ragnar_store_atlas: Visualize a store using Embedding Atlas

In ragnar: Retrieval-Augmented Generation (RAG) Workflows

Visualize a store using Embedding Atlas

Description

Visualize a store using Embedding Atlas

Usage

ragnar_store_atlas(
  store,
  ...,
  host = "localhost",
  port = 3030,
  launch.browser = interactive()
)

Arguments

store	A RagnarStore object to inspect.
...	Passed to shiny::runApp().
host	Host to run the Embedding Atlas server on.
port	Port to run the Embedding Atlas server on.
launch.browser	Whether to launch the browser automatically.

Note

This function requires the embedding-atlas Python package (>= 0.20.0) in your reticulate Python environment, the duckdb R package (>= 1.4.0), and nanoarrow (>= 0.8.0) to transfer data from the DuckDB store to Python.

Examples

## Not run: 
# Start Ollama, then run this in a terminal:
# ollama pull embeddinggemma:300m-qat-q4_0

# Create an in-memory store using local embeddings
store <- ragnar_store_create(
  embed = embed_ollama(model = "embeddinggemma:300m-qat-q4_0")
)

# Read and embed a chapter from R for Data Science
chunks <- "https://r4ds.hadley.nz/data-transform.html" |>
  read_as_markdown() |>
  markdown_chunk()
ragnar_store_insert(store, chunks)

# Launch the Embedding Atlas app
# Interrupt R (Esc or Ctrl+C) to stop the server
ragnar_store_atlas(store)

## End(Not run)

ragnar documentation built on Sept. 11, 2026, 1:06 a.m.