NaLU AI extracts what the user actually said β name, email, postal code, yes/no β without confusing greetings with data. Integrates in 30 seconds.
β R$ 0,0058 Β· Less than a penny to never save "Good Morning" as a name again.
Agent message + user reply. Two fields. Nothing else.
Multi-layer semantic extraction. Normalized and validated result.
obtained: true + validated value. No regex, no hallucination.
AI validators β multi-layer semantic extraction
Extracts full name, ignores greetings and titles.
Extracts email and fixes domain typos (gmail→gmail.com).
International postal code (non-BR).
Detects yes/no in any language and indirect phrasing.
Date of birth in any format. Detects minors.
Detects intent to speak with a human (urgency 1-3).
Differentiates service cancellation vs. current operation.
Extracts company name. Detects legal suffixes.
Validates CPF with mod 11. Formats XXX.XXX.XXX-XX.
Extracts ZIP code and returns enriched address.
Extracts phone with area code. Validates ANATEL DDDs.
Validates CNPJ with mod 11. Formats XX.XXX.XXX/XXXX-XX.
Analyzes conversational context. Detects counteroffers, handoffs, cancellations.
With NaLU AI's validate_reply, the bot understands that "48" in the context of an installment offer is a counter-proposal β not a dollar amount. Cost per analysis: $0.0020. Less than a penny.
With validate_handoff, the bot identifies that the customer wants to speak with a human β even when they don't say it directly. Cost per detection: $0.0012.
curl https://api.naluai.dev/v1/extract/name \
-H "Authorization: Bearer YOUR_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"agent_input": "What is your name?",
"user_input": "Good morning! My name is John Smith",
"language": "en"
}'
# Response:
# {
# "obtained": true,
# "extracted_value": "John Smith",
# "confidence": "high",
# "certain": true
# }
$0.0012 per validation on Starter plan. β R$ 0,0058
No credit card. No deadline. Setup in 30 seconds.
Create free account β