What Enhanced adds: System AI configuration
Open source: each project owns its models
Open source does not provide system-level AI. Each project configures its own LLM provider; the model, key, quota, and invocation are the project's responsibility. The Studio assistant also uses the project's model configuration.
That means every project has to prepare:
- A model endpoint and API key;
- Model names, context windows, and capability information;
- Quota, billing, and usage tracking;
- A way to distribute credentials when several people collaborate.
Enhanced: the platform provides System AI
Enhanced moves model access up to the platform, so projects and users share one System AI layer:
- Shared model catalogue: the management console maintains available models, context windows, modalities, cost, and thinking levels.
- Personal AI settings: users choose a default model, reorder provider preference, and subscribe to plans or use balance.
- Usage and orders: inspect tokens, cost, latency, and first-token latency per request; manage subscriptions and recharge orders in one place.
- Management console: admins maintain the model catalogue, group visibility and price, user bindings, subscription refunds, and Store review.
- Studio assistant integration: after signing in to a project, the assistant uses System AI directly without a separate model key.
Comparison
| Capability | Open source | Enhanced |
|---|---|---|
| System-level AI credentials | None | Provided by the platform |
| Model catalogue | Maintained per project or provider | Maintained centrally in the console |
| Default model | Project configuration | Personal default → group default → project default |
| Provider preference | Not applicable | Users can order available groups |
| Subscriptions and balance | Handled per project | Plans, balance, usage, and orders |
| Management entry point | None | Models, groups, bindings, subscriptions, and Store review |
| Studio assistant | Uses the project's model configuration | Uses System AI automatically |
What stays the same
- Project code, the Workflow API, and data structures do not change.
- Projects can still configure their own LLM provider the open-source way.
- When a caller explicitly chooses a provider, personal provider preference does not override it.
- System AI models are limited to the groups an admin has opened; if there are none, personal AI settings are empty.
Who it is for
- Teams that do not want to maintain model keys in every project;
- Organizations that need one place for models, quotas, usage, and subscriptions;
- Users who want the Studio assistant to work without a separate AI login.