Agent Lifecycle¶
Ravi’s default execution model is the ReAct loop: Think → Act → Observe.
Loop phases¶
Think¶
The model receives the current prompt context and decides whether it can answer directly or needs tools.
Act¶
If tool calls are returned, the runtime validates arguments and executes tools in order.
Observe¶
Tool results are appended to memory so the next model call can incorporate what happened.
Guardrails¶
The loop has three natural control points:
- input validation before the model call
- tool-call validation before execution
- output checks before the final response is returned
Why it matters¶
This structure keeps the agent predictable, inspectable, and easy to stream to a UI.