Cohere

export COHERE_API_KEY="your-cohere-api-key"

import catsu

client = catsu.Client()

response = client.embed(
    model="embed-v4.0",
    input="Hello, Cohere!"
)

print(f"Dimensions: {response.dimensions}")  # 1536

# For search queries
query_response = client.embed(
    model="embed-v4.0",
    input="What is natural language processing?",
    input_type="query"  # → search_query in Cohere API
)

# For documents
doc_response = client.embed(
    model="embed-v4.0",
    input="NLP is a field of AI that focuses on...",
    input_type="document"  # → search_document in Cohere API
)

# Truncate from the end (keep beginning)
response = client.embed(
    model="embed-v4.0",
    input="Very long text that exceeds the token limit...",
    truncate="END"
)

# Truncate from the start (keep ending)
response = client.embed(
    model="embed-v4.0",
    input="Long text...",
    truncate="START"
)

# Don't truncate (raises error if too long)
response = client.embed(
    model="embed-v4.0",
    input="Text",
    truncate="NONE"
)

# English-only model
english_response = client.embed(
    model="embed-english-v3.0",
    input="English text only"
)

# Multilingual model
multilingual_response = client.embed(
    model="embed-multilingual-v3.0",
    input="Texto en español"  # Spanish text
)

# Light multilingual model (faster, smaller)
light_response = client.embed(
    model="embed-multilingual-light-v3.0",
    input="Texte en français"  # French text
)

texts = ["Document 1", "Document 2", "Document 3"]

response = client.embed(
    model="embed-v4.0",
    input=texts,
    input_type="document"
)

print(f"Embedded {len(response.embeddings)} documents")

import asyncio

async def main():
    client = catsu.Client()

    response = await client.aembed(
        model="embed-v4.0",
        input="Async Cohere embedding",
        input_type="query"
    )

    print(response.embeddings)

asyncio.run(main())