I collected embeddings in a list or a pandas column, and np.array(embeddings) fails with an inhomogeneous shape error.
Why this happens
NumPy can only build a 2D array if every row has the same length. This error means at least one embedding has a different length. Common causes are a failed API call that returned an empty list or None, embeddings from two different models in the same list, or embeddings saved to CSV as strings and loaded back.
Answer 1: Find the rows with the wrong length
from collections import Counter
lengths = [len(e) if e is not None else 0 for e in embeddings]
print(Counter(lengths))
bad = [i for i, n in enumerate(lengths) if n != 384]
print(bad[:10])
This tells you whether you have missing values, a second model, or something else, and which rows to fix.
