The Pedagogical Operating System
A zero-trust platform for building, reviewing, distributing, installing, and running research-grade pedagogical apps with validated runtime code, user-scoped storage, settings, lifecycle hooks, and gated LearnAdapt telemetry.
App Runtime v2 inside PedOS 1.1
PedOS 1.1 Lumina is the platform release for the SDK, review pipeline, Plugins Directory, My Apps installation, and analytics. App Runtime v2 is the compatibility layer inside that release. Installed plugins run as PedOS apps, declare a manifest, permissions, lifecycle hooks, settings, storage needs, host surfaces, and evidence profile, then launch through the guarded runtime SDK only after installation.
App Shell
Every installed app opens inside a consistent PedOS shell with My Apps, Settings, Analytics, and Plugins Directory navigation.
Settings & Storage
Apps use approved user-scoped settings and lightweight storage through window.PedOSApp, not ad hoc plugin-local endpoints.
Lifecycle Hooks
Install, uninstall, activate, and deactivate events are recorded and dispatched through LearnAdapt hooks for auditable behavior.
Runtime Permissions
Context, settings, storage, telemetry, and host-surface access are manifest-declared and mediated by the platform.
Host Integration
Apps can be organized in My Apps, launched from the dashboard sidebar, configured from app settings, and measured in app analytics.
Telemetry Gate
Approved installed apps emit profile-appropriate xAPI through the PedOS telemetry gate; preview and uninstalled apps cannot persist learning data.
From Prompt to Production
Learn how the PedOS plugin pipeline works end-to-end. Agentic Studio now begins with a deterministic preflight check that maps the brief to the current SDK contract, flags future-version KIV requests, estimates account credits, and only then hands the accepted brief to the agent development team.
Why an Operating System?
Legacy educational software operates as a closed loop. Researchers who want to test a novel scaffolding idea — for example, testing how dyslexic students respond to specialized model variants — typically have to build an entirely new learning platform from scratch.
PedOS changes this paradigm. It lets researchers build "Cognitive Apps" — discrete modules with custom UI, telemetry, and LLM logic — that plug into a shared platform with built-in BKT, xAPI, and ethical guardrails.
Dual-Layer Architecture
-
PHP UI Hooks (Layer 1) Native LearnAdapt hooks and guarded plugin routes. Plugins can render approved UI and emit telemetry only through the installed-plugin gate.
-
Python Extension Registry (Layer 2) Mount approved structured models into the Multi-Agent Subsumption flow of the platform core.
Zero-Trust Subsumption Guardrails
A hallucinating AI plugin could commit pedagogical malpractice. PedOS mitigates this with strict Tiered Sandboxing — plugins cannot bypass ethics or corrupt mastery data.
Ethics & Safety — LOCKED
Core OS only. Evaluates harm, abuse, or cheating attempts. Plugins can never override this tier.
Bayesian Knowledge Tracing — READ-ONLY
Maintains the probabilistic mastery model P(L). Plugins can read this diagnosis but cannot manipulate the official telemetry.
Developer Zone — APPROVED PLUGINS
Custom scaffolding, assessments, visualizations, and LoRA adapters. Runtime access is mediated by the LearnAdapt router, SAST validation, versioned deployment, and telemetry gates.
Platform Capabilities
Plugin SDK
Standardized app package format with manifest.json, runtime permissions, settings schema, user storage, hook system, and profile-appropriate telemetry contracts.
Agentic Studio
AI-powered plugin generator for non-coders. Describe what you want in natural language — a multi-agent pipeline builds, validates, and deploys it.
SAST Validation
Automated security scanning (banned functions, SQL injection, XSS). Minimum score: 60/100 to pass. Enforced before admin review.
xAPI & Telemetry
Approved installed apps emit validated telemetry through the runtime gate. Utility apps stay lightweight; learning-evidence apps provide BKT-ready xAPI fields while core BKT state remains protected by the platform.
BKT Engine
Bayesian Knowledge Tracing with N400 cognitive overload detection via EEG sensor fusion. Real-time mastery probability updates.
RBAC & Auth
5-tier role hierarchy: Student → Educator → Researcher → Admin → System Admin. Route-level guards, CSRF protection, session timeouts.
Role-Based Access Control
Every protected route in PedOS is guarded by centralized auth middleware with a 7-level role hierarchy:
| Role | Level | Permissions |
|---|---|---|
| Student / Parent | 1 | Use installed plugins, submit responses, view own learning surfaces |
| Educator / Mentor | 2 | Manage learning workflows, dashboards, and instructional interventions |
| Institution Admin / Designer | 3 | Coordinate institutional or design-level learning operations |
| Researcher | 4 | Access labs, experiments, analytics, and research workflows |
| Developer | 5 | Build and inspect plugin-oriented development workflows |
| Admin | 6 | Approve or reject plugins, manage users, and review audit trails |
| System Admin | 7 | Configure platform settings, providers, and deployment controls |
Plugin Submission Lifecycle
A governed app-directory model: admin approval is required before any plugin goes live on the platform.
AI Path (Non-Coders)
Describe your plugin in the Agentic Studio. The multi-agent pipeline plans, builds, reviews, validates, and prepares it for admin approval.
Manual Path (Developers)
Follow the SDK specification. Create your plugin package with manifest.json + php/init.php, then submit for admin review.
PedOS 1.1 Lumina
Explore the Ecosystem
Browse community-built Cognitive Apps, read the developer SDK, or generate your own plugin with AI.