Skip to main content
Version: 1.49

Creating an external deployment with Mistral

Create an Mistral Deployment using the External Deployment type. In this article, you will find out how to integrate the Mistral Nemo model in Deeploy. In this example we will focus on deploying an Mistral agent, but next to an agent also the le Chat model and embeddings can be integrated in Deeploy.

Prerequisites​

Deploy an agent using Mistral​

We use the Mistral Nemo model that is developed by Mistral and available on their platform. In order to create an external Deployment in Deeploy you need the following:

  • The model endpoint from a deployed model
  • The API key to communicate with the model

The model inference endpoint and API key​

To get a Mistral Nemo inference endpoint, follow these steps:

  • On Le Plateforme, create a new agent and API key
  • select the Mistral Nemo model, configure the temperature, system prompt and few-shot prompts and click deploy
  • Navigate to agents and copy/save the API id
  • You need the https://api.mistral.ai/v1/agent/completions endpoint in combination with the API id and API Key to inference the agent

Connecting Mistral Nemo in Deeploy​

Connect the Mistral Nemo to Deeploy via the UI or locally with the Python client

Complete the following steps in the Deeploy UI

  1. Log in
  2. Navigate to the Workspace where you want to add gpt-4o as an external deployment
  3. Click Create and select External
  4. Follow the steps, in step 3 you are expected to add:
    • The retrieved inference endpoint (e.g., https://api.mistral.ai/v1/agent/completions)
    • Select Bearer as authentication method
    • Paste the API Key in the password field
  5. Once the connection check is successful, continue and complete the Deployment creation process