Skip to main content
An AI provider is an external AI service that NekoHub calls when running Asset Skills. Skills use AI providers to automatically generate metadata for your images — for example, producing a caption or descriptive alt text from an uploaded photo.
AI providers are optional. If no active AI provider is configured, Asset Skills that require AI will not run. You can use NekoHub fully without setting up an AI provider — skills simply won’t produce AI-generated enrichments until one is active.

AI provider profiles

An AI provider profile stores the connection details for one AI service. You create and manage profiles from the /ai-providers page in the admin console or via the /api/v1/system/ai/providers API. Each profile contains:
FieldDescription
nameA label for this profile (e.g. OpenAI GPT-4o).
apiBaseUrlThe base URL of the AI service’s API (e.g. https://api.openai.com).
apiKeyThe API key used to authenticate with the AI service.
modelNameThe specific model to use for skill requests (e.g. gpt-4o).
defaultSystemPromptAn optional system prompt prepended to every skill request sent through this profile.
isActiveWhether this profile is the currently active provider.

Active provider

Only one AI provider profile can be active at a time. When a skill runs, it uses whichever profile is currently marked active. You can switch the active profile at any time — the change takes effect on the next skill execution. If no profile is active, skill runs that require AI will fail gracefully without affecting the asset itself.

Testing a profile

Before marking a profile active, you can test it from the admin console. The test sends a small request to the configured AI service and confirms that the credentials, model name, and endpoint are working correctly. A successful test returns a sample caption generated by the model.

AI providers and Asset Skills

AI provider profiles power the Asset Skills feature. When you run a skill on an asset — either manually from the admin console or automatically on upload — NekoHub sends the image to the active AI provider and stores the result as a derivative or structured result on the asset. Common skill outputs include:
  • Image captions stored as structured results with kind: caption
  • Processed image variants stored as asset derivatives
You can view all available skills and their descriptions from the admin console’s asset detail page or via GET /api/v1/assets/skills.