No similar documents were found in file with ID 'file-xxxxx'

I’ve uploaded a file for classifications purpose and for a while I was able to work with the APIs but since this afternoon I keep getting the error where the message says “No similar documents were found in file with ID ‘file-xxxxxxxxxxx’. Please upload more documents or adjust your query.”

This is happening with curl and client libraries:
Here is the file information:

 {
   "id": "file-mng89Ky3Jyk4UzLxiBhJ7tUm",
   "object": "file",
  "bytes": 12622,
  "created_at": 1622179726,
  "filename": "ab_testing.jsonl",
  "purpose": "classifications",
  "status": "processed",
 "status_details": null
 }

And here is the error I’ve been getting:

{
  "error": {
  "code": null,
  "message": "No similar documents were found in file with ID 'file-mng89Ky3Jyk4UzLxiBhJ7tUm'.Please upload more documents or adjust your query.",
  "param": null,
  "type": "invalid_request_error"
  }
}

I deleted the file and uploaded it again but I am seeing the same error. Any help is appreciated.

Hello! Thanks for the question. Lemme see if I can clarify things.

When using the search endpoint with a file, we first filter down possible candidate documents before doing the expensive search call with our engines. That filtering step is keyword based so if there wasn’t an overlapping word between your query and the documents, you probably wont see any results. Does that help?

We’ve typically found that smaller datasets struggle a little more with files. I think your dataset is <100 entries so you can just submit all of these documents straight to the /search endpoint if you’d like.

2 Likes

Thanks for the quick response. I think I am bit lost here. I am not performing a “search” operation but trying to do a “classifications” task where I’ve uploaded the sample (about 60) “text” and corresponding “label” in the jsonl file. I am not really sure what you mean by “overlapping word”. Earlier in the afternoon I was getting the result as “unknown” when it could not match any of the labels which was quite accurate.

1 Like