RAG
Retrieval-Augmented Generation: ground a model answer in documents fetched at query time.
Overview
Retrieval-Augmented Generation answers a question by first retrieving relevant passages from a knowledge source and then having a language model compose the answer from them. It solves two problems at once: a model cannot know facts outside its training data, and fine-tuning it on new facts is expensive and must be repeated whenever the facts change. It exists because most factual answers are cheaper to look up than to memorise. The cost is a new dependency chain - embeddings, a vector store, reranking and a generation step - which is why it is the most interconnected node in this graph.
Technology Graph
How RAG connects to the rest of the stack.
Uses and depends on
- Embedding concept Techniques RAG uses Embedding High ✓ source
- LLM concept Models RAG uses LLM High ✓ source
- Reranking technique Techniques RAG uses Reranking Medium ✓ source
- GraphRAG technique Techniques GraphRAG based on RAG High ✓ source
- Grounding technique Techniques Grounding based on RAG High ✓ source
- Self-RAG technique Techniques Self-RAG based on RAG High ✓ source
Used by
Applied to
- AI Search application Applications RAG used for AI Search High ✓ source
- Agent Memory technique AI Agents Agent Memory used for RAG High ✓ source
- BGE model Embedding Models BGE used for RAG High ✓ source
- Chatbot application Applications RAG used for Chatbot High ✓ source
- Chroma database Vector Databases Chroma used for RAG High ✓ source
- Chunking technique Techniques Chunking used for RAG High ✓ source
- Cohere Embed model Embedding Models Cohere Embed used for RAG High ✓ source
- Command R model Models Command R used for RAG High ✓ source
- Context Engineering technique Prompting Context Engineering used for RAG High ✓ source
- Context Window concept Models Context Window used for RAG High ✓ source
- DeepEval library Evaluation DeepEval used for RAG High ✓ source
- GTE model Embedding Models GTE used for RAG High ✓ source
- Haystack framework LLM Frameworks Haystack used for RAG High ✓ source
- HyDE technique Techniques HyDE used for RAG High ✓ source
- Instructor model Embedding Models Instructor used for RAG Medium ✓ source
- Jina Embeddings model Embedding Models Jina Embeddings used for RAG High ✓ source
- LangChain framework LLM Frameworks LangChain used for RAG High ✓ source
- LlamaIndex framework LLM Frameworks LlamaIndex used for RAG High ✓ source
- Nomic Embed model Embedding Models Nomic Embed used for RAG High ✓ source
- OpenAI Text Embedding 3 model Embedding Models OpenAI Text Embedding 3 used for RAG High ✓ source
- PostgreSQL database Databases PostgreSQL used for RAG High ✓ source
- Prompt Caching technique Infrastructure Prompt Caching used for RAG High ✓ source
- Qdrant database Vector Databases Qdrant used for RAG High ✓ source
- Query Rewriting technique Techniques Query Rewriting used for RAG High ✓ source
- Ragas tool Evaluation Ragas used for RAG High ✓ source
- SQuAD dataset Datasets SQuAD used for RAG High ✓ source
- TruLens library Evaluation TruLens used for RAG High ✓ source
- Typesense database Search Engines Typesense used for RAG High ✓ source
- Hallucination concept Models RAG solves Hallucination Medium ✓ source
Tools
Software that implements or integrates with RAG.
LLM Frameworks
Alternatives
-
Fine-tuning
technique
An alternative to RAG · source
Comparisons
Side-by-side breakdowns with the alternatives above.