Configuration for a task agent.

Fields (Non-exhaustive)

Non-exhaustive structs could have additional fields added in future. Therefore, non-exhaustive structs cannot be constructed in external crates using the traditional Struct { .. } syntax; cannot be matched against without a wildcard ..; and struct update syntax will not work.

max_iterations: u32

Maximum provider call iterations before the agent is forced to fail.

Default: 100 (high default to avoid artificial limits on complex tasks).

system_prompt: Option<String>

Override the system prompt.

When None, crate::system_prompts::reasoning_agent_prompt is used.

temperature: f32

Temperature for AI calls (0.0 – 1.0).

max_tokens: u32

Maximum tokens for a single AI response.

validation_config: Option<ValidationConfig>

Quality checks to run before accepting completion.

Set to None to disable validation entirely (useful in tests).

loop_detection: Option<LoopDetectionConfig>

Loop detection settings. None disables. Default: 5-call window, enabled.

goal_revalidation_interval: Option<u32>

Inject goal-reminder every N iterations. None disables. Default: Some(10).

max_replan_attempts: u32

Abort after this many REPLAN cycles. Default: 3.

max_total_tokens: Option<u64>

Abort when cumulative tokens reach this ceiling. Default: None.

max_cost_usd: Option<f64>

Abort when cumulative cost (USD) reaches this ceiling. Default: None.

timeout_secs: Option<u64>

Wall-clock timeout for the entire execute() call, in seconds. Default: None.

allowed_files: Option<Vec<PathBuf>>

Per-agent file scope whitelist.

When Some, the agent receives a scope-violation error for any file operation targeting a path that is not prefixed by at least one entry in this list. When None, file access is unrestricted.

Uses Path::starts_with for prefix matching, which is component-aware: "/src" allows "/src/main.rs" but denies "/src_extra/file.txt".

plan_budget: Option<PlanBudget>

Optional pre-execution budget check.

When Some, the agent asks the provider to produce a structured JSON plan before starting execution. The plan is validated against the budget constraints; if any constraint is exceeded the run fails immediately with FailureCategory::PlanBudgetExceeded before any file or tool side-effects occur.

Set to None (the default) to skip the planning phase entirely.

context_budget_tokens: Option<u64>

Context budget in tokens.

When the estimated conversation token count exceeds this value, the on_context_pressure hook is called so the consumer can summarize or evict messages.

Only effective when lifecycle hooks are set on the AgentContext. Default: None (no context pressure callbacks).

Trait Implementations

impl Clone for TaskAgentConfig

fn clone(&self) -> TaskAgentConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TaskAgentConfig

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

Formats the value using the given formatter.

impl Default for TaskAgentConfig

fn default() -> Self

Returns the “default value” for a type.