bedrock data automation failing

The issue you're experiencing with syncing .mp3 and .mp4 files in your Bedrock Knowledge Base Data Automation is likely not related to file format support, as you've mentioned these files work in the Demo section. Instead, the problem may be related to permissions or configuration settings.

Here are a few things to check and try:

1. Ensure that your IAM role associated with the Bedrock Knowledge Base has the necessary permissions to access and process these specific file types in your S3 bucket. This includes s3:GetObject and s3:ListBucket permissions.

2. If you're using server-side encryption on your S3 bucket with a KMS key, make sure the IAM role has kms:Decrypt permissions for that key.

3. Verify that the files don't exceed the maximum ingestion job file size specified in Amazon Bedrock endpoints and quotas.

4. Check that your knowledge base configuration, including the vector embeddings model, vector store, and file formats, is properly set up for these media file types.

5. If you're using Amazon OpenSearch Serverless as your vector store, confirm that the vector index is configured with the 'faiss' engine.

6. Ensure that your S3 bucket and the model customization job are in the same AWS region.

7. Try refreshing the Bedrock console page or waiting a few minutes, as sometimes the console may not immediately reflect the actual state of operations.

If the issue persists after checking these points, you may want to try initiating a sync using the AWS CLI or SDK with the UpdateDataSource API call, followed by a StartIngestionJob call. This could potentially bypass any console interface issues.

If none of these solutions work, it would be advisable to contact AWS Support for further assistance. They can provide more specific troubleshooting based on your account setup and configuration.
Sources
Bedrock knowledge base data source sync is not working | AWS re:Post
Bedrock Data source sync does not work | AWS re:Post