Nabs.OpenAi 8.0.80

Nabs OpenAI Library

WARNING: This library is still in development and is not yet ready for use.

An opinionated version with OpenAI APIs.

What this scenarios does this library address?

AI can be described as a set of task that are performed by a machine are are typically performed by humans.

This api is an attempt to solve a specific interaction between our users and the OpenAI API.

At this stage my understanding is the there are some core parts that need to be fulfilled:

• Data that describes the context of the conversation.
• A model that can be used to interpret the data and generate a response.
• A way to manage the interactions between the user and the model such that they can be put together in a way that makes sense.

The Context

We will attempt to create a context that describes the area of specialization of the System. This includes:

• A description of the area of specialization.
• A description of the organization behind the area of specialization.
• The specific approach this organisation has to the area of specialization.
• Terms that define the meaning of key concepts in the area of specialisation.
• The personas representing the organization who will be responsible for responses.
• The content that will be used to inform the generation of responses.
• The policies that govern the way responses are generated.
• The characteristics of the users interacting with the system.

The Model

For this purpose we have selected OpenAI's GPT-4 model. Because the context is quite specific, we will need to fine-tune the model to limit the scope of the responses to the context. We will not be advocating for custom models that require a lot of resources to train.

The Interaction

The interaction need to make sense and also be cost effective. There are two aspects to this interaction:

• Creating and managing embeddings
• Keeping track of the conversations