Choosing Between Keyword Search, Vector Search, Keyword-Vector Hybrid Search, Semantic Search, and Keyword, Vector, Semantic Hybrid Search when doing RAG with Python SDK for Azure OpenAI

Bao, Jeremy (Cognizant) 105

If you want to do RAG just with simple keyword search, you do the following, right?:

response = client.chat.completions.create(
    model = ...,
    messages = ...,
    extra_body = {
		dataSources = [
            { 
                "type": "AzureCognitiveSearch", 
                "parameters": { 
                    "endpoint": ..., 
                    "key": ..., 
                    "indexName": ...,
                }
            }
        ]
    }
)

The following is if you wish to do vector search, right? Does it do only vector search, or keyword-vector hybrid search? If it carries out one of those, how do you do the other type of search?

response = client.chat.completions.create(
    model = ...,
    messages = ...,
    extra_body = {
		dataSources = [
            {
                "type": "AzureCognitiveSearch",
                "parameters": {
                    "endpoint": ...,
                    "key": ...,
                    "indexName": ...,
                    "vectorProfile": ...,
                    "vectorField": ...
                }
            }
        ]
    }
)

How would you modify this to do semantic search and keyword, vector, and semantic hybrid search?

navba-MSFT 20,810 Reputation points Microsoft Employee

2024-04-04T03:18:54.6033333+00:00

@Bao, Jeremy (Cognizant) Welcome to Microsoft Q&A Forum, Thank you for posting your query here!

.

Question 1: If you want to do RAG just with simple keyword search, you do the following, right?

Answer: Yes, your first code snippet is correct for performing a simple keyword search using the Azure OpenAI service. You can refer to the Azure OpenAI Service documentation for more details.

.

.

.

Question 2: The following is if you wish to do vector search, right? Does it do only vector search, or keyword-vector hybrid search? If it carries out one of those, how do you do the other type of search?

Answer: Your second code snippet is correct for performing a vector search. By adding the vectorProfile and vectorField parameters, you’re instructing the Azure Cognitive Search to perform a vector search. This does not perform a hybrid keyword-vector search, it only performs a vector search. To perform a hybrid keyword-vector search, you would need to implement a custom solution that combines the results of a keyword search and a vector search. You can refer to the Vector Search documentation for more details.

.

.

.

Question 3: How would you modify this to do semantic search and keyword, vector, and semantic hybrid search?

Answer: Azure Cognitive Search supports semantic search capabilities which can be enabled by setting the queryType parameter to ‘semantic’. However, this is typically used in the context of a search query, not in the context of generating chat completions. Combining keyword, vector, and semantic search would require a complex custom solution. You would need to perform each type of search separately (keyword search, vector search, semantic search) and then merge and rank the results based on your specific needs. You can refer to the Semantic Ranking documentation and Hybrid Search documentation for more details.

.

.

Hope this helps. If you have any follow-up questions, please let me know. I would be happy to help.
navba-MSFT 20,810 Reputation points Microsoft Employee

2024-04-09T03:04:13.0066667+00:00

@Bao, Jeremy (Cognizant) Just following up to check if my suggestion helped. Please let me know if you have any further queries. I would be happy to help.
Bao, Jeremy (Cognizant) 105 Reputation points

2024-04-09T04:47:13.8333333+00:00
I have switched to using the requests library in Python to make API calls instead of using the Python SDK. I have:

s = requests.Session() retries = Retry( total = 5, backoff_factor = 2 ) s.mount("https://", HTTPAdapter(max_retries=retries)) response = s.post( url = request_endpoint, data = request_body, headers = headers ).json()

This lets me choose 'simple', 'semantic', 'vector', 'vectorSimpleHybrid' or 'vectorSemanticHybrid' for query_type for the object in the data_sources array in the request_body JSON. Is this correct?
navba-MSFT 20,810 Reputation points Microsoft Employee

2024-04-10T05:41:34.7833333+00:00

@Bao, Jeremy (Cognizant) If the below answer helped, Please do not forget to "Accept the answer” and “up-vote” wherever the information provided helps you, this can be beneficial to other community members.

Accepted answer

navba-MSFT 20,810 Reputation points Microsoft Employee

2024-04-09T05:00:33.2133333+00:00

@Bao, Jeremy (Cognizant) Yes, your understanding is correct. Using the requests library in Python to make API calls is a common practice and it provides you with flexibility. The query_type parameter in the data_sources array in the request_body JSON indeed allows you to specify the type of search you want to perform. The options ‘simple’, ‘semantic’, ‘vector’, ‘vectorSimpleHybrid’, and ‘vectorSemanticHybrid’ correspond to different types of searches you can perform with Azure OpenAI service.
Please sign in to rate this answer.
navba-MSFT 20,810 Reputation points Microsoft Employee

2024-04-10T05:42:04.0733333+00:00

@Bao, Jeremy (Cognizant) If the above answer helped, Please do not forget to "Accept the answer” and “up-vote” wherever the information provided helps you, this can be beneficial to other community members.
Sign in to comment

Share via

Choosing Between Keyword Search, Vector Search, Keyword-Vector Hybrid Search, Semantic Search, and Keyword, Vector, Semantic Hybrid Search when doing RAG with Python SDK for Azure OpenAI

0 additional answers