True when the run has a conversation to page against.
Disarms the manager (per-run reset).
Executes one tool call with timing + full error containment.
Markdown tool documentation injected into the system prompt (the
{{ _CONVERSATION_TOOLS }} template block), rendered only when the run has a
conversation.
Arms the manager for a run. Pass null conversationId to disable (programmatic runs).
The run's history floor (ExecuteAgentParams.ConversationHistoryFrom).
When set, every tool sees only the rows written at or after it — the same window the
run's messages were loaded from — so a tool can't page back past it.
The run's branch (ExecuteAgentParams.ConversationBranchID); null is the trunk.
Every tool sees only the rows on that branch's path.
The conversation's branch rows, used to resolve the branch path.
Wires the recursive-sub-call seam (BaseAgent owns prompt execution).
Per-run manager for conversation-history retrieval tools. Initialized with the run's conversationId (the availability gate — no conversation, no tools) and serving all reads from the ConversationEngine cache.