My OpenAI assistant via make don't use the existing uploaded files

I have created an assistant using OpenAI. I’ve provided him some data (personal files that i would it to use for generating response).

I can use the correct prompt and everything directly from the playground but it doesn’t work from Make.

I’ve tried

  • to use a thread that got the info, no more success
  • to link the file using the ID, name, failed too
  • to activate “Knowledge”

I’ve noticed that everytime i test my call to my assistant it creates a new empty vector store on OpenAI, and seen that’s it’s a new feature from April.

Has anyone made this work using assistant, Make and file stored and linked to the assistant ?
Or maybe, there is some more development to be made on this particular action ?
Thanks in advance

1 Like

Welcome to the Make community!

It is definitely possible. I’ve created a Telegram assistant bot with GPT-4-turbo that can read uploaded files. Not only that, it can retrieve relevant previously-uploaded files to the assistant, based on keywords in the chat message (Make filter).

Screenshot_2024-04-02_220445

Here is an example of it in action:

Screenshot_2024-04-02_220423

For more information, see

samliewrequest private consultation

4 Likes

What is your configuration on both side (Make and OpenAI ) to make sure your assistant will look into the files already uploaded ?

1 Like

Got the same issue as here : Problem of Retrieving Information from File Search, Playground Vs. API Usage - #7 by fsobarzo - API - OpenAI Developer Forum

Any though on API version that’s been used by Make ?

1 Like

I am also seeing the same issue it appears that the document you provided is empty and I am unable to retrieve the information directly from it. If you have another document or can provide more details, please upload it or specify the information, and I will assist you accordingly."

This was not happening last week. Last week the Make Scenerio connecting to my OpenAI assistant was working. Could the Make module for OpenAI possibly not be updated to talk to OpenAI Assistant v2 latest version?

I am using the setup with the assistant selected, this week I’ve added the File ID and selected under tools to use “Knowledge Retrieval”. Last week I didn’t have to select the File ID or select Knowledge Retrieval. Even with this setup, it doesn’t work when running through Make Scenario; however, in the OpenAI playground using the same assistant it does retrieve files using the exact same prompt message.

I was thinking of it too.

I’ve test the API using python code and can use my assistant using the files stored. Anyone at Make got more information on OpenAI plugin supporting the latest version ?

1 Like

@moulii I’m experiencing the same thing. When I call the Assistant from Playground it works as expected, producing results from ‘File search’ - When I call from Make.com, I get responses that it can’t find any files. (e.g. “It looks like there aren’t any documents…”)

It seems obvious that Make hasn’t updated the OpenAI module since the recent Assistants V2 changes

In the Make.com UI, in ‘Message an Assistant’ > Tools (Select All, Knowledge Retrieval, Code Interpreter), I would expect to see something that reflects the changes, maybe something like ‘File store’ as an option under Tools.

The OAI migration guide says that if we are still using the v1 of the API, it should still work until the end of 2024.

That means existing scenarios using files (knowledge retrieval) shouldn’t just break yet, and there is no rush to upgrade immediately.

1 Like

Samliew, what about a case where the Assistant was created as a v2 on GPT4-Turbo using File search?

Retrieval and File search are not the same. If you switch a v2 Assistant to v1, the vector store from the File search is not magically “included.”

That’s a great idea! I’m sure others would agree that an OpenAI Assistants v2 integration is a nice feature to have.

You can submit this suggestion to the Idea exchange, under App improvement ideas.

Don’t forget to search for it first, just in case someone already suggested it, so that you don’t end up creating a duplicate.
Screenshot_2024-01-17_130153

You may also want to contact support to expedite this request.

For technical issues like this, directly contacting support can often lead to a faster resolution. You can open a new ticket, or if you are unable to login for some reason, you also can reach support using the contact form on the website.

1 Like

Anyone got any info about this ? or made it work maybe ?

I’ve been trying for hours, there’s no workaround for this. Only thing that will fix it is the make team updating their module to v2. It’s been like this for weeks, surely it can’t be that hard to change the module to openai assistant v2?

1 Like

It’s never that simple :slight_smile:

Anyone at make team can give us some insights on this ?

Hey @here :wave:

I wanted to share some news regarding this update. As you can see in our Idea Exchange post, this update is currently planned. I’ll be keeping you in the loop providing updates on our progress.

Thanks for your patience and understanding! :pray:

3 Likes

Hello Vendi. Please see how many people have written under this post, but no one has yet been able to answer when the update will be ready, many people are waiting for it. Could you answer this question or somehow influence the acceleration of this update?

Hello everyone @here

I just want to confirm that this is unfortunately a known issue of the “4o” model. Our dev team is aware of this and is working on updating our ChatGPT module to the latest API version.

At this moment, there is no specific ETA we could share but we will let you know once the update is rolled out.

Thank you very much for your patience and understanding. :pray:

As a temporary workaround, you could use an older assistant model or use the “Make an API Call” module with the custom API call to OpenAI API with the “OpenAI-Beta: assistants=v2” header.
The below guide might provide general help on how to use the “Make an API Call” modules.

:arrow_right: How to make an API call: tutorial

2 Likes

We just updated the OpenAi Assistant module to support GPT4-o and functions. Give it a try.

You can check this post for more details: Open AI Assistant Module with GPT-4o and functions