We’re currently experiencing a persistent issue with uploading files to the Assistants API. At the moment, we are unable to process any files uploaded by users.
{
"file_ids":["file-8NFEV7zB79V14Z41nqg9qUue"],
"content":"___Uploaded a file named 'dummy-category-data.csv'___",
"role":"user"
}
The response we received in the API request is ok, but the chat message says:
We continue to encounter persistent technical issues, unfortunately making it impossible to access and read the files you have uploaded. Under normal circumstances, I would recommend contacting customer support for assistance with this matter. Since we are operating within a simulated environment and that option is not available, we are at an impasse.
Are we experiencing a platform issue? How can I get it working again? It was functioning well in the past weeks, but it has stopped working.
Even after migrating the SDK and the payload to Assistant v2, the error persists:
{
"attachments": [
{
"file_id": "file-8NFEV7zB79V14Z41nqg9qUue",
"tools": [{ "type": "code_interpreter" }]
}
]
"content":"___Uploaded a file named 'dummy-category-data.csv'___",
"role":"user"
}
I apologize for the inconvenience, but it seems there continues to be an issue accessing the file, despite having the correct file information. This appears to be a technical problem beyond my control, and I am currently unable to access the file to review the content and suggest categories.
In order to move forward, we may need to seek assistance from the technical support team to resolve this file access issue or try re-uploading the file again to see if that resolves the problem.
You might want to paste instructions and parameters into a new assistant, one with no vector data store, and interact with code interpreter. Then, there will be no confusion about files interjected with different methods. When its code interpreter can process your CSV from the mount point, then you can consider separately what knowledge would be useful in a vector store. all created thru beta v2
Hi Jay, thanks for the quick response! As you can see below, the assistant doesn’t have vector data enabled. This error is intermittent but it’s now happening more frequently, making testing impossible. As the chat message suggest it may be a platform issue
{
created_at:1712076479,
description:null,
id:"asst_45pF...",
instructions:"You are an expert in ...",
metadata:{},
model:"gpt-4-1106-preview",
name:"60993c98-f321-44e7-aaf0-5af152187b7f",
object:"assistant",
response_format:"auto",
temperature:1,
tool_resources:{
code_interpreter:{file_ids:[]},
file_search:null
},
tools:[
{
type:"code_interpreter"
},
{
function:{
description:"Create categorie rules ...",
name:"create_category_rule",
parameters:{
properties:{
category_name:{
description:"The name of the category for which you want to create rules.",
type:"string"
},
required:["category_name"],
type:"object"
}
}
type:"function",
}
}
],
top_p:1
}
I ran into the same frustrating issue many of you have faced with file uploads in the Assistants API. Despite seeing the files in the Playground and everything working fine there, the API kept telling me no files were uploaded when I tried accessing them via Make API (formerly Integromat).
Here’s a Quick Breakdown of the Problem:
Files uploaded perfectly and were visible in the Playground.
The Assistant in the Playground handled the files without any issue.
However, accessing these files through the API returned an error: “It appears that no files have been uploaded yet.”
The Workaround That Worked for Me: I managed to get around this by setting up a vectorstore from the uploaded file data. Including both the file_ids and the vectorstore in my API calls did the trick!
