NotFoundError: Error code: 404 - {'error': {'code': '404', 'message': 'Resource not found'}} when using Azure Openai with Langchain

Question

I am facing an issue resource not found error. This is what I have tried:

1. Checked the version, azure_openai_api_key, modelname, version and everything is correct.
2. The model was deployed yesterday so more than 5 minutes have passed.
3. I already have an embeddings model in Azure openAI and embeddings are created and stored in local.

I have cross checked the logs in Azure and the console says that the requests are reached.

I am not sure why I am getting this error even though everything seems to be correct and in line with Langchain and Azure OpenAI documentation.

import os
import dill  # Import dill instead of pickle
import streamlit as st
from dotenv import load_dotenv
from langchain_community.document_loaders import PyPDFLoader
from langchain_community.vectorstores import FAISS
from azure.storage.blob import BlobServiceClient
from langchain_openai import AzureOpenAIEmbeddings
from langchain_openai import AzureOpenAI
from langchain.chains import RetrievalQA
from langchain.prompts import PromptTemplate

load_dotenv()

//ex: entries in .env
AZURE_OPENAI_ENDPOINT="https://jkazureopenaiservice.openai.azure.com/"
AZURE_OPENAI_API_KEY = "key"
DEPLOYMENT_NAME = "gpt-4o"  # The name of your model deployment in Azure
AZURE_OPENAI_API_VERSION = "2024-05-13"
AZURE_OPENAI_DEPLOYMENT = "gpt-4o"


# Azure OpenAI configuration
AZURE_OPENAI_API_BASE = os.getenv("AZURE_OPENAI_API_BASE")
AZURE_OPENAI_API_KEY = os.getenv("AZURE_OPENAI_API_KEY")
AZURE_OPENAI_API_VERSION = "2024-05-13"
DEPLOYMENT_NAME = os.getenv("DEPLOYMENT_NAME")


# Path to the folder containing your PDFs
pdf_folder_path = './localKnowledgeHub'

# Initialize a list to hold all documents
all_documents = []

# Load each PDF file
for filename in os.listdir(pdf_folder_path):
    if filename.endswith('.pdf'):
        file_path = os.path.join(pdf_folder_path, filename)
        loader = PyPDFLoader(file_path)
        documents = loader.load()
        all_documents.extend(documents)

# Initialize the embeddings model
embeddings = AzureOpenAIEmbeddings(
    model="text-embedding-3-large"
)

# Create the FAISS vector store
#vector_store = FAISS.from_documents(all_documents, embeddings)
#vector_store.save_local("vectorstore")

# Load the vector store locally
vector_store = FAISS.load_local("vectorstore", embeddings, allow_dangerous_deserialization=True)

# Initialize the Azure OpenAI language model
llm = AzureOpenAI(
    api_key=AZURE_OPENAI_API_KEY,
    api_version=AZURE_OPENAI_API_VERSION,
    deployment_name=DEPLOYMENT_NAME,
    temperature=70
)

# Define the prompt template
PROMPT_TEMPLATE = """You are an AI Assistant. Given the following context:
{context}
Answer the following question:
{question}
Assistant:"""

PROMPT = PromptTemplate(
    template=PROMPT_TEMPLATE, input_variables=["context", "question"]
)
# Create a retrieval QA chain using LangChain with the custom prompt
qa_chain = RetrievalQA.from_chain_type(
    llm=llm,
    chain_type="stuff",  # 'stuff' means directly use documents retrieved
    retriever=vector_store.as_retriever(),
    chain_type_kwargs={"prompt": PROMPT}
)

# Set up the Streamlit UI
st.title("Chatbot")
st.write("Ask me anything a.")

# Input box for user questions
user_question = st.text_input("Question:")

# When the user submits a question
if user_question:    
    # Get the answer from the QA chain using the invoke method
    try:
        result = qa_chain.invoke({"query": user_question})
        answer = result['output_text']
    except Exception as e:
        answer = f"An error occurred: {str(e)}"
    
    # Display the answer
    st.write("Answer:", answer)

Answer 1

According to this document the latest preview and GA versions are 2024-07-01-preview and 2024-06-01.

And whatever you given the api version is not supported, so you are getting no resource found error.

Try any one of the above api version.

I tried you code with sample pdf file give one of the above api version.

Output:

You can also find sample code in your ai studio chat playground, where you will the api version supported.

and the code.

It is also recommended to use azure_ad_token_provider instead of using api key.

Answer 2

Instead of AzureOpenAI, try using AzureChatOpenAI.

from langchain_openai import AzureChatOpenAI

llm = AzureChatOpenAI(
    azure_deployment="name_of_your_deployment",  
    api_version="2024-05-01-preview", 
    temperature=0,
    max_tokens=None,
    # other params...
)

Details here: https://python.langchain.com/v0.2/docs/integrations/chat/azure_chat_openai/