EN
RegisterSign in

How RAG works

From course materials to grounded answers

RAG combines generative AI with the knowledge sources an educator chooses, so AI support in a course is controllable, explainable and context-aware.

The workflow

Four steps of a course-specific RAG application

  1. 1Retrieve

    A learner’s question searches the course knowledge the educator has curated — lecture notes, slides, papers, exams — and returns the most relevant passages.

  2. 2Augment

    The retrieved passages and the learner’s profile are added to the prompt, so institutional knowledge shapes a relevant, context-aware response.

  3. 3Generate

    A European LLM writes the answer from that context and shows the sources it used, so learners can check every claim.

  4. 4For education

    Educators design and validate each step in the RAG4Edu Toolbox — with GDPR and AI Act checks, and without advanced AI skills.

Course knowledge slides · lecture notes papers · exams · datasets curated by the educator Ingest & index embeddings · metadata Learner question or group task Retrieve top-k relevant passages Augment prompt context + learner profile Generate European LLM Answer with cited sources Educator designs · validates GDPR · AI Act check no / low-code RAG4Edu pipeline: modular, documented, EU-hosted models, learner-aware augmentation. Educators build and configure it through the RAG4Edu Toolbox — without advanced AI skills.
Inside a course-specific RAG application: the educator curates the knowledge base, retrieval grounds every answer in it, and generation runs on European LLMs.

Why it can be trusted in a course

  • Grounded in sourcesAnswers are built on the knowledge base the educator curated for the course.
  • Citations providedLearners see which passages an answer draws on and can check them.
  • European and compliantEU-based LLMs and EuroHPC infrastructure, GDPR-compliant design, aligned with the EU AI Act.
  • Open and reusablePipeline, toolbox and templates are published under open licences.

Tools

What RAG4Edu builds

The project develops open components that educators and institutions can reuse.

RAG4Edu Pipeline and FrameworkConceptual and technical specification, a deployable operational prototype (ingestion, semantic retrieval, generation) and learner-aware augmentation mechanisms.
RAG4Edu Chatbots and ToolboxAuthoring tools to create and configure course-specific RAG chatbots and agents, multilingual case-study applications, templates and documentation.
Group Formation Design Framework and toolCriteria, data sources, decision logic and transparency mechanisms for AI-supported group formation, with collaborative learning design templates.
Integration and support servicesSpecification and integration of EU-based LLMs and AI services, embeddings, token management and compliance features for the pilots.
Learn it in the training Results & resources