AI features, subscriptions and BYOK¶
openkoutsi's AI features — activity analysis, daily training status, and AI plan/workout generation — normally run against the model your instance's administrator has configured. Some instances charge for AI that uses the server's own model, while keeping everything else free. This page explains what changes on those instances and how to check your own status.
The rest of openkoutsi stays free
Subscription gating only affects AI features that use the instance's model. Uploading activities, syncing with Strava/Wahoo, fitness metrics, plans and everything else keep working exactly as before.
What changes on a gated instance¶
When your administrator turns the gate on, using the instance's AI model requires an AI-access subscription (an entitlement granted to your account). If you don't have one, you have two options:
- Bring your own AI model (BYOK). Point openkoutsi at your own OpenAI-compatible endpoint under Settings → AI / LLM. BYOK always works, even on a gated instance, because the request goes to your provider, not the instance's — see Using your own AI model.
- Subscribe. Ask your administrator for access. When you try to use an AI feature without a subscription (and without BYOK), openkoutsi shows an "AI features on this server require a subscription" message instead of running.
Checking your status¶
Settings → AI / LLM shows a line describing your current AI access:
| You see | What it means |
|---|---|
| Included | The instance doesn't gate AI — everything just works. |
| Subscription active (optionally until …) | You hold an AI-access entitlement; the instance model is available. |
| Using your own AI server | You've set up BYOK; your own model is used and no subscription is needed. |
| Not available — subscribe or add your own server | The instance gates AI and you have neither a subscription nor BYOK. |
Auto-analysis on a gated instance
If you enabled automatic activity analysis or daily training status but don't have access, those toggles stay saved but do nothing — the settings page explains why. They resume automatically once you subscribe or set up BYOK.
For administrators¶
The gate is off by default, so a fresh self-hosted instance behaves exactly as before until you enable it.
Enabling the gate¶
In the admin console → instance settings, turn on Require subscription for instance LLM. From then on, only users with an AI-access entitlement (or their own BYOK model) can use the instance's AI features.
Granting and revoking access¶
In the admin console → users, each user has an LLM access control. Grant access (optionally with an expiry date), or revoke it later. Granting is manual in this phase — you decide who gets access, e.g. after payment arranged out of band.
Reading usage stats¶
The LLM usage card aggregates how many tokens each user has consumed on the instance's model, with input and output tokens shown separately and a per-provider breakdown. Pick a period (with week/month granularity) to see tokens per user per month — useful for working out the average AI cost per user. Calls made through users' own models (BYOK) are never counted, since the instance pays nothing for them.
For what each AI feature sends to the model and where it goes, see Your data & AI.