ReAct Agent Implementation — Reasoning & Action Loop

I will implement a ReAct agent — building the thought-action-observation reasoning loop with your specified tools, configuring the system prompt for effective reasoning elicitation, implementing stopping conditions, and adding reasoning trace logging for transparency and debugging. ReAct agents that produce high-quality reasoning require careful system prompt design — the prompt must elicit explicit thought steps before action selection, clear action format specification matching your tool schemas, and explicit observation integration before the next thought step — without this structure, the LLM collapses the reasoning loop into direct action selection, losing the benefit of explicit reasoning.

The implementation covers system prompt design for ReAct reasoning elicitation, tool schema integration, thought-action-observation loop implementation, stopping condition logic, reasoning trace logging, and LangChain or custom implementation depending on your stack.