Knowledge Base

As we witness the rapid advancement of AI in this era, AI also play a pivotal role in every organization to assist the customer and the employees.

For instance if an employee of an organization inference a Foundational model about the hierarchy of particular person. The FM may unable to generate the results as this is domain specific prompt that is scoped only within this organization and FM may no be trained on such internal data.

To extract some domain specific data we can make use of a feature AWS Bedrock Knowledge Base. Knowledge base is fully managed service that helps to integrate the company proprietary information into the generative-AI applications using Retrieval Augmented Generation (RAG) technique.

Without using KnowlegeBase

Here the FM model was unable to generate the output and it is asking for more context as it is more specific within that organization.

After using KnoweldgeBase

Below are the steps that happens with knowledge Base

1. The user prompt the query asking for the manager of John Doe.
2. The query goes to the vector DB say Opensearch DB
3. The proprietary documents can be uploaded to the a data source like S3.
4. We use a agent FM model to convert the raw data to vector embedding that can be stored in the DB.
5. The vector DB searches for this query in the DB and augments it with the prompt and send it to the FM model.
6. The FM model then can able to find out the data and return back the response as the manager of John Doe