Module agent 

Agent Loop Implementation

The agent loop handles the core conversation cycle:

1. User sends a prompt
2. LLM generates a response (possibly with tool calls)
3. If tool calls present, execute them and send results back
4. Repeat until LLM returns without tool calls

This implements agentic behavior where the LLM can use tools to accomplish tasks agentically.

Structs

AgentConfig

Agent configuration

AgentLoop

Agent loop executor

AgentResult

Result of agent execution

IntentDetectionResult

Result from IntentDetection harness

ParallelStepResult

Result of a single parallel step execution, emitted as structured JSON so the frontend can render it dynamically in the user’s language.

TargetHints

Target hints from IntentDetection harness

ToolCommand

Adapter that implements SessionCommand for tool execution via the queue.

Enums

AgentEvent

Events emitted during agent execution