tool_call_bugTier 1 · 70% confidence

ai-agents-tool-call-bug-reusing-the-same-tool-function-name-in-consecutive-c2a36c09

agent: ai_agents

When does this happen?

IF Reusing the same tool function name in consecutive POST requests to /v1/chat/completions causes the server to hang with infinite token generation for the GPT-OSS-120B model.

How others solved it

THEN Ensure each tool call request uses a unique function name to avoid the infinite generation loop. Alternatively, downgrade to vllm version 10.2.0 where the bug is not present, or use a different model (e.g., Mistral, Qwen) which are unaffected.

curl -X POST http://localhost:8000/v1/chat/completions \
  -H "Authorization: Bearer ${VLLM_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-oss-120b",
    "messages": [{"role": "user", "content": "Hi"}],
    "tools": [{"type": "function", "function": {"name": "unique_name_per_request", "description": "...", "parameters": {...}}}]
  }'

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics