Expand description
Parallel tool execution within a single agent turn.
When the LLM returns multiple tool_use blocks, this module partitions them into
execution “waves” based on their concurrency class and runs safe subsets in parallel
via tokio::JoinSet.
Structs§
- Pending
Tool Call - A single pending tool call extracted from the LLM response.
- Tool
Batch Context - Mutable context accumulated across tool executions within a turn.
- Tool
Metrics - Metrics counters for tool execution.
Functions§
- execute_
tool_ batch - Execute a batch of tool calls, optionally in parallel.