Running Snotra on Melious, inside the EU

I have connected Snotra to Melious for the first time. I did it for one reason. When a folder holds documents about real people or real customers, “where does this text go?” is a question I want to answer with one word, and with Melious that word is Europe.

Melious is an inference provider that routes requests to hosts in several EU countries and, by its own account, never falls back to a data centre in the US or Asia. It offers a data processing agreement for the GDPR. I can’t check from my desk where a request really ends up, and neither can you. But a promise in a contract is something a company can be held to, and that is more than most model providers put in writing.

Step 1: Get an API key at Melious

Snotra has no Melious template yet, and it doesn’t need one. Melious speaks the OpenAI API, so all Snotra needs is an address and a key.

1. Create an account at melious.ai.
2. Open the API keys page with Get API Key in the top navigation, or go straight to melious.ai/account/api/keys.
3. Create a new key and copy it. It starts with sk-mel-. Keep it somewhere safe until it is in Snotra, since you need it in the next step.

Step 2: Add the model in Snotra

The screenshots in Set up your first model show the same dialog, if you have never added a model before.

1. Open the settings with Snotra AI › Settings… on macOS, View › Settings… on Windows and Linux, or Cmd+, / Ctrl+,. They open on Models.
2. Click Add model. Under Provider choose OpenAI-compatible, under Template choose Custom endpoint.
3. Enter https://api.melious.ai/v1 as Server URL and paste your key into API key.
4. Leave API style on Chat Completions only. Give the entry a Display name you will recognise in the chat, such as Melious.
5. Click Load models. Snotra asks Melious which models your key can use, and the list fills. Pick GLM 5.3 Flash.
6. Click Apply in the dialog, then Apply at the bottom of the settings.

The new entry now shows up in the chat’s model menu. Choose it for a chat, and that chat’s requests go to Melious. Like every key in Snotra, the Melious key is stored encrypted in your operating system’s user profile and only goes to the address it was saved for. I set this up in Snotra 1.29.0, and tools and permissions work the same as with any other model.

Step 3: Enter the prices by hand

Snotra estimates what a chat costs from the public price table that LiteLLM keeps on GitHub. Melious isn’t in that table, so the new entry has no price, and without one the cost estimate and the usage report have nothing to count. A price of your own fixes that, and it always wins over the table.

1. Look up the prices of GLM 5.3 Flash on the Melious model hub, for input and output tokens.
2. In Snotra’s settings, stay on Models and scroll down to Prices for the cost estimate. The Melious entry says No price known.
3. Click Set price next to it and fill in Input and Output. Snotra expects US dollars per million tokens, so if Melious shows euros, convert them first. The two cache fields can stay empty, and then cost what input costs.
4. Click Save price. The entry now says Your price.

Pressing Update prices later fetches a newer LiteLLM table but leaves your own price alone. If Melious changes its prices, edit the entry with Edit price.

The part I didn’t expect

I had picked Melious for where it runs, not for how fast. Then GLM 5.3 Flash answered before I had properly looked away from the window.

Most of what I hand to an agent during a day is small. Rename a few files, or pull the open points out of a document. With the large models I had got used to waiting for that, and for a task that takes me a sentence to describe, the wait is often what makes me do it myself instead. GLM 5.3 Flash works through those tasks so quickly that the waiting is gone, and the answers have been good enough that I haven’t missed a bigger model for them. I haven’t measured it in tokens per second, so take this as an impression. It is a clear one.

For long planning chats or a tricky refactoring I still pick a larger model for that chat. For everything else Melious with GLM 5.3 Flash is now my default. Snotra keeps the files on my computer, Melious keeps the requests in Europe, and together they are the best setup I have had for everyday work so far.

Snotra has no arrangement with Melious, by the way. I pay for the tokens like anyone else.

Your turn

If you try it, I would like to hear how it goes, in the Discussions or as an issue. Which EU provider should get its own template in Snotra’s list next to OpenRouter?