Runs open models
Llama, Qwen, Mistral, vision models and more — pulled once with ollama pull and served from
your own hardware. CPU works; a GPU makes it fly.
Secretia
Your own AI, in the family’s voice.
Secretia is the local AI of the Escriba family — a private chat that runs open models on your own server. It bundles Ollama (the model engine) and Open WebUI (the chat front) into one ready-to-deploy distribution, dressed in the Escriba look and federated with Lockatus so it shares the suite’s single sign-on. No tokens billed to a third party, no prompts leaving your machine.
What it does
Section titled “What it does”100% local & private
The engine and the chat run on your server. Your conversations and documents never leave it, and there are no per-token bills from a cloud provider.
One sign-on with Lockatus
Native OIDC: the whole family logs in once through Lockatus — multi-user, 2FA and per-app roles. Optional; without it, the standalone admin login still works.
The Escriba look
A thin branding layer dresses the front in the family’s identity — the “Secretia” name, the logo and the magenta accent — without touching Open WebUI’s code.
Talks to images too
Pull a multimodal model (like Llava) and Secretia reads images alongside text — handy next to the rest of the suite’s document work.
Ready to cook
One deploy on your own infrastructure: a Docker Compose, or three tidy services on EasyPanel
(ollama + open-webui + the branding layer).
Part of the Escriba Suite
Section titled “Part of the Escriba Suite”Secretia is a self-hosted, open-source (MIT) member of the Escriba family. Like the rest of the suite it runs on your own hardware, keeps the work local, carries its own accent colour, and shares single sign-on through Lockatus. It is the place where the family’s data work meets a model you fully control.