System Overview

Organization: Pedagogical Innovation Center (CIP), Polytechnic of Porto (IPP)
Status: Active Development

Executive Summary

This documentation covers an AI-enhanced Moodle learning platform that integrates Large Language Model (LLM) agents with vector database technology to deliver personalized learning experiences. The system consists of four primary workflows:

Data Ingestion - Automatic processing of course materials
Content Summarization - RAG-powered summaries for students
Doubt Clarification - AI-assisted question answering
Agentic Quiz Generation - Adaptive assessments with feedback

Platform Capabilities

The platform enables:

✅ Automatic ingestion of diverse course materials (PDFs, PPTs, ePub, text documents)
✅ Real-time retrieval-augmented generation (RAG) for personalized summaries
✅ Intelligent doubt clarification based on course context
✅ Adaptive quiz generation with formative feedback

System Architecture

The platform operates as a distributed system with three main components:

Component	Description
Moodle LMS	Learning Management System hosting course content and student data
LangFlow Orchestration Layer	Workflow automation and LLM agent orchestration
Vector Database (Qdrant)	Semantic search and retrieval of ingested materials

Architectural Diagram

┌─────────────────────────────────────────────────────────────┐
│                      Moodle LMS                              │
│                  (Course Management)                         │
│  ┌────────────────────────────────────────────────────────┐  │
│  │ Courses │ Resources │ Activities │ Student Progress    │  │
│  │ Files   │ Metadata  │ Assessment │ Interactions        │  │
│  └────────────────────────────────────────────────────────┘  │
└─────────────────────────────────────────────────────────────┘
              │                     │
              │ Webhooks            │ API Queries
              ▼                     ▼
┌─────────────────────────────────────────────────────────────┐
│                    LangFlow Workflows                        │
│  ┌────────────────────────────────────────────────────────┐  │
│  │ • Data Ingestion Pipeline                              │  │
│  │ • Content Summarization Agent                          │  │
│  │ • Doubt Clarification Agent                            │  │
│  │ • Quiz Generation Agent (with feedback loop)           │  │
│  └────────────────────────────────────────────────────────┘  │
└─────────────────────────────────────────────────────────────┘
              │                     │
              │ Embeddings          │ Vector Queries
              ▼                     ▼
┌─────────────────────────────────────────────────────────────┐
│                   Qdrant Vector DB                           │
│  ┌────────────────────────────────────────────────────────┐  │
│  │ Collections:                                            │  │
│  │ • course_materials (vectors + metadata)                │  │
│  │ • course_metadata (course_id, section_id, file_type)  │  │
│  └────────────────────────────────────────────────────────┘  │
└─────────────────────────────────────────────────────────────┘

Data Flow Principles

Course-scoped filtering: All Qdrant queries are filtered by course_id to ensure student access only to their own course materials
Metadata preservation: File origin, section, module, and mimetype are retained for context
Agentic orchestration: LLM agents control workflow logic, tool invocation, and decision-making
Fallback strategies: When insufficient materials exist, agents provide educational guidance

Next Steps

Ready to explore the platform? Start with:

Core Technologies - Learn about the technology stack
Data Ingestion - Understand how content is processed
Moodle Integration - See how it connects to Moodle

System Overview ​

Executive Summary ​

Platform Capabilities ​

System Architecture ​

Architectural Diagram ​

Data Flow Principles ​

Next Steps ​