How can I fix ModuleNotFoundError for 'openai.embeddings_utils' when using the OpenAI embeddings API?

Below is an explanation and a couple of options to resolve your issue:

Why the Error Occurs

The error message

PYTHONCopied
ModuleNotFoundError: No module named ‘openai.embeddings_utils’

indicates that the module you’re trying to import isn’t part of the installed OpenAI Python package. In many of the cookbook examples (like those on the OpenAI documentation site), helper code (such as functions for computing cosine similarity) might be provided as a separate module (e.g. in a file named embeddings_utils.py) in the example repository and isn’t shipped with the official client library.

Options to Fix the Issue

1. Download the Auxiliary Utility Module
   The example you’re following may assume that you have cloned the OpenAI Cookbook repository which includes the file (or module) embeddings_utils.py. You can download or copy that file from the repository and include it in your project’s path.

2. Implement the Required Functions Locally
   If you only need to compute, say, cosine similarity or a similar helper function, you can implement it yourself. For instance, here’s a simple cosine similarity example in Python:

   PYTHONCopied
   import numpy as np
   
   def cosine_similarity(a, b):
       a, b = np.array(a), np.array(b)
       return np.dot(a, b) / (np.linalg.norm(a) * np.linalg.norm(b))

   You can then use this function to compare the embedding vectors of your queries and documents.

Summary

The error is due to an import line in the sample that expects a helper module which is not part of the core OpenAI package. Either obtain that module from the OpenAI Cookbook repository or implement the needed helper functionality locally. This should resolve your error when working with text search using embeddings.

If you have further questions on using the embeddings API or troubleshooting your implementation, let me know!