Structs§
- Agent
- The core agent — runs the ReAct loop (reason, act, observe).
- Agent
Handle - Handle for controlling the agent from outside.
- Incomplete
Tool Recovery - Next
Action Assessment - Planned
Tool Call - Recovery
Checkpoint - Recovery
Ledger - Recovery
Reconciliation - Timing
Event - Tool
Plan - Turn
State - Minimal visible state for a turn. This is the first slice of making turn state explicit; later slices can add durable IDs and recovery cursors.
Enums§
- Agent
Command - Commands sent to the agent (from UI or orchestrator).
- Agent
Event - Events emitted by the agent during execution.
- Continue
Reason - Why the agent is allowed to spend another turn.
- Incomplete
Tool State - Loop
Decision - Next
Action Debug View - Recovery
Checkpoint Kind - RunFinal
Status - Stop
Reason - Semantic reason the runtime stopped intentionally.
- Timing
Stage - Tool
Execution Mode - Tool
Risk - Tool risk visible before execution. This is deliberately conservative and can be refined as tool metadata grows.
- Turn
Phase - Coarse-grained phase of a single agent turn.