Optional AI

Slite supports AI-assisted short-code creation and social-preview generation through xsai. The AI integration connects to any standard OpenAI-compatible API (such as OpenAI, DeepSeek, Ollama, or local inference engines).

AI is entirely optional; normal short-link creation and routing do not depend on it.

Enabling AI

AI is disabled by default. Both NUXT_AI_BASE_URL and NUXT_AI_MODEL must be explicitly configured:

NUXT_AI_BASE_URL=https://api.openai.com/v1
NUXT_AI_MODEL=gpt-4o-mini
NUXT_AI_API_KEY=sk-...

Until both variables are populated, AI endpoints return HTTP 501 immediately without initiating any outbound network requests. Restart the container after changing configuration.

• Base URL: Must include the provider's API path prefix where required (e.g. /v1).
• Model: Must be a model identifier supported by that provider.
• API Key: NUXT_AI_API_KEY may remain empty when connecting to self-hosted or local providers (e.g. Ollama) that do not require authentication.

Supported endpoints

• GET /api/link/ai?url=... — suggests concise, memorable short codes based on the destination URL passed via query.
• GET /api/link/og-ai?url=...&locale=... — suggests OpenGraph titles and descriptions based on the destination URL passed via query. Supports an optional locale query parameter to specify the preferred response language.

If the upstream model call fails or times out, Slite falls back automatically to URL-derived heuristics rather than failing your link creation flow.

Data privacy and boundary

URL-only transmission

Slite only transmits the destination URL you supply to the configured AI provider. Slite never scrapes, downloads, or forwards third-party webpage content to the AI model.

Before enabling AI, review the terms, data retention policies, and pricing of your chosen model provider. Never submit URLs containing embedded private tokens or sensitive parameters.