Is it possible to use Azure AI Fountry through AI Assitant or any other way?
Answered
Hey, my company finally arrived on the AI hype train and now I'm sitting on an API URL and an API key via Azure AI Foundry. But it seems the Jetbrains AI Assitant doesn't support this. Is there a way I can use my newfound access to accelerate coding directly in the IDE?
I've seen plugins like ProxyAI but it would be quite sad to have to use 3rd party plugins for something that seems to exists for other providers.
Please sign in to leave a comment.
Hello,
At the moment, using your own API key or account for LLM models isn’t supported with individual subscriptions. This option is currently available only for organizations through the AI Enterprise offering.
But we have OpenAPI for local LLMs: LLM-15339 Third party providers: use OpenAI API for all
And there is a request about BYOK: LLM-11585 Bring your own LLM API key for individual/personal license of AI Pro – please vote and follow for updates.
Apologies for the inconvenience.
Possibly related to https://intellij-support.jetbrains.com/hc/en-us/community/posts/17750680977938-Intellij-AI-Assistant-Self-Host-in-Azure-Cloud
Update: I found a way to do it via LiteLLM (server) but it's quite wonky, especially since IntelliJ keeps defaulting to their own servers each time I stop LiteLLM for a second. :|
Hi, I created an account specifically to comment on this because I ran into the same issue and I have found a potential workaround/solution :)
I added my Azure endpoint in Settings > Tools > AI Assistant > Providers & API Keys > 3rd Party Providers
Provider: OpenAI-compatible.
Regarding the URL, I initially wanted to use the original Azure endpoint of my deployment + API key, but it didn’t work (connection error)
URL: https://[your_deployment].cognitiveservices.azure.com/openai/responses?api-version=2025-04-01-preview
So, I modified it to look like this:
URL: https://[your_deployment].cognitiveservices.azure.com/openai/v1
API Key: (not sharing mine :) )
When I tested the connection, it was successful. :)
However, when I went to the chat and clicked the combobox to select a model, I got a list of all models in Foundry.
Unfortunately, I couldn’t find any option to specify a particular deployment.
Suspecting a problem I typed a simple message like "Hi" in the chat, and of course, I received an error saying that no deployment was found.
I started digging into the idea.log.
To get detailed Azure OpenAI request logs in PyCharm, open Help > Diagnostic Tools > Debug Log Settings and add:
#com.intellij.ml.llm
#ai.koog
After analyzing the logs, I discovered that PyCharm uses default deployment names.
However, all my deployments have custom (non-default) names.
To test, I added a 5.2 Codex model with the exact name that PyCharm seemed to expect (if model has a date in name deploy it in the same way)
Things improved slightly! :)
But now, I encountered a new error:
"The requested operation is unsupported."
After further analysis of the logs, I figured out that the issue lies with Codex models.
I suspect that Codex deployments fail in because those models does not support the specific chat/streaming operations that JetBrains AI Assistant is attempting to use.
Models like gpt-4o or gpt-5.6-sol (those I have quickly tested), works perfectly. :)
Cheers!