Generate vector embeddings for input data using Pinecone's hosted embedding models.
Arguments
- model
The embedding model to use. Available models include: - "multilingual-e5-large" (1024 dimensions) - "pinecone-sparse-english-v0" (sparse vectors) See Pinecone documentation for the full list of available models.
- inputs
A character vector of texts to embed.
- input_type
The type of input. Either "passage" for documents to be indexed, or "query" for search queries. Default is "passage".
- truncate
How to handle inputs longer than the model's max token length. Options: "END" (truncate from end), "NONE" (error if too long). Default is "END".
- tidy
Whether to return a tidy tibble format (default: TRUE). If FALSE, returns the raw API response structure.
Value
List with http response, content (embeddings), and status_code. When tidy = TRUE, content is a tibble with columns: - values: list column containing the embedding vectors
Examples
if (FALSE) { # \dontrun{
# Embed documents for indexing
result <- embed(
model = "multilingual-e5-large",
inputs = c("The quick brown fox", "jumps over the lazy dog"),
input_type = "passage"
)
# Embed a query for searching
query_embedding <- embed(
model = "multilingual-e5-large",
inputs = "What does the fox do?",
input_type = "query"
)
# Use the embedding for a vector query
vector_query("my-index", vector = query_embedding$content$values[[1]], top_k = 10)
} # }