Bug: Agent incorrectly identifies tool calls as empty responses #127
Description
Bug: Agent incorrectly identifies tool calls as empty responses
Description
The current agent sometimes incorrectly identifies a response from the LLM as empty when it actually contains tool calls but no text response. This leads to the agent unnecessarily "scolding" the LLM when it's actually functioning correctly.
Expected Behavior
The agent should only consider a response truly empty if it contains neither text content nor tool calls.
Current Behavior
The agent considers a response empty if it doesn't contain text content, even if it contains valid tool calls.
Impact
This issue causes unnecessary error handling and potentially degrades the user experience by triggering error messages when the LLM is actually working as intended.
Proposed Solution
Update the empty response detection logic to check for both the absence of text content AND the absence of tool calls before considering a response truly empty.