Appearance
AI Designer
Two different things
- AI that helps you design (the Studio assistant): see The AI assistant.
- AI inside your application: agents and chat flows you build.
Available today
- AI provider settings (Studio, AI provider settings): choose the model each profile uses for your tenant, and add your own API key on the API keys tab. With your key every AI feature in the tenant (the Studio assistant, Generate with AI, agents) runs on your account. The key is encrypted when saved and never shown again, only its last four characters. Without a key the platform's key is used, if the platform has one. A self-hosted provider such as Ollama needs no key.
- Agent Designer: an agent has a goal, instructions, the tools it may call, and an autonomy level. A gate enforces what it may do without approval, with delegated authorization (it acts with the user's rights, not more) and an audit of every run.
- Tool calling: agents call registered tools (platform operations).
- Chat flows: designed conversations.
- Knowledge (retrieval) for agents that answer from your documents.
- Generate with AI on lists (pages, entities, workflows, themes).
Step by step: a helper agent
- AI provider settings: confirm a provider and model.
- Agent Designer, New: name, purpose, instructions.
- Choose tools (read-only first), set autonomy to ask before acting.
- Try it in the test panel; read the audit entry.
- Publish and give a role permission to use it.
Planned (placeholders)
Structured-output schemas as a Studio setting, per-tenant token limits and cost controls dashboards, an AI step in the workflow designer, embeddings management screens. Planned.
Security
Human approval for risky actions, permissions per agent, audit of every action. Never give an agent more tools than its task needs.