Integrating Search Functionality with OpenAI Models on Azure AI Platform

Hi Isuan Agarwal,

Integrating a search functionality with Azure OpenAI models to mimic how ChatGPT Browsing works involves building a Retrieval-Augmented Generation (RAG) pipeline. This process allows your AI to fetch fresh information from the web or internal data and generate informed responses. Here’s a step-by-step implementation guide tailored to Azure:

Step-by-Step Implementation:

Step1: Set Up Azure OpenAI

·   Create an Azure OpenAI resource from Azure Portal.

·    Deploy a model like gpt-4, gpt-35-turbo, etc.

·    Get endpoint and API key for calling the model.

Step2: Set Up the Search Functionality

Option A: Web Search (Live Internet Data)

·    Use Bing Search API (via Azure Cognitive Services).

o   Bing Search v7

o   Get your subscription key and endpoint.

Option B: Internal Document Search

·    Use Azure AI Search or OpenSearch:

o   Index your own documents (PDFs, Word, text).

o   Use vector search and semantic search to find relevant content.

Step3: Create a RAG (Retrieval-Augmented Generation) Pipeline

User Query → Search Engine (Bing or Azure Search) → Fetch Results (top 3) → Combine with Prompt → Send to Azure OpenAI → Return Response

Step4: Code the Flow (Example with Bing + Azure OpenAI in Python)

·   Search with Bing API

import requests

def bing_search(query, subscription_key, endpoint):
    headers = {"Ocp-Apim-Subscription-Key": subscription_key}
    params = {"q": query, "textDecorations": True, "textFormat": "HTML", "count": 3}
    response = requests.get(endpoint, headers=headers, params=params)
    return [res['snippet'] for res in response.json()['webPages']['value']]

·   Combine with Prompt and Call OpenAI

import openai

def generate_answer(prompt, model, openai_endpoint, api_key):
    openai.api_base = openai_endpoint
    openai.api_key = api_key
    response = openai.ChatCompletion.create(
        model=model,
        messages=[{"role": "system", "content": "You're a helpful assistant."},
                  {"role": "user", "content": prompt}]
    )
    return response['choices'][0]['message']['content']

·   Full Pipeline

def ask_question_with_search(query):
    search_results = bing_search(query, bing_key, bing_endpoint)
    context = "\n".join(search_results)
    prompt = f"Answer the following based on the web results:\n\n{context}\n\nQuestion: {query}"
    return generate_answer(prompt, model, openai_endpoint, openai_key)

Step5: Host the Application

·   Use FastAPI or Flask for API endpoints.

·   Optionally deploy on Azure App Service or Azure Container Apps.

Hope this helps. Do let us know if you any further queries.

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Thank you!