---
title: "Set up Elara with a hosted or local model | Oplema"
description: "Connect Elara to a hosted model, your own API key (BYOK) or a local model such as Ollama, and see what each option costs and how to fix common failures."
url: "https://oplema.com/docs/elara/setup"
updated: "2026-11-04"
---

# Set up Elara

Elara can use a hosted model, your own API key or a model on your own machine. All three need you to be signed in, and all three store your conversation with Oplema.

## Use a hosted model

The model button at the top of the Elara panel

1. In the Elara tab, select the model button at the top of the panel.
2. In the menu that opens, point to a provider: Oplema, OpenAI, Anthropic, DeepSeek, Z.ai / GLM, xAI / Grok, Mistral, Google Gemini or MiniMax.
3. Choose a model. Each one is marked Lowest cost, Medium cost or High cost.

The default is Elara Flame 1.0, under Oplema. Hosted models spend your usage. With no subscription or on Starter, a high-cost model shows the note "High-cost model: your usage balance may run out quickly."

## Use your own key (BYOK)

This works with any service that has an OpenAI-compatible API, meaning a base URL and an API key that accept `POST /chat/completions`. Your connection also needs `GET /models`, because that fills the model list.

1. In the menu that opens, select BYOK.
2. Open Add connection.
3. Enter the base URL in the field that shows `https://…/v1`, ending before `/chat/completions`. It must start with `https://`, for example `https://api.example.com/v1`.
4. Enter your key in API key and select Add.
5. Pick it from the Select connection… list, then pick a model from the list that shows Select model….

To remove a connection, select it, then select the Remove connection button. Your provider bills you for requests, and they use no Oplema usage.

## Use a local model

1. Start your model server on your computer.
2. In the menu that opens, select Local.
3. Open Add connection and enter the URL in the field, such as `http://localhost:11434/v1`. It must be on `localhost` or `127.0.0.1`. There's no key.
4. Select Add, then pick a model from the list that shows Select model….

Your browser calls the model directly, so the model must be running on the same computer as your browser.

### Ollama and browser access

Browsers block calls to a local server unless the server allows the page's address. Ollama calls this setting `OLLAMA_ORIGINS`. Add the address you open Oplema from, then restart Ollama. Ollama's [FAQ](https://docs.ollama.com/faq) has the steps for each operating system.

Details depend on your model server. Check its documentation if the connection fails.

## If something goes wrong

| What you see | What to do |
| --- | --- |
| "Select an available model connection to continue." | Pick or add a connection under BYOK or Local, or go back with Hosted models. |
| "The selected model is unavailable. Select another model." | Your connection no longer lists that model. Choose another. |
| `external endpoints must use https` | BYOK URLs must use `https://`. |
| `local connections require a localhost base URL` | Local URLs must be on your own computer. |
| Elara replies but never edits | Use a model that supports tool calling. Elara needs it for every edit. |
| A dialog asks "Switch to …?" | You changed model in an existing chat. Your next message sends the chat history to the new model, and cost may rise. Select Switch and send or Cancel. |

### Ask for a flow. Elara builds it.

Elara is Oplema's AI assistant. Ask it to change a flow or a dashboard.

[Try Elara in Oplema](https://oplema.com/editor?path=%2F%3Fopen%3Dstrata) [About Elara](https://oplema.com/elara)
