LlamaIndex Embeddings Cohere

0.8.0 verified Fri May 01 auth: no python

LlamaIndex integration for Cohere embedding models. Current version 0.8.0, requires Python >=3.10 and <4.0. Released as part of the llama-index plugin ecosystem; version bumps may follow llama-index core releases.

pip install llama-index-embeddings-cohere

Common errors

error ImportError: cannot import name 'CohereEmbedding' from 'llama_index' ↓

cause Incorrect import path; trying to import directly from the top-level `llama_index` package.

fix

Use from llama_index.embeddings.cohere import CohereEmbedding.

error cohere.core.api_error.ApiError: status_code: 400, body: {'message': 'input_type is required'} ↓

cause Missing `input_type` parameter when using Cohere embedding models that require it (e.g., embed-english-v3.0).

fix

Pass input_type to the embedding call or during initialization.

error TypeError: __init__() got an unexpected keyword argument 'embed_batch_size' ↓

cause The `embed_batch_size` parameter was removed in newer versions of this library.

fix

Remove embed_batch_size from the constructor arguments.

Warnings

breaking Version 0.8.0 changed import path from `llama_index.embeddings.cohere` (previously may have been `llama_index.embeddings.cohere.base`). Ensure correct import. ↓

fix Use `from llama_index.embeddings.cohere import CohereEmbedding`.

deprecated The `embed_batch_size` parameter was removed; batching is now handled automatically. ↓

fix Remove `embed_batch_size` when constructing CohereEmbedding.

gotcha Cohere requires `input_type` parameter (e.g., 'search_query', 'search_document', 'classification'). Omitting it may cause errors from the API. ↓

fix Always specify `input_type` when calling `get_text_embedding` or during initialization.

Imports

CohereEmbedding

from llama_index.embeddings.cohere import CohereEmbedding

Correct import path as of version 0.8.0

CohereEmbedding
wrong
```
from llama_index import CohereEmbedding
```
correct
```
from llama_index.embeddings.cohere import CohereEmbedding
```
Direct import from llama_index package not supported; must use submodule path

Quickstart

Initialize Cohere embedding model, get a sample embedding.

import os
from llama_index.embeddings.cohere import CohereEmbedding

embed_model = CohereEmbedding(
    api_key=os.environ.get("COHERE_API_KEY", ""),
    model_name="embed-english-v3.0",
    input_type="search_query",
)
print(embed_model.get_text_embedding("Hello world!")[:5])