Issue Summary

On rare occasions when search does not return results or when search is triggered multiple times in a row, the AI Agent's answers do not completely align with the knowledge connected to the agent. This manifests in two primary ways: # The agent adds specific details to its answer that are not included in the provided knowledge sources.

1. The agent makes assumptions that are not fully covered or supported by the connected knowledge.

Steps to Reproduce
Note: As this is noted as a "rare occasion," specific reproduction steps may vary, but generally involve: # Configure an AI Agent with a specific set of knowledge documents.

1. Query the agent on a topic covered by those documents.
2. Observe the response for details or assumptions that exist outside the scope of the provided text.

Expected Results
The AI Agent should strictly adhere to the provided knowledge base and avoid adding external details or making unsupported assumptions.

Actual Results
When search is selected by the agent several times in a row or when zero search results are returned by the agent search, the AI Agent occasionally "hallucinates" details or extends its logic beyond the provided context.

Workaround
Currently there is no known workaround for this behavior. A workaround will be added here when available