Struct SubAgentConfig

Source

pub struct SubAgentConfig {Show 32 fields
    pub name: String,
    pub description: String,
    pub system_prompt: String,
    pub tools: Vec<Arc<dyn Tool>>,
    pub context_strategy: Option<ContextStrategy>,
    pub summarize_threshold: Option<u32>,
    pub tool_timeout: Option<Duration>,
    pub max_tool_output_bytes: Option<usize>,
    pub max_turns: Option<usize>,
    pub max_tokens: Option<u32>,
    pub response_schema: Option<Value>,
    pub run_timeout: Option<Duration>,
    pub guardrails: Vec<Arc<dyn Guardrail>>,
    pub provider: Option<Arc<BoxedProvider>>,
    pub reasoning_effort: Option<ReasoningEffort>,
    pub enable_reflection: Option<bool>,
    pub tool_output_compression_threshold: Option<usize>,
    pub max_tools_per_turn: Option<usize>,
    pub tool_profile: Option<ToolProfile>,
    pub max_identical_tool_calls: Option<u32>,
    pub max_fuzzy_identical_tool_calls: Option<u32>,
    pub max_tool_calls_per_turn: Option<u32>,
    pub session_prune_config: Option<SessionPruneConfig>,
    pub enable_recursive_summarization: Option<bool>,
    pub reflection_threshold: Option<u32>,
    pub consolidate_on_exit: Option<bool>,
    pub workspace: Option<PathBuf>,
    pub max_total_tokens: Option<u64>,
    pub audit_trail: Option<Arc<dyn AuditTrail>>,
    pub audit_user_id: Option<String>,
    pub audit_tenant_id: Option<String>,
    pub audit_delegation_chain: Vec<String>,
}

Expand description

Configuration for adding a sub-agent to the orchestrator.

Used by OrchestratorBuilder::sub_agent_full to avoid a long parameter list.

Fields§

§name: String

Unique name for this sub-agent.

§description: String

Human-readable description of what this sub-agent does.

§system_prompt: String

System prompt to give this sub-agent.

§tools: Vec<Arc<dyn Tool>>

Tools available to this sub-agent.

§context_strategy: Option<ContextStrategy>

Context window management strategy for this sub-agent.

§summarize_threshold: Option<u32>

Summarize conversation when message count exceeds this threshold.

§tool_timeout: Option<Duration>

Per-tool execution timeout for this sub-agent.

§max_tool_output_bytes: Option<usize>

Maximum tool output size in bytes for this sub-agent.

§max_turns: Option<usize>

Per-agent turn limit. When None, uses orchestrator default.

§max_tokens: Option<u32>

Per-agent token limit. When None, uses orchestrator default.

§response_schema: Option<Value>

Optional JSON Schema for structured output. When set, the sub-agent receives a synthetic __respond__ tool and returns structured JSON.

§run_timeout: Option<Duration>

Optional per-agent run timeout. When None, no timeout is applied to this sub-agent’s run.

§guardrails: Vec<Arc<dyn Guardrail>>

Guardrails applied to this sub-agent’s LLM calls and tool executions.

§provider: Option<Arc<BoxedProvider>>

Optional per-agent LLM provider override. When None, the sub-agent inherits the orchestrator’s provider. Use this to route sub-agents to different models (e.g., Haiku for cheap tasks, Opus for complex ones).

§reasoning_effort: Option<ReasoningEffort>

Optional reasoning/thinking effort level for this sub-agent.

§enable_reflection: Option<bool>

Enable reflection prompts after tool results for this sub-agent.

§tool_output_compression_threshold: Option<usize>

Tool output compression threshold in bytes for this sub-agent.

§max_tools_per_turn: Option<usize>

Maximum tools per turn for this sub-agent.

§tool_profile: Option<ToolProfile>

Tool profile for pre-filtering tool definitions for this sub-agent.

§max_identical_tool_calls: Option<u32>

Maximum consecutive identical tool-call turns for doom loop detection.