FAISS vector database integration with Azure OpenAI embeddings failing

[Ethan99]

I keep running into issues when trying to build a vector database using FAISS with Azure OpenAI embeddings. Getting two different errors that I cannot figure out.

First error is:

AttributeError: 'str' object has no attribute 'create'

Second error is:

NotFoundError: Error code: 404 - {'error': {'code': '404', 'message': 'Resource not found'}}

Here is my code setup:

from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.vectorstores import FAISS

embedder = OpenAIEmbeddings(
    openai_api_key="sk-proj-xxxxxxxxxx", 
    model='text-embedding-ada-002', 
    deployment='my-embedding-model', 
    client="azure", 
    chunk_size=5
)

# Sample documents for indexing
documents = ["Vector databases are powerful tools", "FAISS provides efficient similarity search"]
vector_db = FAISS.from_texts(documents, embedder)

I also tried loading from PDF files but same issues:

from langchain.document_loaders.pdf import PyPDFLoader

file_loader = PyPDFLoader("./documents/sample_document.pdf")
doc_pages = file_loader.load_and_split()

vector_store = FAISS.from_documents(doc_pages, embedder)

Working in Jupyter notebook environment. What configuration am I missing for Azure OpenAI integration?

[SparklingGem]

you’re mixin up OpenAI and Azure stuff. try usin AzureOpenAIEmbeddings instead of OpenAIEmbeddings. also, check if your azure_endpoint and api_version are right. that 404 error usually means your endpoint is wrong or the deployment name ain’t matching what’s in your Azure portal.

[JackHero77]

Your Azure OpenAI setup has some issues. Make sure you’re using the azure_endpoint parameter and set openai_api_type="azure". Drop the client="azure" parameter - it’s not valid for this embeddings class. Your endpoint should look like https://your-resource-name.openai.azure.com/. Double-check that your deployment name matches exactly what’s in the Azure portal (it’s case-sensitive). I had the same errors until I updated my API version - try api_version="2023-05-15" or newer. That AttributeError you’re seeing means the client isn’t initializing properly because of missing or wrong Azure parameters.