Searching a Knowledge Base

Lesson Overview

This video demonstrates how to use Knowledge Stores in AirOps to retrieve your own documents and data for use in workflows. It provides a step-by-step guide on creating a knowledge base, adding content to it, and using it in a workflow to generate informed and relevant outputs.

• 0:00: Introduction to Knowledge Stores and their purpose
• 0:22: Creating a Knowledge Store and adding content
• 1:21: Using a Knowledge Store in a workflow
• 1:59: Setting up an LLM step to answer questions based on retrieved data

Key Concepts

Knowledge Stores

Knowledge Stores are a feature in AirOps that allow you to bring your own documents and content into workflows. This enables you to provide models with your specific domain knowledge, resulting in more differentiated and relevant outputs.

To create a Knowledge Store:

1. Click on "Your Data" in the left-hand sidebar
2. Click "Create Knowledge Store"
3. Add content by uploading files, scraping websites, importing via sitemap, or connecting to Google Drive or a database

Semantic Search

When you have content in your Knowledge Store, you can perform a semantic search to find relevant chunks of text. This is done by clicking "Test Retrieval" and entering a query. AirOps will then return the most semantically similar chunks of text from your Knowledge Store.

Using Knowledge Stores in Workflows

To use a Knowledge Store in a workflow:

1. Create a new workflow
2. Add a "Knowledge Base Search" step
3. Select your desired Knowledge Store and specify the number of results you want
4. Pass in a question or query to find relevant chunks of text
5. Add an LLM step to generate an output based on the retrieved chunks

By using a Knowledge Store in a workflow, you can provide the LLM with specific information relevant to your use case, enabling it to generate more accurate and tailored responses.

Key Takeaways

1. "Knowledge Bases... are stores for your data that you can then provide to your workflows and to your LLM steps to make sure they have all the information they need to create informed, relevant content."
   • Knowledge Stores allow you to bring your own data into AirOps workflows, enabling more personalized and accurate outputs.
2. "So I'm gonna go ahead and create a blank workflow, and the first thing I'm gonna do is set up a question... I'm now gonna go ahead and retrieve chunks from my knowledge base that are semantically similar to the question that is being asked."
   • To use a Knowledge Store in a workflow, start by creating a question, then use a Knowledge Base Search step to find relevant chunks of text based on that question.
3. "I'm now gonna add an LLM step and for the LLM step, I'm gonna ask it to take the retrieve chunks from my AirOps docs knowledge base and answer the question that the user has."
   • After retrieving relevant chunks from your Knowledge Store, use an LLM step to generate a response based on that information, providing a more informed and accurate output.