Using embedding in AI

In this guide, we will explain how to generate an AI response using an embedding, also known as a knowledge base, within our platform.

Accessing the Knowledge Base

1. Find the Integration: On our dashboard, locate the integration for OpenAI.

2. Locate Embeddings: Here you can add or edit embeddings. Please refer to our previous video for detailed instructions on modifying embeddings.

Utilizing Embeddings in a Workflow

Using the Master AI Flow

1. Add an Action: Go to the master AI flow we shared previously and add an action.

2. Select Integration: Under "add action", choose the integration with OpenAI.

3. Edit the Action: Select embedding match and completion as the action to perform.

Setting Up the Request

• Input: Use an initial input, such as "hi".

• Introduction: Provide context to the AI and instructions, for example, "You are a helpful assistant."

• Type: Ensure it matches the type for the embeddings you've added. You may leave this empty if you have a single type.

• Maximum Response Length: Set this according to your requirements.

Testing the Request

• Test Request: Press the test request button to initiate the process.

• AI Processing: The AI will use the provided input to search for relevant information within the knowledge base.

• Generating Response: Based on the retrieved information, the AI will generate a completion as the output for the user.

Saving and Using the Response

• Save the Output: Follow the logic used previously, save the generated completion.

• Review Questions: Ensure that questions relate back to the AI appropriately for consistency.

Embedding Match Explanation

The embedding match function operates similarly to a query response. However, the AI strictly fetches relevant information for potential future usage. For instance, you may want to save this information for creating a custom check completion in the future.