I’m using embedding model mxbai-embed-large.

embed_response = ollama.embeddings(model=em_model, prompt='instring')
embed_list = embed_response['embedding']

retrieval_results = collection.query(query_embeddings=[embed_list],
n_results=30)

embed_list looks fine and is a list of 1024 numbers as I’d expect.

retrieval_results[‘documents’][0] comes back as a list of lists as it should be, with 30 documents

Simlarly, retrieval_results[‘ids’] and ‘metadatas’ are full of data like I’d expect.

But retrieval_results[‘distances’][0] is [0.0, 0.0, 0.0, 0.0 … 0.0]

Where are my distances? I just switched to cosine distance but had the same problem with i2.