Emitted when an agent resumes execution after receiving human input.
Emitted when an agent pauses execution to wait for human input (Blocking Tool).
Records an error encountered during any phase of execution.
Records the final AI response message generated by the agent.
The user's inferred intent.
Records status changes of a specific todo item.
Logged by Agent Core upon receiving an END stream event.
Logged by Agent Core upon receiving an ERROR stream event. Content should be Error object or message.
Logged by Agent Core upon receiving a METADATA stream event. Content should be LLMMetadata.
Logged by Agent Core when LLM stream consumption begins.
The agent's step-by-step plan to address the intent.
Records updates to the plan structure (intent, title, or list changes).
Records changes made to the agent's persistent state.
Emitted if a suspension times out (reserved for future use).
Records events specifically related to the synthesis phase (e.g., the LLM call).
The agent's internal monologue or reasoning process.
The generated concise thread title.
Records the LLM's decision to call one or more tools (part of the plan).
Records the actual execution attempt and result of a specific tool call.
Represents the type of an observation record, capturing significant events during agent execution.