How to configure the new n8n AI Assistant Code Sandbox (Service URL & API Key) on Cloudron?

[firmansi]

With the recent n8n upgrade that introduces the n8n AI Assistant feature, I am looking to set it up on my Cloudron instance.

However, I am a bit confused about the required environment variables or configurations, specifically the Service URL and API Key for the AI Assistant.

Could anyone advise on the following:

• Is this AI Assistant service automatically deployed and configured within the Cloudron n8n Docker package?
• If not, what are the correct steps or environment variables needed to configure the Service URL and API Key manually in the Cloudron dashboard?

Any guidance or examples from anyone who has successfully set this up would be greatly appreciated!

Thanks in advance!

[james]

Hello @firmansi

I will have to look into this.
A quick read of https://docs.n8n.io/deploy/host-n8n/configure-n8n/set-up-ai-assistant makes me think this might need an extra service.

[robi]

It's for plugging in an external LLM via those variables, as in API URL and API KEY

[firmansi]

It got me a bit confused since the documentation mention that it's already included in recent version, then i assume that the service already up and running in cloudron n8n, but even it's not there yet in latest n8n cloudron update, please advise how we can do manually in terminal

[robi]

Did you read the docs above?

Use the .env file

[firmansi]

@robi already done that, that's why i asked the question about the Service URL and API Key, since what i understand from the documentation that if it's already included in the instalation then the service url and api key already there in written in .env, but if i have to add manually then what should i fill in the .env configuration

[andreasdueren]

@james yes looks like you need to implement a sandbox environment in the package

[firmansi]

@james is there any update if this possible in N8N package in Cloudron, there's also a new development in N8N that there's a feature called Agent, but I also don't see how this can be done in Cloudron

[james]

Hello @firmansi
I still need to look into this.

[firmansi]

@james hai, is there any update regarding to this, is it doable with Cloudron?

[firmansi]

@james I am so sorry for asking you this again, is the Sandbox already built in in recent N8n update in Cloudron

[james]

Hello @firmansi
There are currently other issues at hand.

[james]

The sandbox can't run inside the app: upstream's runner is privileged Docker-in-Docker.
You can use Daytona (hosted), or run the n8n sandbox service on a separate Docker host and point n8n at it.

Env example for Daytona:

export N8N_INSTANCE_AI_MODEL=anthropic/claude-opus-4-8
export N8N_INSTANCE_AI_MODEL_API_KEY=<llm-api-key>
export N8N_INSTANCE_AI_SANDBOX_ENABLED=true
export N8N_INSTANCE_AI_SANDBOX_PROVIDER=daytona
export DAYTONA_API_URL=https://app.daytona.io/api
export DAYTONA_API_KEY=<daytona-api-key>

[robi]

@james You could if using sysbox runtime.

It gives Docker the ability to run things it cannot with the default runtime and shifts the container toward the more isolated VM side.

Enables DinD too.

Docker acquired Nestybox, makers of sysbox.

[firmansi]

@robi is it mature stack? or relatively new one

[robi]

@firmansi Yes, was mature before the acquisition. Feel free to search the forum for 'sysbox' to see the trail.