AI Agents

AI Agents - AI partners that help you streamline every task

AI Agents on TypingMind act as experts in various industries. It is designed to perform tasks and provide insights within specific domains by following your instructions and custom data.

Why build an AI Agent on TypingMind?

Developing AI Agents on TypingMind provides a huge advantage in terms of customization and management:

• Access 100+ pre-built AI Agents in different industries from technical, marketing, design, entertainment, etc.
• Develop your own AI Agent collection that is specialized in your domain and share access with all team members
• Control member access to certain AI Agents

Utilize our 100+ pre-built AI agents

The default AI agents are a set of pre-built AI agents from TypingMind.

Based on your specific needs, you have the option to allow your members to use them or not as follows:

1. Go to AI Agents
2. Turn on / off the option “Show built-in AI Agents” to control the visibility of these AI agents on the Chat interface.

If you turn on this option, users will be able to add the AI agents from the default list.

In case you don’t want your users to be distracted by these AI agents, you can switch the toggle to OFF mode. It will disappear from the Chat UI.

Build your AI Agent collection

You can create exclusive built-in AI Agents for your team members so they can quickly access the resources without manually creating them from the chat interface.

• Go to AI Agents from the Admin Panel
• Click “Create AI Agent”

Your created AI agents will be displayed on the chat interface as follows:

You can also connect knowledge base or training files to your AI Agent. Learn more

What you can do with TypingMind's AI Agent?

1. Assigned a based chat model with custom parameters

You can assign your AI agent the following:

• A specific chat model
• Custom Parameters like Temperature, Top_p, Max tokens, etc.

Toggle the Model & parameters under the Base model section to assign models and custom parameters.

This allows the AI Agent to always perform best with its custom model and settings.

2. Set skills for AI Agent with plugins and text-to-speech

You can assign your AI Agent with:

• Specific plugins, for example, Web Search
• A custom voice with text-to-speech

Toggle the Plugins or Text-to-speech under the Skills section to assign plugins or a voice for AI Agent.

3. Set knowledge for AI Agent

You have three options to connect your training files or knowledge base for the AI Agents:

• Upload training files directly—upload your PDF, XLSX, CSV, TXT files, etc., to the AI Agents.
• Access to your connected knowledge base via the Knowledge Base section in the Admin Panel. Learn how: Connect knowledge base to your AI Agents
• Use Dynamic Context via API - this allows you to retrieve content from an API and inject into the system prompt. This can be used to add live information to the AI or implement Retrieval-Augmented Generation (RAG) from your own data sources (e.g., vector store database).

Toggle the Training files, Knowledge Base Access or Dynamic Context under the Knowledge section to set up the knowledge for AI Agent

Here are some important notes for training files:

• Assign a model to your AI Agent before adding training files.
• Your uploaded documents will be included directly in the model's system instructions.
• Make sure your document fits within the context length of the chosen model.

4. Use Few-shot prompting

Few-shot prompting is a technique to help the AI agent learn how to respond to users in a specific way by providing demonstrations in the prompt to steer the model to better performance.

The prompts will be automatically inserted at the beginning of every conversation, right after the system instructions (but not included in the system instruction). This is useful when you want the AI to always respond in a very specific format.

5. Improve the chat experience with Welcome messages and Conversation starters

The Welcome message allows the AI model to send the first greeting message when you choose to interact with it. This ensures you have a more engaging and personalized experience from the start.

The Conversation starters give you some ideas / suggestions on the first message you should ask the AI Agent.

6. Control user access and usage for each AI Agent

User access to the AI Agents will be controlled using tags. Each user can be given a tag, and you can use these tags to restrict usage and access:

• Control AI Agent visibility: decide if the AI agent should be visible to a specific group of users or not.
• Control number of messages and characters users can send while using the AI Agent: restrict usage to ensure fair usage and control cost more efficiently.

Here's how to set up:

Step 1: Assign tags for your members

Follow these steps to assign tags to your members:

1. Log into your chat instance's Admin Panel.
2. Navigate to the User list under the User Management
3. Identify the member you wish to assign tags to by their email and click on them to Update tag
4. Enter the tag you want to assign them and click Save

Please note: You can assign multiple tags to each member. Learn more about Set tags for users

Step 2: Control visibility

Here’s how to set tags for each AI agent:

• Within the AI Agent editor, scroll down to the Visibility section and select Visible only to users with tags from the drop-down list
• The User tags section shows up, enter tags that you have just set for specific members at step 1

Example: If you want only members with the 'marketing' tag to access the Marketing Expert AI agent, add the "marketing" tag to that agent.

Only the members with the “marketing” tag will see “Marketing Expert” on the chat interface.

Step 3: Control usage

Here's how to control the number of messages and characters users can send while using the AI Agent:

• Within the AI Agent editor, scroll down to the Usage limit section
• Create a limit group where you can set up message and character limit for users while using the AI Agent and Save
• In the Apply For Users section, select User with tags, enter tags that you have just set for specific members at step 1

How to develop a good AI agent

A good AI Agent can help you generate relevant and high-quality responses.

There are some specific key points you will need to look at to customize a great agent that best suits your goal.

• Define the AI's Role: determine the tone and manner of the AI's responses. If you're developing an AI for the healthcare industry, you might define its role as a "fitness advisor." This ensures ChatGPT generates more accurate responses based on healthy and fitness data
• Set up system instruction: the system instruction forms the base for all the AI's responses. Ensure it aligns with the role you've defined.
• Add few-shot prompting: few-shot prompting (also called training examples) guide how the AI should respond to different inputs. Moreover, the accuracy and relevance of the AI's responses greatly depend on the quality of the training examples.
• Connect to your custom knowledge base: integrating the custom knowledge base allows the AI Agent to access specific information relevant to its role, including proprietary databases, industry-specific knowledge, or real-time data sources.