Is there a list available somewhere, or a reference, of what types of information a field agent can read? For example:
When I create a field agent that says “go search the internet for information about the company” (name of company in a company field) it works beautifully.
It can also read text in another field (in the same record)
But I am having a LOT of trouble getting it to read Google Docs (if the link is provided) or .pdf attachments (in another field).
I do a lot of work with venture capital, and I want to create an agent that pulls a company name, and then goes out to Pitchbook to find company funding history and return that result to Airtable. I have a Pitchbook enterprise account via SSO. But I cannot get this to work… The agent just says it cannot access that information even though I'm logged into Pitchbook in the same browser instance (in another tab).
Can a field agent summarize the text of an interview if it’s an .mp3 or .mp4 file? Do I have to upload the entire audio file to Airtable, or can I point it to a link?
Understanding the spatial limitations on how this work would be VERY HELPFUL!
Thanks in advance.
Page 1 / 1
I’m no field agent expert, just learning with everybody else. Knowing what skills the agents have would be very useful. Maybe everyone is still figuring it out lol.
I’d have better results letting Omni write the prompt for me. My use case is analyzing pdf (or doc) attachments to extract specific info. My own prompts were generating summaries and eloquent analysis. What I needed was very specific transposition of data. I let omni rewrite my 3 sentence prompt into a 3 paragraph prompt that did what I needed it to do.
Soon the machine will prompt ME to do a task, I’ll do it without thinking, and the robot takeover will be complete!
Hi @neildkane,
I can imagine google docs will be harder for the field agent to analyze since it could have permissions or it might not be in a file format like pdf that it can easily read its data. I do see there is a new google drive feature on the field agents that you might be able to use to give access to your google docs. For PDF’s it should work for the most part, would you be able to share the prompt you are using?
For your pitchbook question, unfortunately even if you are logged in to it in your browser, the field agent doesn’t use your browser, but instead it fetches the website from its own internal server, so it wont have any of your credentials.
For the mp3 or mp4 files, I’ve tried this and the field agent doesn’t detect the file saved in your attachment field, looks like currently it only works for pdfs, images or text files. Has anyone has any luck with any sort of audio file?
Hey @neildkane,
As mentioned above, it is pretty much PDF. That being said, I have a couple of use cases for which I will be building a [input format] to PDF soon. In my case, these are mostly images which are taken whilst submitting a form.
This can be done using Zapier, Make, or N8N (more on these automation tools here). Basically, when a record gets created, attachment (image in my case) will be pushed to pdf.co via my automation, to get back a pdf. Then I’d update the record and replace the original attachment with its pdf version. Then the Airtable AI fields will be doing the rest.
