Does Vector Index for 'CososDB for mongo DB Vcore' seamlessly reflect content changes and does this carry over to Azure Open Ai Studio?

Question

Does Vector Index for 'CososDB for mongo DB Vcore' seamlessly reflect content changes and does this carry over to Azure Open Ai Studio?

Junheok Cheon 120

Hello,

I am currently developing a chatbot that uses vector index from Cosmos DB for Mongo DB Vcore.

I have mainly two questions:

Does vector index in the Cosmos DB for Mongo DB Vcore seamlessly reflect changes made in the content of the data base? (For instance, would it reflect deletion of some entries, addition of new entries?)
If I am adding Vector Index from Cosmos DB for Mongo DB V core as a knowledge source (Azure Open AI Studio), would I need to redeploy the chatbot whenever there is changes made to vector index or does this automatically happen?

These questions are crucial for me as I am trying to automate the process of chatbot and I would need to know how these things work for me to further progress.

Thanks!!

Junheok Cheon 120 Reputation points

2025-01-08T10:30:16.87+00:00

Thanks for the reply!
I understand that I need to create vector embedding for each entry before I push it into the database. Then I would assume that when I add new entries (which will have vector embedding as part of its properties), vector index would reflect those new entries, and when I ask the question to the bot that uses this vector index as a knowledge source, it will reference off of the newly added entries. Is this scenario correct?
romungi-MSFT 48,906 Reputation points Microsoft Employee Moderator

2025-01-08T11:13:10.0233333+00:00

@Junheok Cheon I think it should as using Azure Cosmos DB for MongoDB vCore as your source for data to be vectorized and search gives the ability to dynamically add and remove data to be vectorized and immediately available for search. If it does not return immediately, you might have to start a new session as using conversation history might influence the response generated by the model.
romungi-MSFT 48,906 Reputation points Microsoft Employee Moderator

2025-01-09T11:19:30.69+00:00

@Junheok Cheon Did the above help answer your question? Thanks!!

1 answer

Your answer

Junheok Cheon 120 Reputation points

2025-01-08T10:30:16.87+00:00

Thanks for the reply!
I understand that I need to create vector embedding for each entry before I push it into the database. Then I would assume that when I add new entries (which will have vector embedding as part of its properties), vector index would reflect those new entries, and when I ask the question to the bot that uses this vector index as a knowledge source, it will reference off of the newly added entries. Is this scenario correct?
romungi-MSFT 48,906 Reputation points Microsoft Employee Moderator

2025-01-08T11:13:10.0233333+00:00

@Junheok Cheon I think it should as using Azure Cosmos DB for MongoDB vCore as your source for data to be vectorized and search gives the ability to dynamically add and remove data to be vectorized and immediately available for search. If it does not return immediately, you might have to start a new session as using conversation history might influence the response generated by the model.
romungi-MSFT 48,906 Reputation points Microsoft Employee Moderator

2025-01-09T11:19:30.69+00:00

@Junheok Cheon Did the above help answer your question? Thanks!!

Answer 1

@Junheok Cheon AI studio primarily is used as an integration point for all sources of AI applications within Azure. For your question about using this data source with Azure OpenAI:

Does vector index in the Cosmos DB for Mongo DB Vcore seamlessly reflect changes made in the content of the data base? (For instance, would it reflect deletion of some entries, addition of new entries?)

I think there are three categories to this generating vector, searching vectors and generating completions. Vectors should be generated when data is inserted into each of the collections in Azure Cosmos DB for MongoDB vCore, and then stored in a property which can be used for vector searches. When users ask questions, the prompts are vectorized and used to perform the vector search in a collection. So, whenever changes are made to DB you will have to ingest and vectorize the data which will be further available for OpenAI. This solution and repo primarily uses this scenario. You can try to replicate this or use the repo as-is to test the same.

If I am adding Vector Index from Cosmos DB for Mongo DB V core as a knowledge source (Azure Open AI Studio), would I need to redeploy the chatbot whenever there is changes made to vector index or does this automatically happen?

If your vector index is updated the search should automatically feed the chatbot. I don't think you need to redeploy your chat endpoint unless you need to use a new source. I think you can test this with a simple change in your DB after your first integration to confirm.

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.

Share via

Does Vector Index for 'CososDB for mongo DB Vcore' seamlessly reflect content changes and does this carry over to Azure Open Ai Studio?

1 answer

Your answer