Skip to main content
Chroma integrates with Cohere’s powerful embedding models, offering excellent multilingual support and semantic understanding.

​
Installation

pip install chromadb cohere

​
Python usage

​
Basic setup

import chromadb
from chromadb.utils.embedding_functions import CohereEmbeddingFunction
import os

# Set your API key
os.environ["COHERE_API_KEY"] = "your-api-key"

# Create embedding function
embedding_function = CohereEmbeddingFunction(
    api_key=os.environ["COHERE_API_KEY"],
    model_name="embed-english-v3.0"
)

# Create client and collection
client = chromadb.Client()
collection = client.create_collection(
    name="cohere_embeddings",
    embedding_function=embedding_function
)

# Add documents
collection.add(
    documents=["This is a document", "Another document"],
    ids=["doc1", "doc2"]
)

# Query
results = collection.query(
    query_texts=["search query"],
    n_results=2
)

​
Available models

# embed-english-v3.0 (recommended)
# Dimensions: 1024
# Best for English content
embedding_function = CohereEmbeddingFunction(
    model_name="embed-english-v3.0"
)

# embed-english-light-v3.0
# Dimensions: 384
# Faster and more efficient
embedding_function = CohereEmbeddingFunction(
    model_name="embed-english-light-v3.0"
)

​
Input types

Cohere models support different input types for optimized embeddings: 
# For search documents
embedding_function = CohereEmbeddingFunction(
    api_key=os.environ["COHERE_API_KEY"],
    model_name="embed-english-v3.0",
    input_type="search_document"  # Optimize for document indexing
)

collection = client.create_collection(
    name="documents",
    embedding_function=embedding_function
)

# Add documents with optimized embeddings
collection.add(
    documents=["Document content"],
    ids=["doc1"]
)

# For search queries, use search_query input type
query_embedding_function = CohereEmbeddingFunction(
    api_key=os.environ["COHERE_API_KEY"],
    model_name="embed-english-v3.0",
    input_type="search_query"  # Optimize for queries
)

# Create multilingual collection
embedding_function = CohereEmbeddingFunction(
    model_name="embed-multilingual-v3.0"
)

collection = client.create_collection(
    name="multilingual_docs",
    embedding_function=embedding_function
)

# Add documents in different languages
collection.add(
    documents=[
        "Hello world",  # English
        "Hola mundo",   # Spanish
        "Bonjour le monde",  # French
        "こんにちは世界"  # Japanese
    ],
    ids=["en", "es", "fr", "ja"]
)

# Query in any language
results = collection.query(
    query_texts=["greetings"],
    n_results=4
)

​
JavaScript usage

import { ChromaClient } from "chromadb";
import { CohereEmbeddingFunction } from "chromadb";

const embedder = new CohereEmbeddingFunction({
  cohere_api_key: "your-api-key",
  model: "embed-english-v3.0"
});

const client = new ChromaClient();
const collection = await client.createCollection({
  name: "cohere_collection",
  embeddingFunction: embedder
});

// Add and query documents
await collection.add({
  ids: ["id1", "id2"],
  documents: ["Document 1", "Document 2"]
});

const results = await collection.query({
  queryTexts: ["search query"],
  nResults: 2
});

​
Configuration

​
api_key
string
required
Your Cohere API key
​
model_name
string
default:"embed-english-v3.0"
Cohere embedding model to use
​
input_type
string
Input type: search_document, search_query, classification, or clustering
​
truncate
string
Truncation strategy: NONE, START, or END

​
Input types explained

Use when indexing documents into your search system. Optimizes embeddings for being searched against.
Use when embedding search queries. Optimizes embeddings for finding relevant documents.
Use when using embeddings for classification tasks.
Use when performing clustering analysis on embeddings.

​
Supported languages

Cohere’s multilingual models support over 100 languages including:
  • European: English, Spanish, French, German, Italian, Portuguese, Russian, Polish, Dutch, Czech, etc.
  • Asian: Chinese, Japanese, Korean, Hindi, Thai, Vietnamese, Indonesian, etc.
  • Middle Eastern: Arabic, Hebrew, Turkish, Farsi, etc.
  • Other: See Cohere’s language support documentation for the full list

​
Best practices

Use the right model

Use multilingual models only if you need multi-language support. English models are faster for English-only content.

Specify input types

Always specify input_type for better embedding quality. Use search_document for indexing and search_query for queries.

Batch efficiently

Cohere supports up to 96 texts per request. Batch your requests for better performance.

Monitor usage

Track your API usage through the Cohere dashboard.

​
Error handling

import time
from cohere.errors import RateLimitError, AuthenticationError

try:
    collection.add(
        documents=["document"],
        ids=["id1"]
    )
except RateLimitError:
    # Handle rate limit
    time.sleep(60)  # Wait before retrying
except AuthenticationError:
    print("Invalid API key")

​
Resources

Free tier: Cohere offers a generous free tier for experimentation. Perfect for getting started with multilingual search.

Build docs developers (and LLMs) love

Get started for freeTalk to us