I am testing the Assistant on GPT-4o, specifically around the file search functionality. I have been trying to get this working via the API, but all round unsuccessful. So I deleted all the files, started afresh with the UI. I created an assistant, connected a vector store, and attached files to the vector store. I queried the document, with the file search tool enabled, but it is never able to answer the questions I have.
Below are the screenshots for an example of the setup.
Getting the exact same issue… however I can get it working in the playground but not via API. It just doesn’t want to use the file search, spent all of today trying to figure it out. Deleted the assistant and set it up again but still the same time. Would be great to see how many other are experiencing this.
The weird thing is I have another assistant that’s linked to the same vector store and its working fine, no problem.
I’ve been struggling with the same problem, I am able to use file search + functions in playground, but whenever I replicate the same conversation via API calls the assistant will tell me it has problems with its internal search. It was working fine for a week, did not changes to the instructions and it suddenly stopped searching.
I really want to push to production my use case, because it’s pretty simple, but w/o file search it is pointless.
If your API key is not set globally on your application it is worth verifying if you’re really using the same key for every request. Since the change of key policy when these project-specific keys were added, I notice that some endpoints were not talking to each other during test time (e.g. created a new Assistant but it was never listed with the other Assistants) and the reason for that was the use of the previous “global user key” in one endpoint and the use of a “project-specific key” in the other endpoint.
