Struct SessionExecutionArgs 

pub struct SessionExecutionArgs {

    pub agent: Arc<Agent>,
    pub session_id: String,
    pub session: Session,
    pub tools_override: Option<Arc<dyn ToolExecutor>>,
    pub provider_override: Option<Arc<dyn LLMProvider>>,
    pub provider_name: Option<String>,
    pub model: String,
    pub fast_model: Option<String>,
    pub background_model_provider: Option<Arc<dyn LLMProvider>>,
    pub reasoning_effort: Option<ReasoningEffort>,
    pub reasoning_effort_source: String,
    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 cancel_token: CancellationToken,
    pub mpsc_tx: Sender<AgentEvent>,
    pub image_fallback: Option<ImageFallbackConfig>,
    pub runners: Arc<RwLock<HashMap<String, AgentRunner>>>,
    pub sessions_cache: Arc<RwLock<HashMap<String, Session>>>,
}

Arguments for spawning a background agent execution.

This is the crate-agnostic equivalent of the server’s SpawnAgentExecution. It holds everything needed to run the agent loop and persist the result, without depending on HTTP types or AppState.

Fields

agent: Arc<Agent>

session_id: String

session: Session

tools_override: Option<Arc<dyn ToolExecutor>>

provider_override: Option<Arc<dyn LLMProvider>>

provider_name: Option<String>

model: String

fast_model: Option<String>

background_model_provider: Option<Arc<dyn LLMProvider>>

Optional provider override for background/fast model calls.

reasoning_effort: Option<ReasoningEffort>

reasoning_effort_source: String

disabled_tools: Option<BTreeSet<String>>

disabled_skill_ids: Option<BTreeSet<String>>

selected_skill_ids: Option<Vec<String>>

selected_skill_mode: Option<String>

cancel_token: CancellationToken

mpsc_tx: Sender<AgentEvent>

image_fallback: Option<ImageFallbackConfig>

runners: Arc<RwLock<HashMap<String, AgentRunner>>>

sessions_cache: Arc<RwLock<HashMap<String, Session>>>