Enroutia in your chat app

The three things every app asks for

Whichever app you pick, it is always the same three. Create the key in the panel and give it a monthly limit while you are there — a chat app on your own machine gets used more than you expect.

One warning about the dropdown: you will also see `whisper`, `embed` and the `speak-*` voices. Those are not chat models — they are transcription, embeddings and speech — and they show up because `/v1/models` returns the whole catalogue, as OpenAI's own does, and the app cannot filter it. Picking one by mistake breaks nothing: a chat model answers and you are billed that model's price, not the one you asked for. For chatting, stay on `auto` or any of the chat models.

Base URL:  https://api.enroutia.com/v1
API key:  vk_live_…
Model : auto

LibreChat (self-hosted, on your server)

The closest thing to ChatGPT you can run on your own machine: multi-user, with agents and search. It connects through a configuration file, with no code to write.

1. Copy the file below as `librechat.yaml`, next to your `docker-compose.override.yml`.
2. Put your key in your `.env` as `ENROUTIA_API_KEY`. The YAML carries the variable name, never the key itself.
3. Mount the YAML into the container (`- ./librechat.yaml:/app/librechat.yaml`) and start LibreChat as usual.
4. “Enroutia” appears in the model selector with the whole catalogue — read from `/v1/models`, so it never goes stale — and new conversations open on `auto`.

# librechat.yaml
version: "1.3.13"

cache: true

endpoints:
  custom:
    - name: "Enroutia"
      apiKey: "${ENROUTIA_API_KEY}"
      baseURL: "https://api.enroutia.com/v1"
      models:
        fetch: true
        default: ["auto"]
      titleConvo: true
      titleModel: "gpt-oss-20b"
      summarize: false
      modelDisplayLabel: "Enroutia"

And in your `.env`, the one line that is ours:

# .env
ENROUTIA_API_KEY=vk_live_…

One detail that costs money: LibreChat names each conversation with a second call you never see, and it is billed like any other. The preset points that one at the cheapest model in the catalogue, so naming an answer never costs more than the answer did.

Jan (desktop app for Mac, Windows and Linux)

It runs on your own computer and handles both local models and remote providers with your own key. This is the one to pick if what you want is an icon in the Dock.

1. Open Jan → Settings → Model Providers → Add Provider.
2. Enter the Base URL and key from above. Jan loads the catalogue by itself.
3. Pick `auto` as the model and start typing.

Open WebUI

Configured through environment variables, with no config file:

OPENAI_API_BASE_URL=https://api.enroutia.com/v1
OPENAI_API_KEY=vk_live_…

What it costs, and what we keep

Calls from these applications are ordinary calls: your prices, your balance, and your usage in the panel with the same per-key detail as everything else. If you want to know what chat costs you next to your automations, give the app a key of its own.

Your conversations are kept by the application, on your server or on your computer. We keep neither the questions nor the answers — only the billing metadata, exactly as with any other call.