Building LLMs for Production

What are Agents: Large Models as Reasoning Engines

The recently released large pre-trained models created the opportunity to build agents — intelligent systems that use these models to plan the execution of complex tasks. Agent workflows are now possible because of the greater reasoning capabilities of large pre-trained models such as OpenAI’s latest models.

These models can be used for their deep internal knowledge to create new, compelling material, think through problems, use external tools, and make plans. For instance, you can create a research agent that will find essential facts from different sources and generate an answer that combines them in a helpful way. Agents are systems that use LLMs to determine and order a set of actions. In ...