- Newest
- Most votes
- Most comments
When working with Amazon Bedrock Knowledge Bases and Salesforce data sources, there are limitations on how fields are processed. Currently, Amazon Bedrock Knowledge Bases supports only one content field, which is split into chunks and converted into vector embeddings. Other fields are treated as metadata and associated with each chunk.
For CSV files, you can control which fields are treated as metadata using inclusion/exclusion fields in a metadata.json file. However, for Salesforce connections specifically, it appears you're experiencing limitations on which fields are automatically included as metadata in your knowledge base.
To improve your data management and ensure more fields are available, you might consider:
-
Using CSV files as an alternative data source format, where you can explicitly designate content fields and metadata fields using columns. This is considered a best practice for large datasets.
-
Adding filters to document fields or attributes to improve the relevancy of responses. Your data sources can include document metadata attributes or fields to filter and specify the fields to embed.
-
If you need to include specific fields from your Salesforce knowledge articles that aren't being automatically included, you might need to export that data to a CSV format first, where you have more control over field inclusion.
The metadata fields in your knowledge base are important as they can be used for filtering during retrieval, which can significantly improve the accuracy and relevance of your responses.
Sources
Include metadata in a data source to improve knowledge base query - Amazon Bedrock
Improve search results and accurate in Amazon Bedrock | AWS re:Post
Relevant content
asked 2 years ago
