![RAG with MongoDB Atlas and Amazon Bedrock | S02E20 | Lets talk about Data Show](https://assets.community.aws/a/2hKVDY002YTOW2tEO5qQLDqLZMr/Scre.webp)
RAG with MongoDB Atlas and Amazon Bedrock | S02E20 | Lets talk about Data Show
Explore how the new integration of MongoDB Atlas with Amazon Bedrock accelerates the development of highly engaging generative AI applications
Ibrahim Emara
Amazon Employee
Published Jun 2, 2024
The webcast demonstrated how to use MongoDB as a vector store with Amazon Bedrock to build a generative AI solution with retrieval-augmented generation (RAG). RAG allows leveraging large language models like those available on AWS while augmenting their knowledge with your own proprietary data stored in MongoDB Atlas. This enables building AI applications that can provide accurate and up-to-date responses by retrieving relevant information from your data sources.
The key steps shown included: creating a MongoDB Atlas cluster, setting up a vector search index, creating a Bedrock knowledge base linked to the Atlas data, generating embeddings of the data, creating a Bedrock agent with instructions on when to use the knowledge base, and then querying the agent which retrieves relevant information from Atlas and generates a natural language response. The ability to combine multiple knowledge bases with an agent was also demonstrated.
While a managed solution like Bedrock abstracts away much of the complexity, understanding the underlying concepts like embeddings, chunking, and semantic search is still valuable. Overall, the integration of MongoDB and Bedrock provides a powerful way to build domain-aware generative AI applications without the overhead of managing the full AI/ML pipeline.
Show Highlights:
- Retrieval-augmented generation (RAG) augments large language models with proprietary data
- Created MongoDB Atlas cluster and vector search index to store data
- Set up Bedrock knowledge base linked to Atlas data and generated embeddings
- Created Bedrock agent with instructions on using the knowledge base
- Queried agent to retrieve relevant info from Atlas and generate natural language response
- Combined multiple knowledge bases with one agent
- Bedrock provides managed experience abstracting away complexities
Loading...
Ibrahim Emara, RDS Specialist Solutions Architect @ AWS
Shane McAllister, Lead Developer Advocate (Global) @ MongoDB
Igor Alekseev, Data & Analytics Partner SA @ 𝐀𝐖𝐒
Any opinions in this post are those of the individual author and may not reflect the opinions of AWS.