Using from Open WebUI

Open WebUI can use Oraicle anywhere it accepts an OpenAI API: one base URL and one key, and every model in the catalogue appears in its picker.

Get a key from the Dashboard. Keys begin with oai-.

In the interface

Settings → Admin Settings → Connections → OpenAI API, then:

Field	Value
API Base URL	https://api.oraicle.me/v1
API Key	your Oraicle key (oai-…)

Save, then reload. Oraicle’s models appear in the model picker at the top of a new chat.

A full model list does not mean the key works

Our /v1/models is public — it serves the catalogue and prices to anyone, so Open WebUI can fill its picker with an empty or mistyped key and look perfectly healthy. The first message you send is the real test. If the list loads but every chat fails, the key is what to check.

With Docker

Set the same two values as environment variables:

docker run -d -p 3000:8080 \
  -e OPENAI_API_BASE_URL=https://api.oraicle.me/v1 \
  -e OPENAI_API_KEY=oai-... \
  -v open-webui:/app/backend/data \
  --name open-webui \
  ghcr.io/open-webui/open-webui:main

Open http://localhost:3000. The first account created is the administrator.

Using Oraicle alongside another provider

Open WebUI accepts several OpenAI-compatible connections at once, so Oraicle can sit next to whatever you already run — a local Ollama, or another hosted API. Add it as an additional connection rather than replacing the existing one; the picker shows the models from all of them together.

What it costs

Every message is billed per token from your prepaid wallet — no subscription, nothing per seat, and nothing to cancel. A new account starts with $0.25 of credit, which is enough to try the interface before paying anything.

Open WebUI sends the whole conversation with each message, so a long chat costs more per reply than a short one. That is true of any OpenAI-compatible backend; it is worth knowing when a thread runs long. Your balance and a per-call history are in the console.

What will not work yet

Leave the RAG embedding engine alone

Open WebUI’s document features need an embeddings endpoint, and ours is not live — there is no /v1/embeddings to point them at. Leave the embedding engine on its default local model. Setting it to OpenAI with Oraicle’s URL will fail when you upload a document. See What’s live and what’s next.

The same goes for image generation and speech: those are configured separately in Open WebUI and are not part of the Oraicle API today. The image oracle is something a model calls during an answer, not an endpoint Open WebUI’s image settings can use.

Live data

Oraicle’s oracles are switched on with a tools entry of the shape {"type": "weather"} — an extension to the OpenAI request that Open WebUI’s chat box does not send. So a plain chat through Open WebUI answers from the model’s own knowledge, the same as any other OpenAI-compatible backend.

To give a model live data, call the API directly or through a client that can set the request body — see Live data (oracles) and the LiteLLM page for one way to pass it.