Module agent 

Agent definition types and workflow composition.

Agents are defined as data (AgentDefinition + AgentKind) and composed declaratively. The framework handles execution — products only define what the agent knows and can do.

Example

let agent = AgentDefinition {
    name: "researcher".into(),
    kind: AgentKind::Llm(LlmAgentConfig {
        system_prompt: "You are a research assistant.".into(),
        tools: vec![Arc::new(WebSearch)],
        lens: Lens::new(["research", "papers"]),
        llm_config: LlmConfig::new("openai", "gpt-4"),
        experience_extractor: None,
    }),
};

Structs

AgentDefinition

Blueprint for creating an agent. Not a running agent — just configuration.

ExtractionContext

Context passed to the experience extractor during the Record phase.

LlmAgentConfig

Configuration for an LLM-powered agent.

Enums

AgentKind

Determines how an agent executes.

AgentKindTag

Compact tag for agent kind, used in HiveEvent variants.

AgentOutcome

Outcome of an agent’s execution.

Traits

ExperienceExtractor

Trait for extracting experiences from an agent’s conversation history.