In Azure OpenAI Studio, I'm working on Chat Playground, trying to connect to my own Azure Cosmos DB for MongoDB (vCore). I can't select an index name; there's no dropdown. What's the issue?

Question

In Azure OpenAI Studio, I am currently engaged in the Chat Playground project, where I am endeavoring to establish a connection to my personal data source via Azure Cosmos DB for MongoDB (vCore). I have encountered a predicament during the selection of the database, as the database name and collection name have already been designated, but I find myself unable to select an index name. Curiously, there seems to be no available dropdown menu for the index name field. I am inquiring as to the root cause of this particular issue.

Answer 1

Even I faced the issue. The problem is in creating the correct type of index. We have to create the cosmoSearch index, then only the index drop-down will be enabled. Try out the below code. This can be executed before or after loading the data into the collection. The field contentVector should be created in each documents.

db.runCommand({

createIndexes: "your_collection_name",

indexes: [

{

name: "vectorSearchIndex",

key: {

contentVector: "cosmosSearch"

},

cosmosSearchOptions: {

kind: "vector-ivf",

numLists: 1,

similarity: "COS",

dimensions: 239

}

}

]

});

Answer 2

I have a similar problem while trying to connect to Cosmos. I already created the index as a Cosmos DB for MongoDB (vCore), it appears in the dropbox but it says password doesn't match. I already made multiple attempts reseting the password but it doesn't work.

Very frustrating.

Answer 3

Same problem ! Index Name is disabled. The index is healthy, I used Studio 3T to create it and it is functioning well. Please help

Answer 4

I am facing the same issue. If anyone finds a resolution please post.

Answer 5

Watch this video to get an example working. It does a good job of explaining how it works.

[https://www.youtube.com/watch?v=MLY5Pc_tSXw&t=1613s](https://www.youtube.com/watch?v=MLY5Pc_tSXw&t=1613s

)

You'll see the json sample data has a vector array. Each item in the collection will contain the vector embeddings. After I followed the steps and loaded the data, I was able to go to Chat Playground and add the DB/Collection and the Index was no longer grayed out. I then successfully tried test questions in the chat. I guess the challenge will be creating a load process for data with the embeddings.

These are links to Github repo where you can download the example.

https://aka.ms/adgen

https://github.com/jayanta-mondal/ignite-demo