[BUG]Advanced document analysis does not allow chunk text to be retrieved correctly

[HirokiNakatani]

🚨 Please Note 🚨

To ensure efficient investigation of the issue, please fill out the fields below with as much detail as possible. Reports that do not follow this template may be closed without notification. We appreciate your cooperation.

🐞Describe the bug

When using advanced document parsing (sonet3.5), I cannot get chunked text correctly

🔄 To Reproduce

1. Prepare a knowledge base that uses advanced document analysis (sonet3.5)
2. Create a bot that links the knowledge base from 1.
3. Talk to the bot

📷 Screenshots

🔎 Logs for Chat Issues

If the issue occurs during a chat interaction, please check the following logs on Amazon Cloudwatch Logs and include the relevant entries in your issue:

• /aws/lambda/BedrockChatStack-BackendApiHandlerXXXX
• /aws/lambda/BedrockChatStack-WebSocketHandlerXXXX

🔎 Logs for Bot Creation/Update Issues

If the issue occurs during bot creation or updating, please check the execution records of the AWS Step Functions state machine named EmbeddingStateMachineXXX and include the details in your issue.

📝 Additional context

When the content is IAMGE, the text seems to be stored in x-amz-bedrock-kb-description, but I think this is not being taken into consideration.

Could you please take the following into consideration?

[statefb]

@HirokiNakatani Thank you for the reporting. Could you create a PR with the guideline? Thank you (P.S. We will release v3 soon, so it might be better to wait for it before creating a PR)

[HirokiNakatani]

Sorry, I don't know how to do PR.