Expand description
Agent lifecycle management and delegation primitives for Gestura.
gestura-core-agents provides the typed orchestration model for spawning
specialist agents, delegating structured work, and tracking task execution
results across local and remote execution modes.
§Design role
This crate owns the agent/delegation vocabulary used by supervisor-style orchestration without pulling in GUI-specific concerns. It is intentionally a domain crate for types and lifecycle primitives, while presentation-layer integration and higher-level runtime coordination remain elsewhere.
§Main concepts
AgentRole: specialist roles such as supervisor, implementer, reviewer, tester, and remote workerAgentSpawnRequest: spawn-time contract including workspace, execution mode, and advertised capabilitiesDelegatedTask: structured delegated work with approvals, dependencies, reviewer/test gates, memory tags, and remote-target metadataTaskResult: normalized result payload including artifacts and tool-call provenanceAgentManagerandAgentSpawner: lifecycle primitives for creating and managing active agents
§Execution model
Delegated work supports several execution environments:
- shared workspace
- isolated workspace
- git worktree-backed execution
- remote execution targets
This lets the rest of the system reason about isolation and provenance using typed metadata instead of ad hoc strings.
§Stable import paths
Most application code should import these types through gestura_core::agents::*.
GUI-specific orchestration wrappers remain in the GUI crate.
Structs§
- Agent
Envelope - IPC envelope for events exchanged with agents
- Agent
Info - Public agent info for status queries
- Agent
Manager - Core agent manager implementation
- Agent
Spawn Request - Spawn configuration for a managed agent.
- Delegated
Task - Task that can be delegated to a subagent
- Delegation
Brief - Structured brief attached to delegated work.
- Orchestrator
Tool Call - Record of a tool call during orchestrated task execution
- Remote
Agent Target - Remote target details for delegated work that may execute via A2A.
- Task
Artifact Record - Artifact produced by delegated task execution.
- Task
Result - Result from a delegated task
Enums§
- Agent
Command - Commands that can be sent to an agent task
- Agent
Execution Mode - Execution mode used by a subagent.
- Agent
Role - Specialist role assigned to a managed agent.
- Agent
Status - Status value for an agent
- Task
Terminal State Hint - Optional terminal-state mapping returned by delegated execution.
Traits§
- Agent
Spawner - Trait for spawning and managing isolated agents