Zero Shot React Description

Zero-Shot React (Reasoning and Acting) is an agentic framework that enables large language models (LLMs) to reason and act without prior examples or training demonstrations.

In a Zero-Shot React setup, the LLM decides:

• When to think (i.e., perform reasoning steps internally)
• When to act (i.e., call a tool, API, or function)

The agent outputs intermediate thoughts, selects tools, and processes observations dynamically. This makes Zero-Shot React highly flexible for tasks like web scraping, search-based question answering, or tool use in dynamic environments — all without fine-tuning.

It was introduced by OpenAI and LangChain to reduce complexity in orchestration and tool invocation, relying solely on smart prompt formatting and few-shot capabilities built into the model itself.