Intent Classification with Confidence

Routes a user message to one of N categories with a confidence score. Returns "unclear" rather than guessing when uncertain.

The prompt

Classify the user message into exactly one category from this list:
{{categories}}

Return JSON:
{
  "category": one of the listed categories OR "unclear",
  "confidence": 0.0-1.0,
  "reasoning": one sentence
}

Use "unclear" if confidence would be below 0.6. Do not invent categories not in the list.

User message: {{message}}

Recommended models

• GPT-5 Nano
• Claude Haiku 4.5
• Gemini 2.5 Flash

31 upvotes · 0 views · Classification