Struct ExecuteRequest 

pub struct ExecuteRequest {

    pub initial_message: String,
    pub event_tx: Sender<AgentEvent>,
    pub cancel_token: CancellationToken,
    pub tools: Option<Arc<dyn ToolExecutor>>,
    pub provider_override: Option<Arc<dyn LLMProvider>>,
    pub model_roster: ModelRoster,
    pub reasoning_effort: Option<ReasoningEffort>,
    pub auxiliary_model_resolver: Option<Arc<dyn Fn() -> AuxiliaryModelConfig + Send + Sync>>,
    pub disabled_filter_resolver: Option<Arc<dyn Fn() -> (BTreeSet<String>, BTreeSet<String>) + Send + Sync>>,
    pub disabled_tools: Option<BTreeSet<String>>,
    pub disabled_skill_ids: Option<BTreeSet<String>>,
    pub selected_skill_ids: Option<Vec<String>>,
    pub selected_skill_mode: Option<String>,
    pub image_fallback: Option<ImageFallbackConfig>,
    pub gold_config: Option<GoldConfig>,
    pub guardian_config: Option<GuardianConfig>,
    pub guardian_spawner: Option<Arc<dyn GuardianSpawner>>,
    pub bash_resume_hook: Option<Arc<dyn BashResumeHook>>,
    pub app_data_dir: Option<PathBuf>,
}

Per-request parameters for agent execution.

Required fields (initial_message, event_tx, cancel_token) must always be provided. The provider is taken from AgentRuntime::provider; tools default to AgentRuntime::default_tools when None.

Fields

initial_message: String

event_tx: Sender<AgentEvent>

cancel_token: CancellationToken

tools: Option<Arc<dyn ToolExecutor>>

Override runtime’s default_tools. When None, uses the runtime’s default tool executor.

provider_override: Option<Arc<dyn LLMProvider>>

Override the LLM provider for this execution. When None, uses the runtime’s shared provider handle.

model_roster: ModelRoster

Cohesive primary + auxiliary model/provider selection. Replaces the old model / provider_name / provider_type / fast_model(+provider) / background_model(+provider) / summarization_model(+provider) clump. Per-role None preserves the same None → Config::get_* fallbacks applied during AgentRuntime::execute.

reasoning_effort: Option<ReasoningEffort>

auxiliary_model_resolver: Option<Arc<dyn Fn() -> AuxiliaryModelConfig + Send + Sync>>

Optional per-round resolver for auxiliary model settings that should be re-read from live global config between rounds.

disabled_filter_resolver: Option<Arc<dyn Fn() -> (BTreeSet<String>, BTreeSet<String>) + Send + Sync>>

Optional per-round live resolver for the disabled tool/skill sets (#136). When None, the snapshotted disabled_tools/disabled_skill_ids are used.

disabled_tools: Option<BTreeSet<String>>

When None, falls back to Config::disabled_tool_names().

disabled_skill_ids: Option<BTreeSet<String>>

When None, falls back to Config::disabled_skill_ids().

selected_skill_ids: Option<Vec<String>>

selected_skill_mode: Option<String>

image_fallback: Option<ImageFallbackConfig>

gold_config: Option<GoldConfig>

guardian_config: Option<GuardianConfig>

Optional guardian adversarial-review gate configuration.

guardian_spawner: Option<Arc<dyn GuardianSpawner>>

Late-bound spawner for the guardian reviewer child (wired by the server; the runner cannot construct a child directly).

bash_resume_hook: Option<Arc<dyn BashResumeHook>>

Late-bound hook that arranges a self-resume after a background-bash suspend (issue #84 Phase 2b). Wired by the server.

app_data_dir: Option<PathBuf>

Bamboo application data directory (typically ~/.bamboo).