Amazon Bedrock has recently added new features to its models and agents to enhance the Recall Augmented Generation (RAG) process. These updates allow for more relevant, accurate, and personalized responses. In addition to Amazon S3, users can now connect web domains, Confluence, Salesforce, and SharePoint as data sources to RAG apps, which is currently in preview mode.
By including web domains, users can provide RAG apps access to public data, such as social media feeds, to improve response quality. With the new connectors, companies can now integrate data sources from Confluence, Salesforce, and SharePoint into their RAG apps for more comprehensive information retrieval.
To add a web domain as a data source, users can navigate to the Bedrock Console and create a knowledge base, selecting the desired data source such as the Web crawler. Configuring the web crawler involves defining source URLs, encryption settings, sync scope, and other parameters. Once the setup is complete, the knowledge base details can be monitored to check the data source synchronization status.
Connecting Confluence as a data source follows a similar process, where users provide the necessary information, choose the hosting method, and select authentication options. Like with the web domain setup, metadata settings can be configured to control content scope and fragmentation.
The new data source connectors are available in all AWS Regions where Knowledge Bases for Amazon Bedrock is available. Users are encouraged to try out these new features in the Bedrock Console and provide feedback through AWS channels or the generative AI developer community.
In conclusion, the integration of web domains, Confluence, Salesforce, and SharePoint as data sources in Amazon Bedrock’s models and agents allows for improved data retrieval and response accuracy. These enhancements provide users with more control over the information utilized in the RAG process, ultimately resulting in more personalized and relevant interactions.
Article Source
https://aws.amazon.com/blogs/aws/knowledge-bases-for-amazon-bedrock-now-supports-additional-data-connectors-in-preview/