I use a dedicated code server for my nextcloud setup. Code run in a debian VM and is up to date.
However I don’t see the AI options, to enable them and choose provider like in the Desktop client.
I’ve added enable.ai in my coolwsd.xml file but it doesn’t change anything. What am I doing wrong to enable the embedded AI features of CODE 26.04? Not talking of Nextcloud assistant or anything external to Collabora.
After enabling it, there are two ways to configure the provider:
Per-user provider: Keep allow_user_settings set to true and leave api_url, api_key, and model empty. After restarting Collabora Online, users should be able to configure their own AI provider from the settings dialog.
Central provider: Configure api_url and model (and api_key if required). The endpoint needs to provide an OpenAI-compatible /v1/chat/completions API. This can be a provider such as Ollama, vLLM, LM Studio, OpenAI, etc. You can set allow_user_settings to false if you want to prevent users from changing the provider settings.
Once the configuration is changed, restart coolwsd.
Also, the AI sidebar will only appear when a provider is actually configured. In other words, all of these need to be true:
ai.enabled is set to true.
A base URL and model are configured, either by the user or centrally.
The user is logged in. AI is not available for anonymous users accessing a public share link.
The integration does not disable AI through CheckFileInfo.
For Nextcloud, there should be nothing additional required on the richdocuments side, as it does not send DisableAISettings.
Also, enable.ai is not a valid configuration key, so it will simply be ignored.
If you can share your current <ai> section from coolwsd.xml (with the API key removed), we can take a look at the configuration and see if anything is missing.
Thanks it worked, I had found these ai.enable instructions over the internet. I sometimes get a HTTP 404 answer when chatting (with Mistral), but it seems to work though