OevuAI — the company · Caelim — the model family
AI that learns as it works.
OevuAI builds Caelim, a family of models you can talk to for free on the web, phone and desktop — or build on through an OpenAI-compatible API. New accounts start with free credit, automatically.
Products
Four ways in. Same models.
Web chat
Open a browser tab and talk. Uploads, speech, and shareable answer pages — free to start, no setup.
A real conversation. The reply carries the model tag — Caelim 2.1 Turbo.
On your phone
The same conversation, carried with you — the mobile client runs the chat you already know, and the Android app (v1.16.4) adds Caelim-2.5 on the go.
Desktop agent, watched from anywhere
A Windows agent takes your task list, works with its tool layer, and reports to a remote console you can open from any device — including your phone.
The task console. Tasks dispatch to Desktop or Cloud; a desktop code connects your machine for local files.
One iframe on your page. Widget keys are issued per deployment, so you revoke one site without touching the others.
Or drop it into your own site
The embeddable widget is the same chat, framed for your pages — with the OpenAI-compatible API behind it when you want full control.
Models
Caelim — the model family developed by OevuAI.
| Model | Role | Status |
|---|---|---|
| Caelim-2.1 Turbo | Lightweight fast tier — everyday questions, high concurrency, low latency | Online |
| Caelim-2.1 Pro | Deep-reasoning tier — complex tasks, long-form writing, deep analysis | Online |
| Caelim-2.1 Beta | Experimental tier — early access to a 3.8M-token context, lower stability | Online |
| Caelim-2.5 | Mobile-first release | Android app |
Served live from the platform's model registry; see Developers for how to list them over the API.
Research
We publish what we build.
Our first dataset is public and checksummed: 79 curated image–text pairs, 298 MB, dual-track annotation. Methods, limitations and versioning ship alongside it.
Three of the twelve public samples from Image Training Corpus v1.0.0.
Solutions
Start from what you need done.
Hand work to a desktop agent
Task lists, tool use, and remote check-ins — you stay in control.
Answer, draft, think out loud
Free chat with the Caelim models, uploads and speech included.
Build on the same endpoint
OpenAI-compatible API plus an embeddable widget.
Developers
Same interface as your existing stack.
Point your OpenAI SDK at https://oevuai.com/v1. Same tooling, same streaming, Caelim models on the other end.
# Chat completion with an OpenAI-compatible client
curl https://oevuai.com/v1/chat/completions \
-H "Authorization: Bearer $OEVAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "Caelim-2.1-Pro",
"messages": [{"role": "user", "content": "Hello"}],
"stream": true
}'Latest
What we've shipped.
Company
Built in public.
The dataset is downloadable, the methods are documented, the releases are versioned — check our work instead of trusting it. OevuAI is an independent team shipping a model family, free apps, and research you can verify.