Creates a new Pinecone index with an integrated embedding model. This enables automatic embedding of text during upsert and query operations using Pinecone's hosted models.
Usage
create_index_for_model(
name,
cloud,
region,
embed,
deletion_protection = "disabled",
tags = NULL
)Arguments
- name
Name of the index (must be unique within project)
- cloud
Cloud provider: "aws", "gcp", or "azure"
- region
Cloud region (e.g., "us-east-1" for AWS)
- embed
A list specifying the embedding model configuration: - model: The embedding model name (e.g., "multilingual-e5-large") - field_map: A named list mapping source field to text field (e.g., list(text = "chunk_text")) - metric: Optional distance metric ("cosine", "euclidean", "dotproduct") - read_parameters: Optional list with input_type for queries - write_parameters: Optional list with input_type for documents
- deletion_protection
Whether to enable deletion protection ("enabled" or "disabled")
Optional named list of tags for the index
Details
Indexes created with integrated embedding models support: - `records_upsert()`: Upsert text that gets automatically embedded - `records_search()`: Search using text queries with optional reranking
The embedding model determines the vector dimension automatically.
Examples
if (FALSE) { # \dontrun{
# Create an index with multilingual-e5-large model
create_index_for_model(
name = "my-semantic-index",
cloud = "aws",
region = "us-east-1",
embed = list(
model = "multilingual-e5-large",
field_map = list(text = "chunk_text")
)
)
# With custom metric and input types
create_index_for_model(
name = "my-index",
cloud = "aws",
region = "us-east-1",
embed = list(
model = "multilingual-e5-large",
field_map = list(text = "content"),
metric = "dotproduct",
read_parameters = list(input_type = "query"),
write_parameters = list(input_type = "passage")
)
)
} # }