Help us improve the AWS re:Post Knowledge Center by sharing your feedback in a brief survey. Your input can influence how we create and update our content to better support your AWS journey.
AWS Bedrock agent "Sorry I am unable to assist you with this request"
I have this weird response from the bedrock agent once in a while.
My take is that it is somehow related to the number of traces exceeding 7 (I never saw 8 traces before).
- Newest
- Most votes
- Most comments
Is it possible to share the trace logs?
This issue can arise for multiple reasons. For example, the maximum limit may have been reached. However, based on the information that splitting the input into smaller sub-inputs resolves the problem, it seems that one of the following could be true:
- The question or task is outside the agent's trained scope or knowledge base.
- The agent hasn’t been trained on the specific information or context required to answer the query.
- The query might be too complex or ambiguous for the agent to process.
To address the issue, which you have already figured out by dividing the input into smaller sub-inputs, I would also recommend the following:
- Ensure that your query is clear, specific, and within the agent’s capabilities.
- Provide additional context or break down the query into simpler parts.
For example, instead of asking: "Tell me everything about AWS services." Try asking: "What are the key features of AWS Lambda?"
Hi, Firstly, I'm afraid that the trace logs cannot be shared due to some confidentiality policy. My sincere apology for not being able to provide more details about the chatbot answer trace.
Regarding your concerns, I validated these points in my test:
- The agent was given enough instruction and information to achieve the task.
- The trace logs shows no error. Therefore, my take is that there was some kind of outer hard threshold that made the agent stop rather than the logic inside of the agent flow. (Could be number of API allowed or number of steps).
I understand that sharing logs is not an option here. Since you mentioned that there is no failure reason in the logs, my thoughts incline towards the service quotas and limits mentioned on below page.
https://docs.aws.amazon.com/general/latest/gr/bedrock.html#limits_bedrock
Relevant content
- asked a year ago
Also, when I type in again or separate my request into smaller sub-requests, it worked without any issue.