Issue with attachment field in agent integration

Hey ​@Matt,

Would you be open to sharing one “delinquent” file by providing a URL where it can be downloaded? Also sharing a prompt for the agent would help. I’d be happy to see if I can duplicate/work around your issue here. (But that’s assuming the data can be shared).

Steve

Hm the only images I’ve had issues with are .webps.  Could you create an automation to output the filetypes of the attachments giving you issues?

As a weird bonus fact, if you rename a .webp file to .png the field agent will be able to read it so it seems like it’s just blocking the webp extension

Hey ​@Matt and ​@TheTimeSavingCo,

It is interesting to see that this is surprising to both of you. In my experience, I don’t think I was ever able to have an agent read an image or file other than pdf. (I am not a heavy user of this specific field agent).

It seems that you have a different experience? It seems that you were actually able to run it on images!

If helpful, what I had to build was a Make (or n8n automation) that would pass my image to a pdf.co module/node to get it converted into a pdf. (I know, crazy, right?).

Having said that, this approach might be useful for any edge case you stumble upon. Feel free to grab a slot using this link if you’d like me to show you around or help build this specific automation (more context below).
 

Mike, Consultant @ Automatic Nation 
YouTube Channel

I was curious, so I ran a few tests…

I used one of Airtable’s Google ads (pulled from the Ads Transparency Center) as an example.
(https://adstransparency.google.com/?region=US&domain=airtable.com)

• The AI field agent was able to extract text just fine from JPG and PNG.
• WebP displays correctly in the gallery, though the field agent still flags it as “unsupported.”
• AVIF isn’t showing up yet—hopefully it’s on Airtable’s roadmap.

Hi guys, 
I will available an example when its happen with me:
https://airtable.com/app7g55xekREbv0tm/shr5di2MbkUR1BtRW

Hi ​@Matt,

I downloaded your files and tested them in my base—they work fine both with an “Analyze Attachment” field and with my own field agent (which also analyzes the images, just using a different approach).

A couple of things you could try:

1. Create a new “Analyze Attachment” field. What you have should work, but this will rule out any odd Airtable bug.

2. Build a second field agent. Click Edit with AI and try a simple prompt like “Extract the text from the attachment.” Airtable will expand that into a stronger prompt automatically.

I also checked your images—they look fine, not corrupted.

Steve

Hi ​@Matt , were you able to resolve the issue with the file type not supported? Please let me know. I am facing a similar issue with the field agent Analyze Attachment or Attachment Summary. 

​@Stephen_Juth  Please let me know if you were able to resolve it with using field agents in airtable

​@SonaliK my reply above mentioned successfully using field agents for this use case. But let me know if you are referring to something else I had said. Thanks.

Hi ​@Stephen_Juth , I am unable to resolve it with using field agents in airtable.  I have a simple table with two fields with field agents - Analyze Attachment and Attachment summary, but both do not work. The Three attachments are .png, .pdf and .jpg. https://airtable.com/appMnHbLFegjHAspl/shrZ36Ld43Tisnnsj

​@SonaliK Hmm… I’m not sure why you are having issues. I downloaded/uploaded your three images (rows 6-8), and my field agents were able to view the files.

​@SonaliK Hmm… I’m not sure why you are having issues. I downloaded/uploaded your three images (rows 6-8), and my field agents were able to view the files.

 

 

Thanks ​@Stephen_Juth , this is hopeful. Do you have a paid account? I had raised a query with Airtable support and I received this response : 

“ I’m sorry for the trouble you’ve been having while trying to use the image generation or image analysis features. These features you may have seen in our marketing materials are not yet live, but they are part of a powerful set of upcoming capabilities we’re actively developing within our new Omni interface. These features will enable users to generate visual content and extract insights from images directly within Airtable, greatly expanding what’s possible with AI-powered workflows.

While these image-related features aren’t available just yet, we’re working hard to bring them to you soon and ensure they meet our quality and performance standards. Once released, they’ll represent a significant leap forward in how Airtable can support creative and analytical use cases alike.

In the meantime, we recommend keeping an eye on our What’s New Page, where we’ll announce availability as soon as these features are ready for general use.”

If you are using a paid account or having some thing for Omni activated, perhaps that is reason it works for you?

​@SonaliK I am using a paid account (team plan). I also have AI features turned on for my main workspace.

​@Stephen_Juth  Thanks for your responses. Airtable support team have confirmed that these features are not release yet and may be availed if the early access is turned on.

Thanks again. 

​@SonaliK Ahh, right! I’m glad Airtable support had pointed this out. I turned this on a while ago and forgot about this setting.