LlamaIndex Embeddings IBM

0.6.0.post1 verified Sat May 09 auth: no python

LlamaIndex integration for IBM watsonx.ai embeddings. Provides wrapper classes to generate embeddings using IBM foundation models via watsonx.ai API. Current version 0.6.0.post1, supports Python >=3.11, <4.0. Packages are released frequently alongside llama-index core updates.

pip install llama-index-embeddings-ibm

Common errors

error ModuleNotFoundError: No module named 'ibm_watsonx_ai' ↓

cause Missing required dependency `ibm-watsonx-ai` not installed.

fix

Run pip install ibm-watsonx-ai.

error ImportError: cannot import name 'WatsonxEmbeddings' from 'llama_index.embeddings.watsonx' ↓

cause Import path changed after version 0.5.0.

fix

Use from llama_index.embeddings.ibm import WatsonxEmbeddings.

error AuthenticationError: Bearer token is invalid. ↓

cause Missing or incorrect API key or project ID.

fix

Ensure WATSONX_APIKEY and WATSONX_PROJECT_ID are set correctly.

Warnings

breaking In version 0.5.0, the import path changed from `llama_index.embeddings.watsonx` to `llama_index.embeddings.ibm`. Update all imports. ↓

fix Use `from llama_index.embeddings.ibm import WatsonxEmbeddings`

deprecated Passing `credentials` as a dict is deprecated. Use explicit parameters (`url`, `apikey`, `project_id`) instead. ↓

fix Remove `credentials` argument and set individual parameters.

gotcha The `model_id` must include the provider prefix (e.g., 'ibm/slate-125m-english-rtrvr'). Omitting 'ibm/' may cause a silent fallback or error. ↓

fix Always prefix your model with 'ibm/'.

Imports

WatsonxEmbeddings

wrong

from llama_index.embeddings.watsonx import WatsonxEmbeddings

correct

from llama_index.embeddings.ibm import WatsonxEmbeddings

The package namespace changed from 'watsonx' to 'ibm' in version 0.5.0+

Quickstart

Initialize WatsonxEmbeddings with credentials and generate an embedding.

from llama_index.embeddings.ibm import WatsonxEmbeddings

embedding = WatsonxEmbeddings(
    model_id="ibm/slate-125m-english-rtrvr",
    url=os.environ.get("WATSONX_URL", "https://us-south.ml.cloud.ibm.com"),
    apikey=os.environ.get("WATSONX_APIKEY", ""),
    project_id=os.environ.get("WATSONX_PROJECT_ID", ""),
)

embeddings = embedding.get_text_embedding("Hello World!")
print(embeddings[:5])