ai-agents-tool-call-bug-sending-two-identical-post-requests-to-v1-chat-com-b67c057d

THEN Use a unique tool function name for each request (e.g., append a timestamp or UUID) to avoid the infinite generation loop. Alternatively, downgrade to vLLM v0.10.2, which does not exhibit this issue, or monitor for a patch in later versions.

# Workaround: unique function name per request
import uuid
tool_name = f"get_weather_{uuid.uuid4().hex}"
tools = [{"type": "function", "function": {"name": tool_name, ...}}]