Raw user message sent by your backend
“I need a lightweight notebook for work while travelling.”
- intent
- purchase_product
- entity
- notebook
- preference
- lightweight
- missing
- budget
- next
- clarify
APIProcess alive196279c
Raw language in · typed intent out
Send the user’s raw message. IntentParse identifies the intent.
Your backend does not determine the intent first. It sends the original user message and selects only a domain schema; IntentParse returns the intent, facts, constraints, preferences, missing fields and provenance.
You send raw input
Forward the user’s original message. Do not classify or rewrite the intent before calling the API.
IntentParse discovers intent
The parser identifies the intent and separates known facts, bounded inference and unresolved dimensions.
Your application decides
The schema limits the domain; your own authorization still controls every action and side effect.
Raw message in · identified intent out
input → /v1/parse → intent
{
"intent": "purchase_product",
"state": "needs_clarification",
"unresolved_dimensions": ["budget"],
"recommended_action": "clarify"
}FAQ
Direct answers about IntentParse.
What does IntentParse do?
IntentParse converts a raw user message into validated structured intent: goal, entities, facts, inferences, constraints, preferences, unresolved dimensions, confidence, recommended action and provenance.
Does the caller need to identify the intent first?
No. The caller sends the original message and chooses only a domain schema. IntentParse identifies the intent.
Does IntentParse pay for production model inference?
No. Production inference is BYOK: the customer selects and pays the compatible model provider. The free playground uses a separately funded demonstration model.
Can IntentParse authorize an action?
No. IntentParse describes intent state. The customer application remains responsible for authorization and every side effect.