Implementation:Cohere ai Cohere python Client Embed

Metadata
Cohere Python SDK
Cohere Embed API
NLP, Embeddings, Vector_Search
2026-02-15 14:00 GMT

Overview

Concrete method for generating text embeddings with automatic batching and concurrent processing.

Description

Client.embed provides embedding generation with automatic batching. When batching is enabled (default True), texts are split into chunks of 96 (embed_batch_size from config.py), processed concurrently using ThreadPoolExecutor.map(), and results are merged via merge_embed_responses() from utils.py. Image embeddings skip batching. The method delegates to BaseCohere.embed for the actual HTTP call.

Usage

Call on a Client or ClientV2 instance. Pass texts as a list of strings, specify the input_type for the use case, and optionally select embedding_types for the output format. Batching is automatic for text inputs.

Code Reference

Source Location: Repository cohere-ai/cohere-python https://github.com/cohere-ai/cohere-python, File src/cohere/client.py, Lines L180-224
Signature:

def embed(
    self,
    *,
    texts: typing.Optional[typing.Sequence[str]] = OMIT,
    images: typing.Optional[typing.Sequence[str]] = OMIT,
    model: typing.Optional[str] = OMIT,
    input_type: typing.Optional[EmbedInputType] = OMIT,
    embedding_types: typing.Optional[typing.Sequence[EmbeddingType]] = OMIT,
    truncate: typing.Optional[EmbedRequestTruncate] = OMIT,
    request_options: typing.Optional[RequestOptions] = None,
    batching: typing.Optional[bool] = True,
) -> EmbedResponse:

Import: from cohere import Client or from cohere import ClientV2 (Client.embed for V1 batching, ClientV2 inherits)

I/O Contract

Inputs

Parameter	Type	Required	Description
texts	Optional[Sequence[str]]	No	Text strings to embed (auto-batched at 96)
images	Optional[Sequence[str]]	No	Image URLs or base64 data (no batching)
model	Optional[str]	No	Embedding model ID e.g. "embed-english-v3.0"
input_type	Optional[EmbedInputType]	No	"search_document", "search_query", "classification", "clustering", "image"
embedding_types	Optional[Sequence[EmbeddingType]]	No	"float", "int8", "uint8", "binary", "ubinary"
truncate	Optional[EmbedRequestTruncate]	No	Truncation behavior
batching	Optional[bool]	No	Enable auto-batching (default True)

Outputs

EmbedResponse (discriminated union):

EmbeddingsFloatsEmbedResponse: id, embeddings (List[List[float]]), texts, meta
EmbeddingsByTypeEmbedResponse: id, embeddings (typed), texts, meta

Usage Examples

from cohere import Client

client = Client()

# Embed documents for search indexing
doc_response = client.embed(
    texts=["Machine learning is a subset of AI.", "Deep learning uses neural networks."],
    model="embed-english-v3.0",
    input_type="search_document",
    embedding_types=["float"],
)

# Access embeddings
embeddings = doc_response.embeddings  # List[List[float]]

# Embed a query
query_response = client.embed(
    texts=["What is deep learning?"],
    model="embed-english-v3.0",
    input_type="search_query",
)

# Large-scale embedding (auto-batched at 96)
large_response = client.embed(
    texts=["text " + str(i) for i in range(1000)],  # Auto-splits into batches of 96
    model="embed-english-v3.0",
    input_type="search_document",
)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment