Skip to contents

RAG with Pinecone Backend

Source: vignettes/pinecone-backend.Rmd
pinecone-backend.Rmd

This vignette shows how to query an existing Pinecone index/namespace from RAGFlowChainR.

1) Load packages

2) Configure Pinecone target 

# Optional if your endpoint requires API key auth
Sys.setenv(PINECONE_API_KEY = "your-pinecone-api-key")

# Format: "https://index-host|namespace"
# Namespace can be left empty: "https://index-host|"
pinecone_target <- "https://your-index-host|my_namespace"

3) Create the RAG chain 

my_llm <- function(prompt) {
  "Mock answer from LLM."
}

rag_chain <- create_rag_chain(
  llm = my_llm,
  vector_database_directory = pinecone_target,
  method = "Pinecone",
  embedding_function = embed_openai(model = "text-embedding-3-small"),
  embedding_dim = 1536,
  use_web_search = FALSE
)

4) Query 

result <- rag_chain$invoke("What does this Pinecone namespace contain?")
result$answer
result$documents

Notes

  • Pinecone indexing/upserts happen outside RAGFlowChainR.
  • Match embedding_function and embedding_dim to the vectors stored in Pinecone.