{
  "context": "Evaluating a Coralogix support response for this user question:\n\nWe can see traces from our OpenAI calls in Coralogix, but AI Center shows no LLM rows. A representative span looks like this: name=\"POST /v1/chat/completions gpt-4o-mini\", kind=CLIENT, attributes={http.request.method=\"POST\", server.address=\"api.openai.com\", url.full=\"https://api.openai.com/v1/chat/completions\", llm.model_name=\"gpt-4o-mini\"}. There are no gen_ai.* attributes on the span. What does AI Center actually look for?",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "mentions-gen-ai-system",
      "description": "The response contains \"gen_ai.system\" (case-sensitive).",
      "max_score": 3
    },
    {
      "name": "mentions-gen-ai-provider-name",
      "description": "The response contains \"gen_ai.provider.name\" (case-sensitive).",
      "max_score": 3
    },
    {
      "name": "mentions-gen-ai-input-messages",
      "description": "The response contains \"gen_ai.input.messages\" (case-sensitive).",
      "max_score": 3
    },
    {
      "name": "trace-span",
      "description": "The response matches the pattern: (?i)(trace|span)",
      "max_score": 3
    },
    {
      "name": "explains-that-ai-center-detects-genai-spans-f",
      "description": "Explains that AI Center detects GenAI spans from legacy gen_ai.system or newer gen_ai.provider.name / gen_ai.input.messages attributes on trace spans, not from generic HTTP span names alone. It should say logs or metrics alone do not create AI Center rows. FAIL if it says model names in span names are sufficient.",
      "max_score": 2
    }
  ]
}