Nevola Engine is an API for content brands that personalize at scale. It turns user profiles and temporal-contextual signals into per-request content your developers ship in an afternoon.
In production with content brands across Europe
It usually starts small. A feature flag here, a model prompt pasted into a controller there, a handful of analytics events to decide who sees what. Six months later you have personalization logic spread across three services, two databases, and a Notion doc nobody trusts. Every new content variant means a deploy.
The expensive part isn't the model. It's the drift. Your profile schema says tone: warm but the prompt still expects voice: friendly, and nobody notices until a subscriber gets the wrong content for a week. There's no replay, so you can't reproduce what a user actually received last Tuesday. There's no audit trail, so when legal asks how a piece of content was generated, the honest answer is a shrug.
A profile is the durable record of who a subscriber is to your brand: preferences, cohort, the external ID you already use. You write to it with POST /v1/profiles and patch it as you learn more. We store a hashed external_user_id and the attributes you send — no other PII touches our database.
POST /v1/profiles
{
"external_user_id": "u_8f21c",
"attributes": { "tone": "warm" }
}
Signals are temporal-contextual values derived from a timestamp plus the context you pass in. GET /v1/signals/temporal/{user_id} returns the computed cohort and weighting the engine will use for that moment — deterministic, versioned, and inspectable. You never guess why a subscriber landed in a given variant; the signal is right there in the response.
GET /v1/signals/temporal/u_8f21c
→ { "cohort": "morning_warm",
"weight": 0.72,
"version": "tcs.v3" }
Templates define the shape of a personalization; sequences chain them into multi-step delivery like an onboarding arc or a daily drop. Call POST /v1/personalize/content for a single piece, or POST /v1/personalize/sequence for the whole run. Delivery confirmations arrive on the personalize.delivered webhook, so your system knows exactly what shipped.
POST /v1/personalize/sequence
{ "profile_id": "u_8f21c",
"sequence_id": "onboarding_7d" }
Create a profile, then ask for content. The engine reads the profile, computes the temporal-contextual signal for the request, and returns finished copy with the signal attached. Same flow whether you call it with curl or the SDK.
# 1. Create a profile
curl -X POST https://api.nevolagroup.com/v1/profiles \
-H "Authorization: Bearer $NEVOLA_API_KEY" \
-H "Content-Type: application/json" \
-d '{"external_user_id": "u_8f21c", "attributes": {"tone": "warm", "cadence": "daily"}}'
# 2. Get personalized content for this moment
curl -X POST https://api.nevolagroup.com/v1/personalize/content \
-H "Authorization: Bearer $NEVOLA_API_KEY" \
-H "Content-Type: application/json" \
-d '{"profile_id": "u_8f21c", "template": "daily_reflection", "context": {"local_time": "2026-03-14T07:02:00-03:00"}}'import { Nevola } from "@nevola/engine";
const nevola = new Nevola({ apiKey: process.env.NEVOLA_API_KEY! });
await nevola.profiles.create({
externalUserId: "u_8f21c",
attributes: { tone: "warm", cadence: "daily" },
});
const result = await nevola.personalize.content({
profileId: "u_8f21c",
template: "daily_reflection",
context: { localTime: "2026-03-14T07:02:00-03:00" },
});
console.log(result.body); // finished copy
console.log(result.signal); // the temporal-contextual signal that shaped itFull request and response shapes are in the API reference.
This is where the engine earns its keep. A brand sending daily reflection content to 40,000 subscribers can't write 40,000 drafts, and a single generic one defeats the point of subscribing. With Nevola, each subscriber gets a narrative drop shaped by their profile and the moment they open it — morning cadence reads differently from evening, a three-month subscriber differently from a week-one signup. The brand defines the template once and lets the engine handle the per-request work, with every delivery logged and replayable.
Adaptive learning paths usually mean a maze of conditionals nobody wants to maintain. A learner who aces the first three modules shouldn't sit through the remedial track, and a learner who stalls shouldn't be pushed ahead. Sequences let you express the path as templates that branch on profile attributes, so the engine serves the next step that fits — and the audit trail shows exactly why.
Context is everything in coaching, and a checkpoint that ignores it feels robotic. A user checking in at 6am before a workout needs a different prompt than one logging in after a hard day. Signals give the app a clean way to read the moment and return a contextualized checkpoint, without hard-coding time-of-day rules into the client.
All plans include the full API. No feature gating.
Read the docs, write your first personalization, and decide for yourself.