Setting up an AI Provider

This page explains how to configure the AI provider required to use Agentiqs.

Configure an AI Provider

There are several ways to make an AI provider available.

Setup method	Details
Import an AI Provider	Import settings shared by an administrator
Connect with Entra ID authentication	Use Azure OpenAI / Microsoft Foundry with a company account
Create an AI Provider	Create provider settings by entering the connection information yourself
Use a local LLM	Use an LLM that runs on your own PC or server, such as Ollama / LM Studio

Please check with the Agentiqs administrator in your organization for the exact setup method.

Import an AI Provider

This section explains how to import an AI provider.

Procedure

1. Click [Settings] > [Sharing] > [Share Settings] > [Import].
   • The [Import Settings] dialog appears.
2. Enter the following data received from your administrator.
   • Passkey
   • Export Data
3. Click [Import] in the [Import Settings] dialog.
   • When the settings are imported successfully, the AI provider import results are displayed.

Connect with Entra ID authentication

When you use Entra ID authentication with an Azure OpenAI or Microsoft Foundry AI provider, you can connect to AI resources with your company account without distributing or configuring API keys individually.

Procedure

1. Import an Entra ID-compatible AI provider by following Import an AI Provider.
2. From [Settings] > [Organization], initiate the connection.
   
3. Enter the client ID provided by your administrator and connect.
4. If the browser asks you to authenticate, sign in with your company Microsoft account.
   • If successful: the screen below is displayed. Close the browser.
     
   • If it fails: report the displayed error message to your administrator.
5. Return to Agentiqs and confirm that the Entra ID connection status shows as connected.

Before you start

• The existing API-key-based method can still be used.
• Depending on your organization, Entra ID settings on the organization side and permission settings on Azure may be required.
• Signing out disconnects the connection. To use it again, reconnect using the same procedure.
• Switching Azure directories is not supported. The organization in the home directory is used.

Create an AI Provider

If you want to configure it yourself by entering the connection information, create an AI provider.

Procedure

• Refer to Administrator's Guide > Create an AI Provider.

Use a local LLM

You can use a local LLM running on your PC or server as an AI provider.
Because data does not leave your PC or server, this is suitable for sensitive use cases and for testing without incurring API charges.

• Supports chat (streaming), agent tool calls, knowledge search (embeddings), and image input, and can be used with the same workflow as an AI provider.
• For supported local LLMs, see System Requirements - Supported LLMs.

Procedure

1. Install a local LLM and obtain the model you want to use (for example: ollama pull llama3.1:8b).
2. Go to [Settings] > [AI Provider] > [Add], then select the target provider from the gallery.
3. Enter the endpoint on the [API] tab (an API key is usually not required).
   Example: Ollama: http://localhost:11434, LM Studio: http://localhost:1234
4. On the [Model] tab, edit the model definition to match the model you obtained.
5. Use [Test connection] to verify connectivity, then save.

Notes when using a local LLM

• What you can achieve in Agentiqs depends on your local environment (GPU / memory) and the model you use.
  • Tool call quality, image input quality, response quality, speed, and more.
  • Some features, such as image generation, web search, and the code interpreter, are not available.
• Usage is counted as $0.

Configure the models you use

The models registered in an AI provider are configured in different locations depending on the purpose.

Setting	Target	Use case
Purpose-specific models	Models used for chat, summarization, sub-agents, and image generation	AI generation and processing
Search model	Embedding model	Search knowledge and chat history

AI provider priority and model roles

AI providers can be reordered by drag and drop.
The order is reflected in the list display, the model selection list, and the default selection at runtime.

• In [Model usage settings (model roles)], you can review and change the models used for chat, summarization, sub-agents, and image generation.
• Assigning models that are appropriate for each task helps optimize model selection.

Purpose	Description
Default	The standard model for chat
Summarization	The model used to summarize and compress conversation history
Sub-agent	The model used when running sub-agents
Image generation	The model used to generate images

Procedure

1. Open [Settings] > [AI Provider].
2. Drag the handle (⋮⋮) at the upper left of the provider card to reorder priority. (Figure ①)
3. In [Model usage settings], review the model for each purpose and select one manually if needed.
   You can return to automatic selection by choosing [Return to automatic]. (Figure ②)

If the specified model is disabled or deleted

A replacement model is selected automatically and you are notified with a ⚠ icon. (Figure ③)

Search model for knowledge and chat history

Select the embedding model used to search knowledge and chat history.

Procedure

1. Click [Settings] > [AI Provider].
   • The AI provider selection screen is displayed.
2. Select the target model from [Model] and save the changes.

Create an AI Provider

Create an AI provider using the same procedure as the administrator.

Procedure

• Refer to Administrator's Guide > Create and Distribute AI Providers.