README.md

metadata

title: AIDidact
emoji: 🎓
colorFrom: blue
colorTo: indigo
sdk: docker
app_port: 7860
pinned: false

AIDidact

AIDidact is a scalable AI-powered microlearning ecosystem for personalized, self-directed learning. The platform is organized around modular 5 ECTS learning units, hybrid recommendations, cheating-resistant assessments, and analytics-driven learner support.

Included Deliverables

• [Architecture blueprint](C:\Users\user\Desktop\AI Flow\AIDidact\docs\architecture.md)
• [Core algorithms](C:\Users\user\Desktop\AI Flow\AIDidact\docs\algorithms.md)
• [Relational data schema](C:\Users\user\Desktop\AI Flow\AIDidact\schemas\schema.sql)
• [REST API specification](C:\Users\user\Desktop\AI Flow\AIDidact\api\openapi.yaml)
• [FastAPI backend scaffold](C:\Users\user\Desktop\AI Flow\AIDidact\backend\README.md)

System Overview

AIDidact supports millions of learners across web and mobile channels. At registration, learners provide prior learning, goals, educational level, readiness signals, and optional learning preferences. That information is transformed into a learner profile vector used to personalize module recommendations, pacing, and feedback.

Each microlearning module is worth 5 ECTS credits and includes Bloom-aligned objectives, chunked 10 to 15 minute content units, interactive activities, formative assessments, and explicit completion criteria. Modules are self-paced, measurable, and adaptive where possible.

Textual Architecture Diagram

flowchart LR
    A["Web App"] --> G["API Gateway"]
    B["Mobile App"] --> G
    C["Admin Console"] --> G

    G --> U["Auth and Consent Service"]
    G --> P["Profile Service"]
    G --> M["Module Service"]
    G --> R["Recommendation Service"]
    G --> S["Assessment Service"]
    G --> T["Progress Service"]
    G --> D["Dashboard Service"]

    U --> PG["PostgreSQL"]
    P --> PG
    M --> PG
    S --> PG
    T --> PG

    P --> K["Event Bus"]
    M --> K
    S --> K
    T --> K

    K --> L["Analytics Processor"]
    L --> NS["NoSQL Log Store"]
    L --> WH["Analytics Warehouse"]
    L --> FS["Feature Store"]

    R --> FS
    R --> VX["Vector Index"]
    R --> RM["Hybrid Recommender Models"]

    S --> Q["Question Generation"]
    S --> E["AI Evaluation"]
    S --> I["Integrity Engine"]

    D --> WH

Key Design Decisions

• Transactional data lives in PostgreSQL; high-volume telemetry flows to NoSQL and warehouse systems.
• Recommendation uses hybrid content-based plus collaborative filtering and refreshes from learner progress.
• Assessment uses blueprint-based randomized questions, AI-assisted open-ended grading, and anomaly detection.
• GDPR compliance is built in through consent management, auditability, pseudonymization, retention rules, and human review paths.

Hugging Face Spaces

This repository is now prepared for Hugging Face Spaces using Docker.

Recommended Space setup:

• SDK: Docker
• Template: Blank
• Hardware: CPU Basic

For the easiest deployment path on Spaces:

• use the included Dockerfile
• keep DATABASE_URL as SQLite for demo use
• optionally switch to an external PostgreSQL instance later via Space Secrets

Main deployment files:

• [Dockerfile](C:\Users\user\Desktop\AI Flow\AIDidact\Dockerfile)
• [backend README](C:\Users\user\Desktop\AI Flow\AIDidact\backend\README.md)
• [environment example](C:\Users\user\Desktop\AI Flow\AIDidact\backend.env.example)

Example User Flow

1. Learner registers and grants required consents.
2. Diagnostic assessment establishes readiness and baseline mastery.
3. Profile vector is generated from prior learning, goals, level, readiness, and preferences.
4. Recommendation service ranks starter modules.
5. Learner completes chunked units, activities, and formative checks.
6. Events stream to analytics and update mastery, dashboards, and next recommendations.
7. Learner completes a unique summative assessment with integrity controls.
8. Module is marked complete only if score, engagement, and integrity thresholds are satisfied.