As part of one of our projects, we need to send an image taken by the user of the Glide app to Gemini.
The available Google Cloud Vision and Google Cloud integrations that allow the use of GCP services seem incomplete.
This is not just about extracting text from the image as provided by the Google Cloud Vision integration. Our requirement is to use Gemini’s multimodal mode.
We have tested using a Cloud Function to encapsulate the call to Gemini, but the authentication part seems complex.
Do you know if there are future development perspectives for the Google Cloud Vision and Google Cloud integrations to support other tools (Cloud Run calls, Cloud Function calls, calls to Gemini, Google Workflows calls, etc.)?
Can anybody guide us on using APIs with OAuth2 by utilizing a Google service account in Glide ?
Thank you.very much to the community !
Interesting use case. I wonder if it’s just for extracting info from the image, would you be able to use OpenAI for that? The built-in OpenAI integration allows you to add images I believe.
Google integrations are complex to set up as far as I aware. It used to take me multiple hours.
Thanks. I work with Houy-Sy on this case.
No, openAI is not an option.
To avoid complex Google integration, we used make (former integromat) and it works.
But we can’t get the image content.
While using the image field, we can get the GCS link, but not the content.
As it’s also a GCP authentification with JWT or Oauth2, it seems to be endless story
yes.
we would like to integrate glide in several projects with google cloud functions, google workflows, gemini, GCS, and other GCP services.
This one with multimodal gemini is the first one and will be use as POC to see if we can do what we wants.
Only Oauth2…
As we already have Make (integromat) in our scope, so, we manage : glide calls a make webhook with a header param, make catch the call, verify this param and send the payload to GCP service.
But as I’m really new to Glide, I can’t find any way to send base64 image content instead of the URL to the image.
I think you have to download the file first with a HTTP module, then convert it to base64 like this.
They have a built-in base64 function, but not sure if it works with your case.
Yeah unless you have that image as part of a form, then you can do an on-submit action.
You can use a Trigger Webhook action, doesn’t have to be a Call API action. As long as you’re having the image URL and the rowID to edit later, I think it should work.