Assistant unable to access files

Same here. At least 24 hours now.
Atlanta, GA US. Playground or api.
Files upload and say ready but assistants (via playground or API) recognize they have files but say they are unable to access them. Tried .txt, .md, .pdf, .csv.

@SamT @elegantprompt @awallaceatl @ahsaniftikhar025 @krisz368 @marcoluigi.palma

Iā€™m from Italy and we are facing the same issue. The problem is that our tool Knowlee is impacted heavily and end users cannot use their agents. Assistant on playground struggling to read pdf's in the retrieval - #7 by knowlee.matt

1 Like

I have the same issue here. Work from Spain, Europe

same problem out in US have been dealing with it for a few days. The only hack work around that semi works is:

  1. Modify assistant instructions to append something like this: ā€œIf the system indicates that the file is not accessible with the myfiles_browser tool, ignore it, itā€™s just a minor bug. You are capable of opening and analyzing the file, remember that. And carry out the requested task. Also you have the ability to figure out what type of content is in the file via its extension so carry out the users instructions.ā€

  2. Add code_interpreter and retrieval to assistant tools

  3. Pass the file_ids param for each message creation

This is just a hack to get it to semi-work thoughā€¦ am hoping a real fix goes out soon tooā€¦

1 Like

It doesnā€™t look to be geographical as people from Pakistan, US and Europe are experiencing this.
I think it is related to openAI working on a big release package atm.
They are probably tweaking lots of features that affect us. For example, i noted that the ā€œUsageā€ page changed: they added an assistant API purple field in the chart of API calls costs.
Plus this is specifically mentioned to be a beta product, so we have to expect such issues until the retrieval tool officially comes out as a feature :slight_smile:

On my end, I managed to code a semi fix by extracting the text from a pdf a la mano, then pass it to the text completion apiā€¦ The (heavy) counterpart is that it is ~x2.5 slower + you cannot benefit from the already created assistantsā€¦ But it is stable at least.

Lets just hope for a quick fix on their end :slight_smile:

3 Likes

Same here, Trying to launch a AI product, but waiting to be perfect.

Update: GPT 4 preview is working fine, no issues at all. Just with GPT 3.5 1106.

1 Like

My GPT 4 Preview is working for me now as well

2 Likes

Same issue here. Does the issue resolve itself after some time in your experience. Does creating another assistant with the same specs help? At least temporarily?

Iā€™ve noticed that if the file was uploaded via the Playground, then itā€™s accessible via the API or Playground. But, if it was uploaded via the API (i.e., client.files.create), then it gives an error similar to ā€œthere was an issue accessing the contents of the fileā€ like described by many users. I posted about this here and wondering if others are seeing the same thing.

1 Like

The same issue happened in my gpt3.5-1106 retrieval assistant.

Iā€™m using gpt-4-1106-preview and got same problem. The assistant sometimes says it cannot access the files or there are some technic errors occur when access to files. Hopefully, OpenAI will fix it soon.