AI models
The "engine" that powers the AI features. Most people never need to touch this — it just works.
Why use it
You normally do not pick anything here. The smart writing, image, and video tools simply run on whatever engine your admin has turned on. This page just explains what is happening in the background, so nothing feels mysterious.
Step by step
Use an AI feature
Write a caption, generate an image, or make a video like you normally would. You do not choose an engine first.
It picks the right one for you
Behind the scenes, Sosyabot sends your request to the matching engine — one for text, one for images, one for video. This is automatic.
You get your result
The caption, image, or video comes back. That's it — no settings to fiddle with.
If you ever see a choice
In a few advanced spots you might see a short list of options. Just leave it on the default unless someone told you otherwise.
TIP
If an AI feature is missing or greyed out, it usually means your admin has not turned that engine on yet. Just ask them — there is nothing for you to fix.
Good to know
- You do not need to understand or manage any of this; the right engine is chosen for you automatically.
- Which engines are available is decided by your admin, not by you.
- Each AI action uses credits: text 1, image 5, video 20.
- Turning an engine off only affects new requests — anything already in progress still finishes.
Related
For developers
Models live in an admin-managed registry. Each model has a provider, model_key, name, category (text | image | video), is_active flag, and api_type (chat default | completion).
User-facing surfaces only ever see active models via GET /api/v1/ai-models/active (public fields: _id, name, provider, model_key, api_type). Admins manage the full registry (list, create, update, delete) and can set a per-model API key override at PUT /api/v1/ai-models/:id/api-key, stored as ai-model-<id>_api-key in Options; if absent, the service falls back to the global QEVRON_API_KEY.
Toggling is_active hides a model from the picker on the next reload but does not stop already-drafted calls. Admin details: /en/admin/ai-models.