llm_guardrails_prompt = As an expert in business intelligence analysis, evaluate the following answer according to these criteria:

QUESTION: {{question}}
ANSWER: {{result}}

Evaluation criteria (rate from 1 to 5):
1. ACCURACY: Are the data and metrics correct?
2. RELEVANCE: Does the answer address the question asked?
3. COMPLETENESS: Are all necessary information present?
4. CLARITY: Is the answer clear and understandable?
5. RELIABILITY: Are there any signs of hallucination or error?
Reply in JSON format:
{
    'scores': {
        'accuracy': X,
        'relevance': X,
        'completeness': X,
        'clarity': X,
        'reliability': X
    },
    'detected_issues': ['list of issues'],
    'recommendations': ['list of improvements'],
    'confidence_level': 'high/medium/low'
}
