Wrappers
In the world of Large Language Models (LLM’s), a “wrapper” is like a helper tool that makes it easier to use the model.
Imagine the LLM is a powerful engine, and the wrapper is the user-friendly interface that lets you control and interact with this engine without needing to understand its complex inner workings.
It simplifies the process of asking the LLM questions or getting it to perform tasks, much like how a phone’s touch screen makes it easy to use without knowing the details of how the phone works.
There are different kinds of wrappers, commonly categorized as “thin” and “fat” wrappers. Here’s a simple explanation of each type:
-
- Thin Wrappers:
- Definition: These are lightweight and add minimal functionality around the LLM.
- Purpose: They mainly handle basic tasks like sending requests to the LLM and receiving responses.
- Example: A simple web interface that allows you to type in a question and see the LLM’s answer.
- Fat Wrappers:
- Definition: These are more feature-rich and add significant functionality around the LLM.
- Purpose: They can manage more complex tasks such as user authentication, input validation, error handling, and integration with other software or services.
- Example: A comprehensive chatbot platform that not only lets you interact with the LLM but also tracks conversation history, supports multiple languages, and integrates with customer service databases.
- Thin Wrappers:
Other Kinds of Wrappers
-
- API Wrappers:
- Simplify the process of making API calls to the LLM, often providing additional features like rate limiting and logging
- Custom Task Wrappers:
- Designed for specific tasks such as text summarization, translation, or sentiment analysis. These wrappers make it easier to use the LLM for particular applications.
- Middleware Wrappers:
- Act as intermediaries that connect the LLM with other systems or applications, facilitating smooth data exchange and processing.
- API Wrappers: