Inkline
Menu

RAG request

Templates
RAG request 6 items and 10 connections. question retrieve(question) nearest 50 candidates rerank(candidates) top 5 context prompt + context answer + citations answer Client API Retriever Vector DB Reranker LLM
RAG request — preview
Sequence

The calls behind one RAG answer: retrieve candidates, rerank to the top five, then prompt the model with that context.

What is inside

  • Retrieve, rerank, generate
  • Citations come back with the answer
Use this template

Opens in the playground, no account needed. Sign up to save it.

Show the Inkline text
diagram sequence "RAG request"
  actor       client "Client"
  participant api    "API"
  participant ret    "Retriever"
  participant vdb    "Vector DB"
  participant rr     "Reranker"
  participant model  "LLM"

  client ->  api   : question
  api    ->  ret   : retrieve(question)
  ret    ->  vdb   : nearest 50
  vdb    --> ret   : candidates
  ret    ->  rr    : rerank(candidates)
  rr     --> ret   : top 5
  ret    --> api   : context
  api    ->  model : prompt + context
  model  --> api   : answer + citations
  api    --> client : answer