Introduction Overview: Empower Your Tech Entrepreneurship Journey with Microsoft Learn
Are you ready to embark on an exciting journey into the world of tech entrepreneurship and artificial inte...
JohnAziz Hey John, thanks for the great article. I've worked through a few samples and have a question I hoped you would have an idea on.
Let's say I index 10,000 documents in AI Search, and I am able to do chat completions on them without issue. In my scenario, I may be able to "prefilter" the documents by the user doing some kind of metadata search on other metadata stored in a different index. For example, the user finds the 10 invoices tied to VendorId=123ABC. Let's say I now have a list of those 10 documents, along with document title and maybe document id in the index.
I'm trying to figure out if it's possible to narrow the scope of the retrieval portion, so I could instruct the chat completion service to only reference those "prefiltered" 10 invoice documents while it's doing the chat completion, and ignore the rest of the 10,000 document set. I've tried to change the "QueryType" to simple and set some parameters on the query for a key/value query, but no luck. I haven't been able to find much documentation on this part of the retrieval either.
Any insights you might have would be greatly appreciated!