Knowledge base

AI TOOLS is designed for developers who want to use the Infomaniak LLM in their applications.

What is AI Tools?

AI Tools is an LLM (Large Language Model), a type of machine learning model that has been trained on a massive dataset of text and code. LLMs are capable of generating text, translating languages, writing various types of creative content, and answering questions in an informative way.

To stay true to its values, Infomaniak offers in its kSuite tools an LLM based on one of the most powerful open-source models available on the market.

This sovereign artificial intelligence can assist you, for example, with drafting, rephrasing, or translating your emails.

AI Tools allows you to connect your compatible libraries or frameworks via API, without an interface.

The API documentation can be found at https://developer.infomaniak.com/docs/api and contains information about available methods, parameters, and examples.

Getting AI Tools

To use AI Tools, you need to create an API token.

Then, to access AI Tools and for example, check its daily usage:

1. log in to the Infomaniak Manager (manager.infomaniak.com) from a web browser like Brave or Firefox
2. click on the ‍ icon at the top right of the interface (or navigate via the left sidebar menu, for example)
3. choose AI Tools (Cloud Computing universe)
   
    

Only one product can exist per organization, but multiple API tokens can be created.

AI Tools can be used by the following users:

• legal representative of the organization
• organization administrators
• organization collaborators with technical rights

Billing

• API billed on credit
• each request consumes one credit or LLM token *
• billing is accessible to legal representatives and administrators
• collaborators with technical rights have access to documentation but not invoices
• a credit card is required to start using the API
• an unpaid invoice will result in service suspension
• a spending limit can be set by a user with billing rights in the organization and on the product to control unforeseen costs
• rates are listed on the sales page

* LLM tokens (not to be confused with API tokens) are basic units used by language models to process and generate text. Each token generally represents a word, a part of a word, or an entity in a text corpus and is used for learning and generating natural language.

This guide pertains to Infomaniak products that integrate artificial intelligence, including kChat, kDrive, Mail, Site Creator, and more.

Responsibilities of the Parties

Data Provided to AI

• Infomaniak's AI is currently based by default on mixtral8x22b
  • GPT4 turbo is an optional choice for kChat / Mail users
  • Audio messages are transcribed with Whisper
  • Llama 3 is available for the AI Tools API
• When using a writing assistant, messages and requests addressed to the AI are exclusively processed in Switzerland by Infomaniak's infrastructure
• No data is processed or shared with third parties

Data Provided by AI

• The language model generates text based on the data it has assimilated during its training, lacking consciousness, intentions, or moral responsibilities
• It is crucial for the user to be aware that the information generated by the model is not always accurate and should not be considered an infallible source of information
• The user should exercise critical judgment and verify the information provided by the assistant with reliable sources when necessary
  • e.g., if a person uses the model for crucial tasks such as medical, legal, or financial advice, they should take additional precautions to confirm the information from the model
• In case of inappropriate or dangerous use of this assistant, the responsibility falls on the user who undertakes these actions

This guide concerns Infomaniak authentication tokens, or tokens, which are used to verify the identity of the application and grant specific permissions to interact securely with Infomaniak's API features and data.

Generate an Infomaniak API Token

Authentication tokens or tokens are used to verify the identity of the application and grant specific permissions to interact securely with API features and data.

To access the Infomaniak API, it is necessary to generate an access token. Its default duration is "unlimited," but you can set a custom duration upon creation.

Be aware that if no calls are made with the token for one year, it is automatically disabled. In this case, you will need to generate a new one. There is no limit to the number of tokens you can create.

During creation, you can choose the application and scope accessible by the token to match your current project needs.

Add / Retrieve a Token

API keys are managed on this page in the Manager:
 

To do this:

1. Click the blue Create a token button
2. Choose the appropriate product from the dropdown menu according to your current needs and project
3. Choose a validity period
4. Click the blue Create a token button
5. Enter your Manager login password as requested
6. Copy the token securely - it will no longer be accessible after this step

Example application: token for the Newsletter API

This guide explains how to customize the type of language used within discussions and interactions with the AI Tools Infomaniak conversational agent.

Introduction

• By default, responses and elements within a conversation with a bot in AI Tools will be of the “usual” type, the assistant is named “Infomaniak Assistant” and its description is clear:
  
• It is entirely possible to invent and extremely customize a style that must be followed to the letter by the bot during interactions
• Once the method is saved, you can reuse it and switch between different ones according to your needs

Customize the AI Tools Language Model

To create a new “system prompt” and use it:

1. log in to Infomaniak Manager (manager.infomaniak.com) from a web browser like Brave or Edge
2. click the icon at the top right of the interface (or navigate using the left sidebar, for example)
3. select AI Tools (in the Cloud Computing universe)
4. click on a language model to start configuring a conversation
   
5. click on Edit the current conversational assistant
6. click on ‍ New below the choice of the conversational assistant
   
7. give a title to the oriented conversation you are configuring
8. provide the necessary prompt for configuring your bot, detailing whether it should behave like... a terminal... a specialized translator... etc. (examples of prompts)
9. save the changes
   

You can now start a conversation and get different responses or elements based on the behavior inherent to the chosen system prompt:

It is also possible to choose a model for each conversation (learn more about models and responsibilities).

All these new settings are available in a sidebar accessible on the right side of the screen from all conversations.