Educational Reference Hub

The practical portal forRetrieval-Augmented Generation

Learn what RAG is, compare tools, design architectures, evaluate quality, and build grounded AI systems that users can trust.

RAG connects language models with external knowledge sources to improve grounding, accuracy, traceability, and domain relevance.

New to RAG Go deep Build efficiently

Tool profiles

Structured entries with verification notes and official links.

Architecture patterns

From basic RAG to enterprise observability and Graph RAG.

Comparison pages

Criteria-based, cautious, vendor-neutral guidance.

Audience pathways

Entry points for beginners, engineers, educators, and decision-makers.

// pipeline overview

Ten stages, one grounded answer

Every verifiable answer travels through these stages, from raw knowledge to cited, evaluatable output.

Knowledge

Documents

Prepare

Ingestion

Prepare

Chunking

Index

Embeddings

Index

Vector DB

Find

Retrieval

Refine

Reranking

Assemble

Context

Answer

Generation

Trust

Citations

Flow logic

RAG turns raw knowledge into searchable evidence, selects the most relevant context, then generates an answer that users can verify through citations and evaluation.

Ground answers in sources

Refresh knowledge without retraining

Expose citations and evidence

Measure quality over time

Portal finder

Find the right RAG topic fast

What is RAG?

concept

Definition, purpose, and the grounding role of Retrieval-Augmented Generation.

RAG vs fine-tuning

concept

When to retrieve knowledge and when to adapt model behavior.

Tools directory

Dify guide

guide

How Dify fits into RAG application building, workflows, knowledge bases, and agents.

Architectures

reference

Basic, advanced, hybrid, metadata-aware, graph, agentic, multimodal, local, and enterprise RAG patterns.

Implementation playbook

playbook

A practical lifecycle for planning, building, testing, launching, and maintaining RAG systems.

Evaluation center

playbook

How to evaluate retrieval, answer faithfulness, citations, latency, safety, and business outcomes.

Glossary

reference

Searchable definitions for embeddings, vector databases, reranking, grounding, Graph RAG, and more.

Choose your pathway

Different users need different levels of depth. Start with the pathway closest to your role, then move into architectures, tools, and evaluation.

Beginner

Understand RAG without getting buried in infrastructure.

Start 1 Start 2 Start 3

Needs

Plain definitions
Visual pipeline
Common mistakes
Glossary

Best sections

Learn RAG
Glossary
Tutorials

Explore the portal

Use the hub as a map: learn the concepts, choose a stack, build a pipeline, evaluate quality, and keep knowledge current.

Learn RAG

Open the learn rag section.

RAG Tools

Open the rag tools section.

Dify Guide

Open the dify guide section.

Architectures

Open the architectures section.

Tutorials

Open the tutorials section.

Use Cases

Open the use cases section.

Glossary

Open the glossary section.

Comparisons

Open the comparisons section.

Resources

Open the resources section.

Implementation

Open the implementation section.

Evaluation

Open the evaluation section.

The RAG stack, end to end

Effective RAG is not a single tool. It is a chain of editorial, retrieval, generation, evaluation, and operational decisions.

Layer 1

Knowledge sources

Authoritative documents, records, webpages, databases, tickets, policies, and archives.

Ownershipfreshnesspermissionsformat qualitysource authority

Layer 2

Ingestion and parsing

Convert files into structured text, tables, page references, images, and metadata.

OCR qualitylayout handlingdeduplicationlanguage detectionupdate cadence

Layer 3

Chunking and metadata

Create retrievable units that preserve meaning, context, and source traceability.

Chunk sizeoverlapsection boundariespage numbersaccess labels

Layer 4

Embeddings and indexing

Represent chunks for semantic retrieval and store them with searchable metadata.

Embedding modelvector databasehybrid indexingnamespace strategyre-indexing

Layer 5

Retrieval and reranking

Find relevant evidence, filter it, and order the best context for generation.

Top-khybrid weightingmetadata filtersrerankerquery rewriting

Layer 6

Generation and citations

Assemble prompts, instruct the model, generate answers, and cite source passages.

Prompt policysource displayunsupported-answer behaviormodel choicetone

Layer 7

Evaluation and operations

Measure quality, detect regressions, monitor traces, and improve the system over time.

Golden questionsfaithfulness checkshuman reviewtelemetryfeedback loops

Why RAG matters

RAG is one of the most practical patterns for building AI systems that need current, private, or specialized knowledge.

Better grounding

Private and domain knowledge

Reduced hallucination

Updatable knowledge

Explainability through sources

Risks every RAG project must control

A serious RAG portal should teach the failure modes, not only the happy path.

Permission leakage

Apply document-level access controls before retrieval, not only after generation.

Stale knowledge

Track last-updated metadata and create refresh workflows for high-authority sources.

Prompt injection

Treat retrieved text as untrusted content and keep system instructions separate.

Unsupported answers

Require citation-backed claims and define a no-answer policy.

Bad OCR or parsing

Sample parsed output visually and preserve source page references.

Overconfident automation

Use human review for legal, medical, financial, policy, or high-impact decisions.

Who is this site for?

The hub is written for both beginners and experts, with careful explanations, structured comparisons, and practical architecture guidance.

Developers

Build reliable retrieval systems and choose maintainable toolchains.

Researchers

Compare RAG patterns, evaluation methods, and retrieval tradeoffs.

Educators

Teach grounding, sources, and practical AI literacy.

Librarians and documentalists

Connect metadata, archives, and documentary authority to AI interfaces.

Businesses

Assess platforms, governance, and production readiness.

Public institutions

Design traceable knowledge services with source accountability.