RAG — Retrieval-Augmented Generation
End-to-end RAG pipeline: ingest → embed → vector store → retrieve → LLM.
Rendering…
Make it your own.
flowchart LR
subgraph Ingestion
Docs[Documents] --> Chunk[Chunker]
Chunk --> Embed[Embedding model]
Embed --> VStore[(Vector store)]
end
subgraph Query
U[User question] --> QEmbed[Embed query]
QEmbed --> Search[Top-K search]
VStore --> Search
Search --> Rerank[Reranker]
Rerank --> Prompt[Prompt template]
U --> Prompt
Prompt --> LLM[LLM]
LLM --> A[Answer + citations]
end
classDef store fill:#dbeafe,stroke:#1e3a8a;
classDef llm fill:#fce7f3,stroke:#9d174d;
class VStore store
class LLM,Embed,QEmbed,Rerank llm