            OFFLINE INDEXING
          

            EMBED · INDEX

            RETRIEVE

            
            ANSWER TO USER

            DATA FEEDS

Documents

Databases

Web pages

Integrations

IDX

Knowledge sources

                  Documents & data
                

IDX

Chunking & embed

                  Split → vectors
                

User query

                  Natural-language request
                

Embedding model

                  Query → vector
                

Vector database

                  Top-k similarity search
                

Re-ranking

                  Relevance scoring
                

Context assembly

                  Prompt + retrieved chunks
                

LLM generation

                  Grounded synthesis
                

Guardrails & citation

                  Grounding & sources
                

Response

                  Cited answer
                

OUTPUT // GENERATED USE CASES

01 / 08GENERATING

Uncover affluent customers

SPECTRUM AI · RAG

Request

Indexing

Retrieval

Generation

            PART 1 · INDEXING
          

Retrieval-augmented generation