Setting up a Dedicated AI

This page explains how to customize Agentiqs' AI to suit your needs.
By setting up knowledge and agents (roles), you can make the AI's answers more accurate and practical.

---

Creating Knowledge

FAQs, user manuals, internal specifications, databases, etc., can be turned into knowledge so that the AI ​​can refer to them when answering questions.

This explains the procedure for creating knowledge in Agentiqs.

Procedure

1. Click the [Knowledge] > [Create New] button.
   • The knowledge base is created and you will be taken to the settings screen.
2. Enter the required information for the settings parameters and click the [Save] button.
   • The knowledge base settings will be saved.

Settings Parameters

Basic Settings

[Overview]

• Title: Set the title of the knowledge base. This will be the key when using the knowledge base from the chat, so please enter a title that matches the content of the knowledge base.
• Description: Set the description of the knowledge base. This will be the key when using the knowledge base from the chat, so please enter a description that matches the content of the knowledge base.

[Index] > [File Settings]

• Target File Extensions: You can specify the files to be included in the knowledge base. If you specify extensions separated by commas, only files with those extensions will be included.
• Exclusion Patterns: You can specify files to ignore. Files and folders that match the rules specified in glob format will not be included in the knowledge base.

[Index] > [Search Settings]

• Search Limit (Number of Results): Specifies the maximum number of search results to display. (Default: 30)
• Relevance Threshold: Specifies the minimum relevance value for search results. A value of 0.7 or higher will broadly capture semantically similar results. (Default: 0.7)

[Data]

• Root Directory: The path registered with [Add Data] will be a relative path from the root directory. Even if the clone destination directory differs for each user, adjusting the root directory will reduce data inconsistencies and make index updates function more smoothly.
• Add Data: Specify a file or folder to add data to be included in the knowledge base.

Files to be converted into knowledge

The following document formats can be imported as knowledge (filtering by file extension is also possible):

• Text files: .txt, .md, etc.
• Office documents: .docx, .pptx, .xlsx
• PDF files
• Source code files (any language)
• Next Design projects: .nproj

About Knowledge Generation Fees

• The AI ​​model usage fee is incurred when you select the data to be converted into knowledge using [Add Data].
  • The fee varies depending on the AI ​​model used and the amount of data imported.
    • Estimate (example): Approximately 1 yen/MB (when using AI model: text-embedding-ada-002)
  • The AI ​​model used is automatically the model defined first in the AI ​​provider.

Regarding Index Updates

• If you replace specifications or other documents that have been turned into knowledge base items with the latest version, the changes may not be reflected in the knowledge base search results until you execute [Index Update].
• Hover your mouse cursor over the knowledge base item you want to update in the knowledge base list and execute [Index Update].
  • Even with index updates, there may be AI model usage fees.

Creating an AI Agent

By defining roles, tone of voice, tools, and knowledge base items in advance, you can obtain consistent answers without having to fine-tune prompts each time.

This explains the procedure for creating an AI agent in Agentiqs.

Procedure

1. Click the [Agent] > [Create New] button (Figure ①→② above).
   • The [Agent Gallery] will be displayed.
2. Select the agent to use as a base.
   • The agent settings screen will be displayed.

You can also create an agent by chatting interactively with the AI.

Click the [Create with AI] button (Figure ③ above). The chat will appear on the right side of the screen.

AI Agent Settings (Basic Settings)

The key settings for configuring an AI agent are as follows:

• System Prompt
  • This is the instruction text passed to the AI ​​when the chat starts.
  • Define the agent's role, response policy, and output format (e.g., bullet points, template).
• Knowledge
  • Select the knowledge the agent will refer to from the pre-created knowledge base.
  • For instructions on creating knowledge bases, see "Create Knowledge Base".
• Tools
  • Select the tools the agent will use from the pre-created tools.
  • To create a new tool, see "Integrate with External Tools".

Regarding the Tools Used

• Next Design Functions is primarily intended for use in workflows.
  Adding it to an AI agent may increase unintended calls and thus increase usage costs.

AI Agent Settings (Advanced Settings)

Context

• This setting is intended to inject dynamic information into prompts before LLM interactions.

• At this stage, it is assumed that the default settings will be used as is.

Tool - Automatic Approval

• This setting determines whether confirmation (approval) is required before executing a tool from the chat.

• ON (automatic approval) is generally recommended. If OFF, approval will be requested each time the tool is executed, and the user must approve each time.

• Consider OFF when handling operations with significant impact, such as file creation/update, or when you want to confirm the AI's judgment before proceeding.

Sub-agent

• This setting is for executing LLM interactions that combine multiple agents.
  This is useful when you need to combine various areas of expertise.

• Use the [Workflow] function when predictability and reproducibility of results are important.

Experimental Implementation - Planning Mode

• When LLM thinks, it uses calls that are adjusted to first consider the execution plan in more detail and over a longer period.