SDK v0.1 Now Available

The Cognitive Learning
API for Developers

Build intelligent learning experiences with our powerful SDK. Integrate Bayesian Knowledge Tracing, adaptive learning, and cognitive analytics in minutes.

terminal
npm install @cognitive-engine/sdk

Quick Start Guide

Get up and running with the PrepSt SDK in under 5 minutes

1

Install SDK

Add the package to your project

2

Initialize

Configure with your API key

3

Start Building

Use our powerful APIs

# Install the SDK
npm install @prepst/sdk

# Initialize in your project
import { PrepSt } from '@prepst/sdk'

const engine = new PrepSt({
  apiKey: process.env.PREPST_API_KEY
})

# Start tracking learning
const result = await engine.track({
  userId: 'user_123',
  conceptId: 'algebra_basics',
  correct: true,
  timeSpent: 45
})
# Install the SDK
yarn add @prepst/sdk

# Initialize in your project
import { PrepSt } from '@prepst/sdk'

const engine = new PrepSt({
  apiKey: process.env.PREPST_API_KEY
})

# Start tracking learning
const result = await engine.track({
  userId: 'user_123',
  conceptId: 'algebra_basics',
  correct: true,
  timeSpent: 45
})
# Install the SDK
pnpm add @prepst/sdk

# Initialize in your project
import { PrepSt } from '@prepst/sdk'

const engine = new PrepSt({
  apiKey: process.env.PREPST_API_KEY
})

# Start tracking learning
const result = await engine.track({
  userId: 'user_123',
  conceptId: 'algebra_basics',
  correct: true,
  timeSpent: 45
})

Authentication

Secure your API requests with API keys and environment variables

API Keys

Generate API keys from your dashboard. Keep them secure and never commit them to version control.

# .env.local
PREPST_API_KEY=ps_live_abc123...
PREPST_SECRET_KEY=ps_secret_xyz789...

Server-Side Only

Always use API keys on the server side. Never expose them in client-side code or public repositories.

// server.ts
import { PrepSt } from '@prepst/sdk'

const engine = new PrepSt({
  apiKey: process.env.PREPST_API_KEY
})

Code Examples

Real-world examples to help you integrate cognitive learning into your application

import { PrepSt } from '@prepst/sdk'

const engine = new PrepSt({
  apiKey: process.env.PREPST_API_KEY
})

// Track a learning event
const result = await engine.track({
  userId: 'student_123',
  conceptId: 'quadratic_equations',
  questionId: 'q_456',
  correct: true,
  timeSpent: 120, // seconds
  difficulty: 0.65,
  metadata: {
    source: 'practice_quiz',
    attemptNumber: 2
  }
})
import { PrepSt } from '@prepst/sdk'

const engine = new PrepSt({
  apiKey: process.env.PREPST_API_KEY
})

// Get mastery predictions for a student
const predictions = await engine.predict({
  userId: 'student_123',
  concepts: ['algebra_basics', 'geometry', 'trigonometry']
})
import { PrepSt } from '@prepst/sdk'

const engine = new PrepSt({
  apiKey: process.env.PREPST_API_KEY
})

// Get adaptive question recommendations
const questions = await engine.getAdaptiveQuestions({
  userId: 'student_123',
  conceptId: 'linear_equations',
  count: 5,
  targetDifficulty: 'optimal', // 'easy', 'medium', 'hard', 'optimal'
  cognitiveLoadLimit: 0.7
})
import { PrepSt } from '@prepst/sdk'

const engine = new PrepSt({
  apiKey: process.env.PREPST_API_KEY
})

// Get detailed analytics for a student
const analytics = await engine.getAnalytics({
  userId: 'student_123',
  timeRange: 'last_30_days',
  includeInsights: true
})

How It Works

Our four-phase approach to cognitive learning combines cutting-edge AI with proven educational science

Phase 1

Cognitive Modeling

We build a probabilistic model of each learner's knowledge state using Bayesian Knowledge Tracing (BKT). This goes beyond simple right/wrong tracking to understand the underlying cognitive patterns.

  • Tracks mastery probability for each micro-concept
  • Models learning rate and retention patterns
  • Identifies cognitive strengths and weaknesses
Phase 2

Adaptive Question Selection

Our engine selects the optimal next question based on your current knowledge state, learning velocity, and cognitive load. Every interaction is personalized.

  • Item Response Theory (IRT) for difficulty calibration
  • Cognitive load monitoring prevents overwhelm
  • Spaced repetition for long-term retention
Phase 3

Real-Time Interventions

The system detects learning plateaus, misconceptions, and optimal challenge points in real-time, triggering targeted interventions when needed.

  • Plateau detection algorithms
  • Misconception identification
  • Dynamic difficulty adjustment
Phase 4

Continuous Learning

Every interaction feeds back into the model, creating a continuously improving understanding of how you learn best. The engine gets smarter with every question.

  • Pattern recognition across learners
  • Predictive analytics for future performance
  • Personalized learning path optimization

Core Concepts

Key terminology and principles that power the cognitive learning engine

Bayesian Knowledge Tracing (BKT)

A probabilistic model that estimates a learner's mastery of a skill based on their performance history. Unlike simple scoring, BKT accounts for guessing, slipping, and learning rate.

Example:

If a student answers correctly, BKT updates the probability they've mastered the concept, considering they might have guessed.

Micro-Concept Mastery

Breaking down subjects into the smallest possible learning units. Instead of 'Algebra', we track mastery of 'solving linear equations with one variable' separately from 'factoring quadratics'.

Example:

SAT Math is decomposed into 200+ micro-concepts, each tracked independently.

Item Response Theory (IRT)

A framework for measuring question difficulty and learner ability on the same scale. Helps select questions at the optimal difficulty level for each learner.

Example:

A question with difficulty 0.7 is ideal for a learner with ability 0.7, maximizing information gain.

Cognitive Load

The mental effort required to process information. Our engine monitors cognitive load to prevent overwhelm and optimize learning efficiency.

Example:

After three difficult questions, the system may present an easier review question to reduce cognitive strain.

Spaced Repetition

A learning technique that increases intervals between reviews of material. Our engine automatically schedules reviews at optimal times for long-term retention.

Example:

A mastered concept is reviewed after 1 day, then 3 days, then 7 days, then 14 days.

Learning Velocity

The rate at which a learner acquires new knowledge. Our engine tracks velocity to predict future performance and adjust pacing.

Example:

A learner with high velocity in geometry but low velocity in algebra receives more geometry content.

Implementation Roadmap

Our phased approach to building the world's first cognitive learning engine

Phase 1: Foundation

Live in SAT Prep
  • Bayesian Knowledge Tracing implementation
  • Micro-concept decomposition (200+ concepts)
  • Basic adaptive question selection
  • Real-time mastery tracking

Phase 2: Intelligence

In Development
  • Advanced IRT-based difficulty calibration
  • Cognitive load monitoring
  • Plateau detection algorithms
  • Personalized intervention triggers

Phase 3: Optimization

Planned
  • Multi-armed bandit for content selection
  • Spaced repetition scheduling
  • Learning velocity tracking
  • Predictive performance analytics

Phase 4: NomaDB

Research
  • Cross-learner pattern recognition
  • Universal cognitive profile database
  • Transfer learning across subjects
  • Global learning optimization

Ready to integrate?

Explore our SDK documentation to start building with the cognitive learning engine API

Best Practices

Guidelines for maximizing the effectiveness of the cognitive learning engine

Data Collection

Do

  • Track every learner interaction with timestamps
  • Record response time and confidence levels
  • Capture partial credit and work shown
  • Log context like time of day and session length

Don't

  • Don't aggregate data before sending to the engine
  • Don't filter out 'bad' sessions or outliers
  • Don't delay data transmission
  • Don't modify learner responses

Question Selection

Do

  • Trust the engine's recommendations
  • Provide diverse question types for each concept
  • Include questions at multiple difficulty levels
  • Tag questions with detailed metadata

Don't

  • Don't override the engine with manual selection
  • Don't limit question pool artificially
  • Don't reuse questions too frequently
  • Don't ignore cognitive load warnings

Intervention Design

Do

  • Respond immediately to plateau signals
  • Provide multiple intervention types
  • Allow learners to opt-out of interventions
  • Track intervention effectiveness

Don't

  • Don't interrupt during active problem-solving
  • Don't force interventions on resistant learners
  • Don't use generic hints or explanations
  • Don't ignore learner feedback on interventions

Frequently Asked Questions

Common questions about the cognitive learning engine

API Reference

Complete reference for all available endpoints and methods

POST/v2/track

Track a learning event and update knowledge state

userIdconceptIdcorrecttimeSpentdifficulty
GET/v2/predict

Get mastery predictions for concepts

userIdconcepts[]
GET/v2/questions/adaptive

Get adaptive question recommendations

userIdconceptIdcounttargetDifficulty
GET/v2/analytics

Retrieve learning analytics and insights

userIdtimeRangeincludeInsights
POST/v2/concepts

Create or update concept definitions

nameprerequisites[]difficultymetadata
GET/v2/knowledge-graph

Get the knowledge graph for a user

userIddepth

Base URL: https://api.cognitive-engine.com

Use Cases

See how leading organizations use our SDK to transform learning experiences

EdTech Platforms

Build adaptive learning experiences for K-12, higher education, and test prep platforms

  • Personalized learning paths
  • Real-time mastery tracking
  • Predictive analytics

Corporate Training

Create intelligent training programs that adapt to employee skill levels and learning pace

  • Skill gap analysis
  • Compliance tracking
  • ROI measurement

Online Courses

Enhance MOOCs and online courses with cognitive intelligence and adaptive content

  • Dynamic difficulty adjustment
  • Engagement optimization
  • Completion prediction

Tutoring Services

Power tutoring platforms with AI-driven insights and personalized recommendations

  • Student profiling
  • Intervention triggers
  • Progress reporting