Models

The Models registry decides which models users see and which provider keys back them. It's an admin surface — the user-facing endpoints only return active models.

Endpoints

GET    /api/v1/ai-models                  admin: list
GET    /api/v1/ai-models/:id              admin: fetch
POST   /api/v1/ai-models                  admin: create
PUT    /api/v1/ai-models/:id              admin: update
DELETE /api/v1/ai-models/:id              admin: delete
PUT    /api/v1/ai-models/:id/api-key      admin: set per-model API key override
GET    /api/v1/ai-models/active           authenticated user: list active models
                                          (public fields: _id, name, provider, model_key, api_type)

Model shape

{
  id: number,            // legacy sequential id
  provider: string,      // "qevron", "openai", "anthropic", "stability", "fal", "runway"
  model_key: string,     // e.g. "gpt-4", "claude-3-5-sonnet", "verinova"
  name: string,          // display name
  category: string,      // "text" | "image" | "video" (default "text")
  type: string | null,
  is_active: boolean,    // default true
  api_type: string,      // "chat" (default) | "completion"
  api_params: object,    // default max_tokens, temperature
  meta: object,
}

Per-model API key override

PUT /api/v1/ai-models/:id/api-key
{ "apiKey": "..." }

Stored in the Options table as ai-model-<id>_api-key. The text-generation service reads this key first; if absent, it falls back to the global QEVRON_API_KEY.

Enabling and disabling

Toggle is_active via PUT /api/v1/ai-models/:id. Disabled models disappear from the user-facing picker on the next reload. Posts already drafted against a disabled model still execute — disabling is a "future calls" gate.

Related

AI Overview

OAuth Credentials

Credits