Skip to contents

Generate embeddings for multiple texts using OpenAI

Source: R/openai_embed.R
oai_embed_batch.Rd

High-level function to generate embeddings for multiple text strings using OpenAI's embedding API. This function handles batching, concurrent requests, error handling, and provides progress reporting for large collections of texts.

Usage

oai_embed_batch(
  texts,
  model = "text-embedding-3-small",
  dimensions = 1536,
  batch_size = 10,
  concurrent_requests = 1,
  max_retries = 5,
  timeout = 20,
  endpoint_url = "https://api.openai.com/v1/embeddings",
  key_name = "OPENAI_API_KEY",
  include_texts = TRUE,
  relocate_col = 2,
  verbose = FALSE
)

Arguments

texts

Vector or list of character strings to generate embeddings for

model

OpenAI embedding model to use (default: "text-embedding-3-small")

dimensions

Number of embedding dimensions (default: 1536 for text-embedding-3-small)

batch_size

Number of texts to process in one API request (default: 10)

concurrent_requests

Number of requests to send simultaneously (default: 1)

max_retries

Maximum retry attempts for failed requests (default: 5)

timeout

Request timeout in seconds (default: 20)

endpoint_url

OpenAI API endpoint URL (default: OpenAI's embedding endpoint)

key_name

Name of environment variable containing the API key (default: "OPENAI_API_KEY")

include_texts

Whether to include original texts in the result (default: TRUE)

relocate_col

Column position to place error columns (default: 2)

verbose

Whether to enable verbose request logging (default: FALSE)

Value

A tibble containing:

  • Embedding vectors as columns (V1, V2, ..., Vn)

  • .error: Logical column indicating if embedding failed

  • .error_message: Character column with error details

  • text: Original texts (if include_texts = TRUE)

Details

This function efficiently processes multiple texts by:

  1. Splitting texts into batches of the specified size

  2. Creating concurrent requests (if configured) for faster processing

  3. Handling individual batch failures gracefully

  4. Pre-allocating memory for embeddings to improve performance

  5. Providing detailed success/failure reporting

If a batch fails, only the documents in that specific batch will be marked as failed, not all documents across all batches. Failed embeddings will be filled with NA values and marked with error information.

The function returns a tibble with embedding columns (V1, V2, ..., Vn), error tracking columns (.error, .error_message), and optionally the original texts.

Examples

if (FALSE) { # \dontrun{
  # Basic batch embedding
  texts <- c("First text", "Second text", "Third text")
  embeddings <- oai_embed_batch(texts)
  
  # Large-scale processing with concurrent requests
  large_texts <- rep("Sample text", 100)
  embeddings <- oai_embed_batch(
    texts = large_texts,
    batch_size = 20,
    concurrent_requests = 5,
    dimensions = 512
  )
  
  # Custom model and settings
  embeddings <- oai_embed_batch(
    texts = texts,
    model = "text-embedding-3-large",
    dimensions = 1024,
    include_texts = FALSE,
    timeout = 30
  )
} # }