Skip to content
More support
Contact us
  • There are no suggestions because the search field is empty.

Connect Langdock to Open-Weight Models on Impossible Cloud Inference

This article explains how to add an open-weight model served through Impossible Cloud as a custom model in Langdock, and how to use it with plugins such as Microsoft 365.

Interactive demo

The demo below walks through the full setup and a short example with Microsoft 365. The steps that follow describe the same configuration in text form.

 

Prerequisites

  • A Langdock workspace with admin rights
  • An Impossible Cloud inference account and API key
  • The model ID you want to use, from the Impossible Cloud model catalogue

Step 1: Get your endpoint URL, API key and model ID

You can find your endpoint URL in the Impossible Cloud console, and create an API key there. The model IDs are listed in the Impossible Cloud model catalogue.

You need three values for the configuration in Langdock:

  • Base URL: the OpenAI-compatible endpoint, ending in /v1, for example https://eu-west-3.inference.impossibleapi.net/v1
  • API key: your Impossible Cloud inference API key
  • Model ID: the exact ID from the model catalogue, for example qwen3.8-27b

Step 2: Add a custom model in Langdock

  1. Open your account menu at the bottom left and click Settings.
  2. Under Workspace, click Models.
  3. Click Add custom model.
  4. In the Choose from configurations dialog, click set up manually.

Step 3: Configure the model

On the Configure your model page, keep the Completion tab selected and enter the following:

  1. Model Name: a display name your team will recognize in the model picker, for example Impossible Cloud - Qwen 3.8 - 27B.
  2. Provider Model Name: the exact model ID from the Impossible Cloud model catalogue, for example qwen3.8-27b.
  3. Provider: the organization that built the model. If it is not in the list, select Other.
  4. Region: select EU.
  5. Description (EN) and Description (DE): optional. A short description helps your team understand what the model is for.
  6. Expand Configure and enter the Context Window Size from the model catalogue.
  7. Expand Pricing and enter the Input token price and Output token price per 1M tokens from the model catalogue. Langdock uses these values for cost tracking and spend limits.
  8. Click Continue.

Step 4: Add the deployment

  1. Under API Key, click New Key.
  2. In the Create new API key dialog, enter the following:
    • Key name: an internal name, for example Impossible Cloud Inference Key. End users do not see this name.
    • Base URL: your Impossible Cloud endpoint URL, including /v1.
    • API Key: your Impossible Cloud API key. Enter the key only, without a Bearer prefix.
    • Hosting Provider: replace the default with Impossible Cloud. Users see this name in the model list.
  3. Click Save key.
  4. In Model Name / ID, enter the model ID again, for example qwen3.8-27b.
  5. Click Test & continue to confirm that Langdock can reach Impossible Cloud.
  6. Review the configuration and click Save model.

The model now appears in the model list under Other, marked as hosted by Impossible Cloud.

 

Note: The API key is stored once in the workspace. All users who select the model use this key, and they never see it.

Step 5 (optional): Set the model as your workspace default

By default, Langdock selects a model automatically for each chat. To use the Impossible Cloud model for every new chat:

  1. On the Models page, open the Default chat model dropdown.
  2. Select the model you configured.

Members can still override the default in their personal settings.

Step 6: Connect your data sources as plugins

Plugins let the model read data from your existing tools and act on it. This example uses Microsoft 365.

  1. Click Back to app, then open Plugins in the left sidebar.
  2. Install the plugins you need, for example Outlook Email, Outlook Calendar and OneDrive.
  3. Sign in with your Microsoft account when prompted.

Plugins use your own permissions, so the model can only access data you can already access.

Step 7: Start working

  1. Start a new chat and check that your Impossible Cloud model is selected in the model picker.
  2. Type @ in the chat input to add a plugin to your request, for example Outlook Email.
  3. Send your request. When the model wants to change data in a connected tool, for example to create an email draft, Langdock asks you to confirm the action first.