If you disable retrieval, doesn’t that defeat the purpose of the upload if you want the Assistant to reference the document?
2 Likes
Is retrieval and .csv files working with the API?
1 Like
No, it’s not. It hasn’t worked since launch in the API or the Playground. I’m quite surprised that OpenAI hasn’t fixed it already.
Same here via the API. Worked when I converted it to JSON . JSONL did not work
1 Like
Getting the same CSV file issue as the OP and others in this thread.
Switching to Code Retrieval, as suggested by some, doesn’t seem like a valid work-around if one is actually needing to use the Assistant function.
It looks like this is a related issue - New Assistants API bug when uploading files (I’m not allowed to include a link, but that’s the post title)
I have same issue. Failed to upload a CSV file.
Facing a similar issue when uploading CSV files.
+1. I tried uploading pdf, txt and csv files and, if Retrieval is enabled, it fails.
With ChatGPT-4 regarding file uploads. Each time I attempt to upload a file during our conversations, I receive the following message:
“I’m sorry, but I don’t see any file uploaded in our chat. Could you please try uploading the file again? If you’re experiencing any issues with the upload, let me know so that I can assist you further.”
This is particularly puzzling as I was able to successfully share CSV files in another conversation before the last update. I am unsure why this issue is presenting itself now, whether it is a temporary glitch or a result of recent changes to the platform’s functionality???
1 Like
Hello!
The problem is related to Retrieval functionality. If you disable it the error goes away… I going to paste data directly while they fix it 
1 Like
How do you paste the data directly? Could you please elaborate?
There’s a temporary solution that involves using the OpenAI API to upload the CSV file and create the assistant programmatically. Here is a snippet of code provided by a user in the OpenAI Developer Forum that outlines the process:
( After running this code, you can go back to the Playground interface and your file and assistant should be available)
from openai import OpenAI
import os
import dotenv
dotenv.load_dotenv()
api_key = os.getenv(“OPENAI_API_KEY”)
if api_key is None:
raise ValueError(“The OPENAI_API_KEY environment variable is not set.”)
client = OpenAI(api_key=api_key)
file = client.files.create(
file=open(“put your file name here with file type at the end aka .csv”, “rb”),
purpose=‘assistants’
)
assistant = client.beta.assistants.create(
name=“Enter your assistant name here”,
description=“blah blah blah, You are great at creating beautiful data visualizations…”,
model=“gpt-4-1106-preview”,
tools=[{“type”: “code_interpreter”}],
file_ids=[file.id]
)
1 Like
@rudi.ranck Thanks! I’ll try it out
+1 I’ve been experiencing this since last night.
Hello dear GPTS Colleagues,
I found a trick. Well when you upload a file, I add at the beginning of txt file an explanation sentence to make the system think it is a plain file and not a file to be inerpreted. Of course this is not the best course of action… but it works a little bit.
On my side, with data, I decided to first transform the data into clear sentences and then it seems easier to analyse.
Hope it helps!
Actually getting only text to describe data does not seem to help cgpt to answer. it is lost into the content though it seems to work fine with regular knowledge. It is natural because the attention process can focus on some specific part of “not” redundant knowledge…
Any update on this. It’s been 2 days now and the comments here seem to have died out - but the error is persisting my side?
It’s still happening. Nothing has changed. It seems it hasn’t caught OpenAI’s attention yet, or I imagine they’d fix it.