The SMART (Sequential Multi-Agent Reasoning Technique) system is a powerful pipeline for enhanced language model capabilities. It includes:

• Planning Agent: Prepares incoming user requests for subsequent agents in the chain.
• Reasoning Agent: Handles internal thought processes, planning, and thinking.
• Tool-use Agent: Interfaces with various tools to gather information or perform actions.
• User-interaction Agent: Manages direct communication with the user.

Key features:

• Adaptive model selection based on task complexity
• Multi-step reasoning process
• Integration with external tools
• Customizable prompts for each agent in the chain
• Advanced Planning and internal thought

SMART enhances the problem-solving capabilities of language models by breaking down complex tasks into manageable steps and leveraging specialized agents for different aspects of the process.