What is Zero-Shot (ReAct) agent
ReAct (Reason+Act) is a collaboration between Language Models (LLMs) and external tools. It synergizes reasoning and acting in language models. The process involves:
- Reasoning: Understanding the context or the question.
- Acting: Taking an action based on the reasoning, such as searching or looking up information.
For instance, if you ask, "Who wrote the novel 'Pride and Prejudice'?", the model first reasons that it needs to identify the author of the book. It then acts by searching the web to find that the author is Jane Austen.
How to use Zero-Shot agent on Spell
To use Zero-Shot agent it’s enough to choose it from the dropdown during agent creation:
The choice of plugins is crucial for this agent because it will try to accomplish its goals through the best choice of tools.
Zero-Shot: additional considerations
Each framework and system has its advantages and disadvantages. Check out the breakdown below to learn what to look out for with Zero-Shot!
Pros:
- Cost: Zero-Shot is very simple in operation and does not require sophisticated reasoning iterations
- Reliability: Because there is no complex reasoning in place, the agent can stay on track
Cons:
- No memory: The agent does not have the context of its own decision making - it’s always about here and now
- Can’t handle complex tasks: This agent is not made to think, but to act. It performs best in straightforward tasks with multiple tools.