Lex GeneralQA Intent includes <answer>\n<intent>FAQ_QUERY</intent>\n<text>
Hi, I have a Lex BOT that uses GeneralQA Intent with the "Knowledge base for Amazon bedrock" Sometimes the beginning of the responses includes: "<answer>\n<intent>FAQ_QUERY</intent>\n<text>"
Is this a bug?
Response snippet:
{
"messages": [
{
"content": "<answer>\n<intent>FAQ_QUERY</intent>\n<text>Yes, there are several financial aid options available to students at .... (rest of the response text) ",
"contentType": "PlainText"
}
- Newest
- Most votes
- Most comments
I understand that you are having a Lex BOT that uses GeneralQA Intent with the "Knowledge base for Amazon bedrock" Sometimes the beginning of the responses includes: "<answer>\n<intent>FAQ_QUERY</intent>\n<text>" This appears to be XML/markup formatting being included in the response that should typically be stripped out. While not technically a bug, it's metadata that shouldn't be visible to end users. To handle this, -Use a Lambda function as a fulfillment hook to clean/parse the response before sending it to the user. Resource: https://docs.aws.amazon.com/lexv2/latest/dg/built-in-intent-qna.html
For further queries , Contact AWS Support , they would be able to help you with better context and resource.
This is not a bug but rather part of the expected output format for the AMAZON.QnAIntent in Amazon Lex.
When using the AMAZON.QnAIntent with a knowledge base (like Amazon Bedrock knowledge base), the response includes structured metadata that helps categorize the type of response being provided. The format you're seeing (`<answer>`, `<intent>FAQ_QUERY</intent>`, `<text>`) is part of the output instruction format that the model uses to structure its responses.
This format indicates:
- `<intent>FAQ_QUERY</intent>` - Shows that the model classified this as a frequently asked question
- `<text>` - Introduces the actual response content
The complete format typically includes additional elements like passage_id and confidence level. This structured format is designed to provide both the answer and metadata about how the answer was derived.
If you don't want this metadata to appear in your user-facing responses, you'll need to process the response in a Lambda function or in your client application to extract just the text portion before displaying it to your users.
Sources
AMAZON.QnAIntent - Amazon Lex
Relevant content
- asked 2 years ago
- asked 3 years ago
- AWS OFFICIALUpdated 4 years ago
Thanks! This seems to happen when the response is long and truncated.