Get 40% offSee the offer

0.9.0 betaDocumentation of the Community Edition, copied on .

Settings#

Two screens hold the settings: the personal page /settings, and Settings > LLM Configuration (/settings/llm).

Settings (/settings)#

Open the menu under your name at the top right and choose "Profile, password and API keys". It is open to every signed-in person; administrators see more tabs.

Tab Who What it holds
Profile Everyone Full name, country and organisation can be changed and saved. The e-mail is shown and cannot be changed. Below it, a Password card: current password, new password, confirmation. A change signs out the other devices.
Governance Administrators The governance mode, read from the server. See remediation.md.
API Keys Administrators Generate a key, copy it, delete it. The key is shown once, at generation. See api.md.
Mail Administrators The mail relay: host, port, username, password, from address. Save, remove, and send a test message. See mail.md.
AI Usage Administrators The budget (tokens per month, tokens per run) and the usage by period. See ai.md.

Changing the governance mode, the mail relay or the AI budget is recorded: see audit.md.

LLM Configuration (/settings/llm)#

Administrators only; the menu entry is hidden from everyone else. It is open in the Community Edition.

  • Model: what answers now, and a form for each of two models, the good one and the small one: provider, model, address of the provider, API key. Save, Test and Remove. A model can be tested only after it is saved. The model is off by default, and what is sent to it is described in models.md and ai.md.
  • AI Execution Speed: how fast the agents run, with custom request and retry delays in milliseconds. A change applies to all users at once.
  • Cost tiering: the good model handles chat, ownership decisions and resources above the cost threshold; the small one the rest. With no small model saved, the good model answers everything.

A change of model applies to new requests; jobs already running finish on the previous model. In the Community Edition nothing is carried out: see remediation.md.