Getting up and running with Umbraco.AI.Agent.Copilot

Questions
1

I know I know - the package says -alpha1 but I was still wondering if following the Winter Keynote, anyone has managed to spin up Umbraco.AI.Agent.Copilot and get it working with reponses?

I am trying to get it working specifically with a Microsoft AI Foundry connection

Under the connection I have set my endpoint https://redacted.services.ai.azure.com/ and my AI Foundy API Key - both copied from the “Keys and Endpoints” tab in my Foundry in Azure

I have created a Profile called Default Profile, picked my connection and set the model to GPT 5 Nano 2025 08 07

I have created an agent called Default Copilot Agent

When I open Umbraco Copilot, I can chose the agent and write a prompt

image
image1270×624 33.3 KB

Getting the exception

Azure.RequestFailedException: Unknown model: gpt-5-nano-2025-08-07
Status: 400 (Bad Request)
ErrorCode: unknown_model

Content:
{"error":{"code":"unknown_model","message":"Unknown model: gpt-5-nano-2025-08-07","details":"Unknown model: gpt-5-nano-2025-08-07"}}

In my Foundry I have deployed “gpt-5-nano”

image
image590×881 40.6 KB

And I was just wondering if anyone has managed to get it up and running and if I’m missing something obvious? :slight_smile:

2

Hi Kasper
Well, I managed to get it to work… :smiley: I found - some models are better at some actions than others - one model was better at doing web fetches - another for looking up media… This is very much one of the things we need to fix for it to go beyond alpha :smiley:
In Azure yeah, you need to deploy your own models - so its a little harder to setup. We successfully set all of the providers up - for the demo however, I didn’t actually use the Azure ones, so I can’t say that for sure… but from your exception it seems to be more of an Azure thing than a copilot thing,

2 Likes
3

Managed to get it up and running - turned out I needed to make a Foundry project in the Preview mode, then it just worked!

I got chat up and running and it -thinks- it’s doing a good job

image
image1270×704 54.8 KB

Very much looking forward to playing more with this

1 Like
4

Yeah when the value is not a plain string it works allot better when it already has a value. Also something to fix ofc

So my guess is if you manually set a value, save and try again it will work

1 Like
5

Just a headsup I had to actually make my Foundry model name gpt-5-mini-2025-08-07 to get it to work - as it didn’t appear in the drop-down as a “deployed” model.

1 Like
7

Just for reference :slight_smile: Insert Response does not work with TipTap · Issue #37 · umbraco/Umbraco.AI · GitHub