Declare
A typed function or API schema defines one narrow responsibility.
NEOSYNTROPY API / TUNABLE INTELLIGENCE
Call schema extraction, reasoning, and routing today. Keep the outputs your application accepts, then use those reviewed examples to tune a compact model behind the same stable endpoint.
const response = await fetch(
`https://api.neosyntropy.com/api/v1/schema`,
{
method: "POST",
headers: {
Authorization: `Bearer ${process.env.NEOSYNTROPY_API_KEY}`,
"Content-Type": "application/json",
},
body: JSON.stringify({
"prompt": "Extract the refund request.",
"input": {
"text": "Refund $40 for order ORD-204. The item arrived damaged."
},
"output_schema": {
"type": "object",
"properties": {
"order_id": {
"type": "string"
},
"amount": {
"type": "number"
},
"reason": {
"type": "string"
}
}
}
}),
},
);
if (!response.ok) throw new Error(await response.text());
const result = await response.json();
// Uncertainty policy: never execute an uncertain or unfamiliar result.
const confidence = Number(result.confidence ?? result.output?.confidence ?? 0);
const outOfDistribution = Boolean(
result.out_of_distribution ?? result.output?.out_of_distribution,
);
if (outOfDistribution || confidence < 0.8) {
throw new Error(
`Unsafe AI result: confidence=${confidence}, out_of_distribution=${outOfDistribution}`,
);
}
console.log(result);Illustrative decoded output. Your application can reject low-confidence or out-of-distribution results before they continue.
/api/v1/schemaYOUR TASK. DEFINED BOUNDARIES.DECLARATIVE TUNING / SMALL MODELS
Start with hosted intelligence, keep the outputs your application accepts, and turn those reviewed examples into a compact model for the same API contract.
How model-backed nodes work โOnly executions that pass the schema, critic, and review gates enter the task-specific dataset.
A typed function or API schema defines one narrow responsibility.
Schema checks and critics turn evaluated executions into a reviewed training set.
NeoSyntropy trains a lightweight adapter on a compact base model.
The same contract runs on your specialized model inside the controlled runtime.
You keep the contract, the execution path, and the audit trail. The model behind the node can improve without rewriting the application around it.
THE API CAPABILITIES
Send an input and output schema to extract validated JSON. The dedicated schema endpoint keeps extraction separate from reasoning and routing.
POST /api/v1/schemaRun ordered steps that advance only after router validation. Tool requests pause with a standard function call and resume when your application submits the result.
POST /api/v1/reasoningSend context, a question, and candidate answers. The rank endpoint returns every candidate in probability order so your application can validate the top match before routing work.
POST /api/v1/rankSign in, create a project, and issue a key with framework:invoke access.
Choose available project models or connect vendor credentials through BYOK.
Copy a server-side HTTP example. Keep your key out of browser code.
Check status codes, validate output, and decide what your application executes next.