Using Llama-index with gpt-35-turbo for QNA with Azure OpenAI Service

llama-index-chatgpt-azure-openai-service.py

import os
import openai
from dotenv import load_dotenv
from llama_index import GPTVectorStoreIndex, SimpleDirectoryReader, PromptHelper, LangchainEmbedding, ServiceContext
from llama_index.llms import AzureOpenAI
from langchain.embeddings import OpenAIEmbeddings

# Load environment variables (set OPENAI_API_KEY and OPENAI_API_BASE in .env)
load_dotenv()

# Configure Azure OpenAI Service API
openai.api_type = "azure"
openai.api_base = os.getenv('OPENAI_API_BASE')
openai.api_key = os.getenv("OPENAI_API_KEY")
openai.api_version = os.getenv('OPENAI_API_VERSION')
# openai.log = "debug"

# Initialize LLM and Embeddings model (model is the actual model name, e.g., gpt-35-turbo, engine is your custom deployment name, e.g., my-gpt-35-turbo)
llm = AzureOpenAI(engine="gpt-35-turbo", model="gpt-35-turbo", temperature=0.0)
embeddings = LangchainEmbedding(OpenAIEmbeddings(deployment_id="text-embedding-ada-002", chunk_size=1))

# Define prompt helper
prompt_helper = PromptHelper(context_window=3000, 
                             num_output=500, 
                             chunk_overlap_ratio=0.1, 
                             chunk_size_limit=1000)

service_context = ServiceContext.from_defaults(llm=llm, embed_model=embeddings, prompt_helper=prompt_helper)

# Load documents
documents = SimpleDirectoryReader('data/qna/').load_data()

# Create index
index = GPTVectorStoreIndex.from_documents(documents, service_context=service_context, prompt_helper=prompt_helper)
query_engine = index.as_query_engine(service_context=service_context, verbose=True)

query = "What is azure openai service? give me back a bullet point list"
answer = query_engine.query(query)

print(f"Query: {query}")
print(f"Answer: {answer}")
print(f"Sources: {answer.get_formatted_sources()}")