Manage Artificial Intelligence servers

Add or delete an Artificial Intelligence server

From the Artificial Intelligence servers table, accessible by clicking on the icon, you can add or delete one or more servers.

If you delete a server linked to a project, it will be removed from that project's configuration.

When creating a server, the Name, API provider, URL, and Model fields must be filled in. The URL is the base URL of the API endpoints of the provider.

Clicking on the ID (#) or the Name of a server opens its details page, where you can configure it fully.

You can perform several actions from the AI server consultation page:

• set or modify the server Name, API provider, URL, Model, and Description;
• set the server authentication API key;
• set provider-specific configuration parameters;
• test the configuration;
• delete the AI server using the [...] button.

SquashTM natively supports several major AI providers, with, additionally, a mechanism that allows configuration of any other provider.

The following paragraphs describe the configuration for each provider.

Configure an Artificial Intelligence server

Configuration for Anthropic

The default URL is https://api.anthropic.com/v1, this is the root URL for the API endpoints to access the models hosted by Anthropic themselves. If you use an Anthropic API hosted somewhere else, you need to change this default URL.

The models provided by Anthropic are listed here.

The API key is to be generated from your Anthropic account page.

The following request parameter can be defined:

• Timeout
  The maximum duration (in seconds) allowed for waiting for the model's answer.

The following model parameters (described on this page) can be tuned:

• Temperature
• Top-p
• Max tokens
  If unset, the default value 1024 is communicated to Anthropic.
• Thinking type
• Thinking budget tokens

Configuration for OpenAI

The default URL is https://api.openai.com/v1, this is the root URL for the API endpoints to access the models hosted by OpenAI themselves. If you use an OpenAI API hosted somewhere else, you need to change this default URL. In the case of Azure OpenAI, see the paragraph below.

The models provided by OpenAI are listed here. Be sure to use a model supporting the v1/chat/completions endpoint, this support is indicated on each model's description page.

The API key is to be generated from your OpenAI account page.

The following request parameters can be defined:

• Organization ID
• Timeout
  The maximum duration (in seconds) allowed for waiting for the model's answer.

The following model parameters (described on this page) can be tuned:

• Temperature
• Top-p
• Seed
• Max completion tokens
• Reasoning effort

Configuration for Mistral AI

The default URL is https://api.mistral.ai/v1, this is the root URL for the API endpoints to access the models hosted by Mistral AI themselves. If you use a Mistral AI API hosted somewhere else, you need to change this default URL.

The models provided by Mistral AI are listed here. Be sure to use a text-to-text model, you can check the features of each model on this page.

The API key is to be generated from your Mistral AI organization page.

The following request parameter can be defined:

• Timeout
  The maximum duration (in seconds) allowed for waiting for the model's answer.

The following model parameters (described on this page) can be tuned:

• Temperature
• Top-p
• Random seed
• Max tokens

Configuration for Azure OpenAI

The default URL is https://YOUR_RESOURCE_NAME.openai.azure.com. You must replace YOUR_RESOURCE_NAME by the name of your Microsoft Azure resource.

The models provided by Azure OpenAI are listed here.

The API key is to be generated from your resource page on Microsoft Azure Portal.

The following request parameters can be defined:

• API version
  The allowed values are 2022-12-01, 2023-05-15, 2023-06-01-preview, 2023-07-01-preview, 2024-02-01, 2024-02-15-preview, 2024-03-01-preview, 2024-04-01-preview, 2024-05-01-preview, 2024-06-01, 2024-07-01-preview, 2024-08-01-preview, 2024-09-01-preview, 2024-10-01-preview, or 2025-01-01-preview.
• Timeout
  The maximum duration (in seconds) allowed for waiting for the model's answer.

The following model parameters (described on this page) can be tuned:

• Temperature
• Top-p
• Seed
• Max tokens
• Max completion tokens

Configuration for Google Vertex AI

The URL field is unused.

The models provided by Google Vertex AI, as well as the regions where each is available, are listed here.

You must generate the JSON private key in your Google Cloud Console.
It should have this format:

{
  "type": "...",
  "project_id": "...",
  "private_key_id": "...",
  "private_key": "...",
  "client_email": "...",
  "client_id": "...",
  "auth_uri": "...",
  "token_uri": "...",
  "auth_provider_x509_cert_url": "...",
  "client_x509_cert_url": "...",
  "universe_domain": "..."
}

The following request parameters must be defined:

• Project ID
  Your Google Cloud project ID.
• Region
  The region where AI inference should take place.

The following model parameters (described on this page) can be tuned:

• Temperature
• Top-p
• Seed
• Max output tokens

Configuration of a custom AI server

The configuration of a custom AI server is described on this page.

Test the AI Server Configuration

This feature allows you to verify that the AI server is correctly configured.

You can modify these prompts for your test, but these changes will not be saved.

Expected Result

If the connection to the AI server and the generation process work correctly, the response should look similar to:

Error Display

If the request fails, an error message is displayed along with the response returned by the API provider.

Common errors include:

• an invalid or missing API key;
• an incorrect model name or endpoint;
• a timeout or network issue;
• a provider-side error.

The panel displays the raw error returned by the API to facilitate diagnosis.

For example: