Introduction
Assistants
Introduction
Assistants
Define the context for how you want AI to execute your instructions
What are Assistants?
- Definition: Assistants are where you define the context for how you want to execute prompts. In your assistant you can configure a persona, audience and language model to execute instructions against.
- Purpose: It allows you to define different contexts around how you want your prompt to execute. The persona defines the role you want AI to play, the audience defines how you want the responses tailored or framed, and the language model binds the instructions to a particular model. This allows you to easily run or test a single prompt in different contexts.
Why is it Important?
- Flexibility: Provides an easy way to execute or test a prompt under different scenarios.
- Predictability: Allows for more predictable responses that specific audiences will understand.
- Ease of use: The complexities of creating an assistant are hidden when using the playground as they are created dynamically.
Components of Assistants
Key elements that are configurable for Assistants:
- Personas: Pre-defined personalities that guide the tone and style of AI responses.
- Audiences: Customized settings that tailor the AI’s communication for different user demographics.
- Language Models: Select the language model that you want instructions to run against.
- Vector Stores: Select a vector store that contains documents and files that can be used to provide the LLM with additional context that it has not been trained on.
Example
1. Creating writing assistant with context from previous blog posts:
Humorous Marketing Assistant
- Creative Writing Assistant: This example sets up an assistant that specializes in creative writing and has been provided with context from multiple blog posts uploaded as documents to the associated vector store .
Summary
Assistants allow you to easily execute prompts in different contexts. To help users that are new to AI get up and running quickly, assistants are generated dynamically when creating playground threads. These assistants can be modified later once users get more familiar with the concepts.