#Assistant Not Using Vector Store

1 messages · Page 1 of 1 (latest)

stuck fossil
#

Having a problem with my assistant using its native OpenAI vector database. Its being called by a front-end web chat. The only way it seems to even consider using the vector store is if I tell it which source document to use in its answer. And then it cites [0] each time. I know this is not how it's supposed to work. I've even changed the prompt to include " Run retrieval and check your sources for the best answers from the documents" and it still doesn't check the docs unless I tell it to in the chat. Has anyone else experienced this? I'm discovering there is absolutely no way to see the queries it's sending to the vector store (or if its sending them at all, unless I'm mistaken). This is ... frustrating.

broken bloom
#

Have you tried telling in the prompt the ids of your files an provide som context to it?