AI
The AI this server uses to write. Works with any OpenAI-compatible provider. Keys are encrypted at rest.
Per-stage models (advanced)
Put a cheap/local model on the bulk chapters and a strong one on structure. Blank = use the model above.
Usage limit ⓘ
Optional, off by default. Only useful on a shared instance where other people generate against your API key above. Set a per-person cap; leave 0 for unlimited (a personal self-hosted copy).
Access & Login
Optional. Leave off for local self-hosting. Turn on OIDC to require sign-in on a shared instance — works with Authentik, Google, Auth0, any provider with a discovery URL.
User provisioning (advanced)
Sign-in methods
Which ways people can sign in. Turn off password login for SSO-only; if you ever lock yourself out, recover with python -m codexmill.auth_reset.
Break-glass setup token (advanced)
A recovery hatch, in case you're ever locked out of Settings (e.g. you lose the admin password, or an SSO allowlist no longer matches anyone). You set a value in the CODEXMILL_SETUP_TOKEN environment variable when you deploy; paste that same value below to unlock Settings for this session. It is never stored in the app and never put in a URL (that would leak it into server logs and browser history) — it only lives in that env var.
Shared AI
Let signed-in users generate with this server's AI (your model and key). Off means only admins can generate.
Users
root is you; admin manages settings and users; user only generates. The per-user Server AI switch appears when Shared AI is on, and picks who may use it.
Add a user
About
CodexMill is a free, open-source, self-hosted story-bible generator. It gives away what paid services charge for — the same templated LLM pipeline, in the open and running on your own machine, so nobody has to pay for it.
Set the server's AI under the AI tab, then head back to the app to generate.