Skip to contents

RAG with Weaviate Backend

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

This vignette shows how to query Weaviate using the Weaviate method in RAGFlowChainR.

1) Load packages

2) Configure Weaviate target 

# Optional if your Weaviate deployment requires token auth
Sys.setenv(WEAVIATE_API_KEY = "your-weaviate-api-key")

# Format: "https://weaviate-host|ClassName"
weaviate_target <- "https://your-weaviate-host|Document"

3) Create the RAG chain 

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

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

4) Query 

result <- rag_chain$invoke("Summarize the most relevant Weaviate documents.")
result$answer
result$documents

Notes

  • Your Weaviate class should include text fields such as page_content, text, or content.
  • Ensure query embeddings use the same vector setup used during ingestion.