Task 05: Use generative AI for recommendations

1. Return to the browser that is signed into the Azure AI Foundry > Azure OpenAI Service.

2. On the left menu, select Shared resources > Deployments.

3. On the Model deployments tab select + Deploy model > Deploy base model.

   

4. Select the Inference tasks menu and select Chat completion.

5. From the Chat completion options, select gpt-4 and then select Confirm.

6. On the Deploy model GPT-4 window, select Deploy.

7. On the Details tab, under Endpoint, copy and paste the following value in a notepad file for future use:

   Default	Value
   Target URI:
   KEY:

You want to help customers to find the best set of products for something, for example, to organize a high-school graduation party.

1. Return to the Visual Studio Code workspace and enter the following in a new query:

   declare @search_text nvarchar(max) = 'help me plan a high school graduation party'
   
   -- Get the search vector for the search text
   declare @search_vector vector(1536)
   exec dbo.create_embeddings @search_text, @search_vector output;
   
   -- Get the top 50 products that are closest to the search vector
   drop table if exists #t;
   with cte as 
   (
       select         
           id, product_name, [description], product_description_vector,        
           row_number() over (partition by product_name order by id ) as rn
       from 
           [dbo].[walmart_product_details]
       where 
           available = 'TRUE'
   ), 
   cte2 as -- remove duplicates
   (
       select 
           *
       from
           cte 
       where
           rn = 1
   )
   select top(50)
       id, product_name, [description],
       vector_distance('cosine', @search_vector, product_description_vector) as distance
   into
       #t
   from 
       cte2
   order by 
       distance;
   
   -- Aggregate the search results to make them easily consumable by the LLM
   declare @search_output nvarchar(max);
   select 
       @search_output = string_agg(cast(t.[id] as varchar(10)) +'=>' + t.[product_name] + '=>' + t.[description], char(13) + char(10))
   from 
       #t as t;
   
   -- Generate the payload for the LLM
   declare @llm_payload nvarchar(max);
   set @llm_payload = 
   json_object(
       'messages': json_array(
               json_object(
                   'role':'system',
                   'content':'
                       You are an awesome AI shopping assistant  tasked with helping users find appropriate items they are looking for the occasion. 
                       You have access to a list of products, each with an ID, product name, and description, provided to you in the format of "Id=>Product=>Description". 
                       When users ask for products for specific occasions, you can leverage this information to provide creative and personalized suggestions. 
                       Your goal is to assist users in planning memorable celebrations using the available products.
                   '
               ),
               json_object(
                   'role':'user',
                   'content': '## Source ##
                       ' + @search_output + '
                       ## End ##
   
                       Your answer needs to be a json object with the following format.
                       {
                           "answer": // the answer to the question, add a source reference to the end of each sentence. Source reference is the product Id.
                           "products": // a comma-separated list of product ids that you used to come up with the answer.
                           "thoughts": // brief thoughts on how you came up with the answer, e.g. what sources you used, what you thought about, etc.
                       }'
               ),
               json_object(
                   'role':'user',
                   'content': + @search_text
               )
       ),
       'max_tokens': 800,
       'temperature': 0.3,
       'frequency_penalty': 0,
       'presence_penalty': 0,
       'top_p': 0.95,
       'stop': null
   );
   
   -- Invoke the LLM to get the response
   declare @retval int, @response nvarchar(max);
   declare @headers nvarchar(300) = N'{"api-key": "<YourOpenAIKey>", "content-type": "application/json"}';
   exec @retval = sp_invoke_external_rest_endpoint
       @url = '<Your GPT4 Endpoint URL>',
       @credential = [https://<your-openai-resource>.openai.azure.com/],
       @method = 'POST',    
       @timeout = 120,
       @payload = @llm_payload,
       @response = @response output;
   select @retval as 'Return Code', @response as 'Response';
   
   -- Get the answer from the response
   select [key], [value] 
   from openjson(( 
       select t.value 
       from openjson(@response, '$.result.choices') c cross apply openjson(c.value, '$.message') t
       where t.[key] = 'content'
   ))
   

2. Execute the SQL statement.

3. View the return message and see what the GPT-4 model was able to recommend.

   Try using different search phrases. For example, for the question: “I am looking for a gift for my friend who is a foodie. She loves to cook and bake. She is always trying out new recipes and experimenting with different ingredients. I want to get her something that will inspire her creativity in the kitchen. Do you have any suggestions?”, you may get:

   For a friend who loves to cook and experiment in the kitchen, consider the Growing up in a Korean Kitchen: A Cookbook which offers a blend of memoir and recipes, perfect for exploring traditional and modern Korean dishes (7779).

   Another great choice could be The New Whole Foods Encyclopedia, a comprehensive resource for selecting, preparing, and using whole foods, which can inspire new culinary creations (10947).

   For a touch of fun and practicality, the Best Choice Products Mason Jars w/Labeling Stickers & Chalk, Set of 5 would be perfect for storing spices or homemade jams (7579).

Congratulations! You’ve successfully completed this task.