Nextcloud Assistant (integration_openai)

Nextcloud's AI features run through the integration_openai app. It connects any number of OpenAI-compatible services, and an admin picks the models each service exposes for each task.

Upstream: source, docs.

Plugboard options

Give Nextcloud its own key. vision handles image analysis and OCR, so it must accept images. Docker needs a reachable address (Serving other machines).

{
  "options": {
    "listen": { "host": "0.0.0.0", "port": 4010 },
    "keys": { "nextcloud": "{env:PLUGBOARD_KEY_NEXTCLOUD}" },
    "models": ["opencode-go/glm-5.3-flash", "anthropic/claude-sonnet-4-6"],
    "aliases": {
      "text": "opencode-go/glm-5.3-flash",
      "vision": "anthropic/claude-sonnet-4-6"
    },
    "apps": { "nextcloud": { "allowedModels": ["text", "vision"] } }
  }
}

Connect

In the admin settings, open Artificial intelligence and add a service:

Field	Value
Base URL	http://host.docker.internal:4010/v1 (the address your Nextcloud server reaches)
API key	the nextcloud key
Models	text and vision for the text modality. Leave the image, transcription and speech modalities off

Skip personal keys unless you want per-user quotas.

What works

• Text tasks: free prompt, chat, summarize, headline, topics, reformulate, change tone, proofread, translate and the rest of the text task list.
• Chat with tools. Plugboard returns the tool calls and Nextcloud runs them (Tool calls).
• Image analysis and OCR, through the vision alias.

Limits

• Image generation, transcription and text to speech are not served. Use another service in the same app for them.
• Run 4 background job workers on the Nextcloud server so AI tasks do not wait on each other.
• Nextcloud's quotas are per service. Plugboard's limits and budgets apply separately.