Upload file to chat

[JorisBeen]

Hi There,
I am trying to create a simple example. I just upload a simple docx word document with the text: this is just a test file.
when i upload this within the chat (and i think i am handling it correct...) i always get the response:

I’m sorry, but I don’t have access to the contents of “testfile.docx.” If you can provide the text or the specific portion you’d like me to read, I’ll be happy to help.

I need to know what i do wrong....

This code i use to upload a file, in this case i use the purpose "assistants"

 public async Task<FileResponse> UploadFileToOpenAI(OpenAIClient openAIClient, string fileName, string purpose, Stream? getFile)
 {
     OpenAI.Files.FileUploadRequest fileUploadRequest = new OpenAI.Files.FileUploadRequest(getFile, fileName, purpose);
     var fileResponse = await openAIClient.FilesEndpoint.UploadFileAsync(fileUploadRequest);
     return fileResponse;
 }

this goes succesfull and i can also see the file on the openai playground in the storage with the file.id.

then when i create the message and add the message to the chat i this:

//remove the markdown text from the message to the assistant.
var message = RemoveMarkdownImagesAndFiles(assistantChatRequest.Message);

string messageContent = string.IsNullOrEmpty(assistantChatRequest.PromptMessage)
                         ? $"users question: {message}"
                         : $"{assistantChatRequest.PromptMessage}, users question: {message}";

List<Attachment> attachments = new();
var contents = new List<Content>
    {
        new Content(messageContent),

    };

imageFileResponses.ForEach(x =>
{
    contents.Add(new Content(new ImageFile(x.Id, ImageDetail.High)));
});

// Add supported files
fileSearchResponses.ForEach(x =>
{
    attachments.Add(new Attachment(x.Id, Tool.FileSearch));
});

return (contents, attachments);

And then this to send the message:

// Create message content and send the message
var messageContent = await CreateMessageContent(referenceId, token, assistantChatRequest);
var messageRequest = new OpenAI.Threads.Message(messageContent.content, Role.User, messageContent.attachments);

await _openAIclient.ThreadsEndpoint.CreateMessageAsync(assistantChatRequest.ThreadId, messageRequest);

// Create a new run for processing the message
var runRequest = new CreateRunRequest(assistantChatRequest.AssistantId);
// Optionally, pass an action if you want to process server-sent events
Func<IServerSentEvent, Task> action = null;
var run = await _openAIclient.ThreadsEndpoint.CreateRunAsync(assistantChatRequest.ThreadId, runRequest, action);

// Return the run ID as the task identifier
return (run.Id, null);

When i send a picture with this, it is able to read the image using the purpose as vision.

please help me to get reading files in the chat.

Kind regards,
Joris

[StephenHodgson]

You're not the only one that has problems with this API endpoint, it is a little difficult to use even for the official python/js library users.

Best bet is to try to do the same thing in the API playground, write out the steps, and try to recreate in code.

I think your best bet is to add the file directly to the assistant itself, but I have not used this feature much. I mostly use realtime.

[JorisBeen]

I will close the discussion as it seems to be working, just like i have described here above.
is only working fine for gpt-4-turbo and the o1 models. the others are not working properly...
could reproduce it in the playground

Thanks for the reply!

[StephenHodgson]

Ok so it only works correctly for O series and gpt-4 series models?