Getting Started

Overview

Chatnode was designed to simplify the process of creating chat bots using your content. Once trained, it will be able to answer questions about your content. It can then easily be deployed on a standalone website hosted by Chatnode, using a snippet of JavaScript code to host it as a pop-up chat window on your site, or as an embedded (iframe).

There is an easy, four-step process:

Create your bot.
Train your bot.
Customize your bot.
Deploy your bot.

Step 1: Create your bot

When you first login to Chatnode, you will see a header with a menu option all the way to the right that says, “New bot” with a plus sign.



Click that button to create a new bot.

Now you have the structure of a bot but it needs to be trained.

Step 2: Train your bot

There are three (3) ways for your bot to learn your content – with a file upload, with text (cut-and-pasted or typed in), or with a URL. You do this by using the buttons in the “Add Data” section.



The first option is to load a document that has your questions and answers in it. For now, the only options are PDF, TXT, or DOC/DOCX. You can load multiple documents and have the bot trained on all of them in one session. Or you can start with one and load other documents over time. Retraining your bot is a trivial matter.

When formatting a document for training, it is important to have your questions and answers set up in pairs. Here is an example one of our clients used to build a bot for students:

Question: How can I tell if someone is listening to me? Answer: You can tell if someone is listening if they are looking at you, responding to what you say, and asking questions about what you're talking about.
Question: What should I do if I have trouble expressing my ideas? Answer: If you have trouble expressing your ideas, you can practice talking about them with a trusted friend or family member. You can also try writing them down first to organize your thoughts.
Question: How can I communicate better with my classmates? Answer: You can communicate better with your classmates by being friendly, listening to what they have to say, and being respectful when sharing your ideas.

When you click the “+ File” button, you will be prompted to choose a file from your computer. When you click “Choose Files”, you can attach a single file, or highlight multiple files to load several at once.



Once you’ve added one or more files, two things will happen:

You will get a summary of how many tokens will be required to train your bot and how many your plan supports.
The “Train Bot” button will light up.



If you meant to add several files, but forgot, do not worry. It is a simple matter to add more files and retrain the bot after it’s been through initial training.

The second training option is to simply enter text that is question and answer pairs. For these screenshots, we used the same few questions from above.



The third option, and one of the most impressive, is to train via URL. Select the “+ URL” button and enter your URL. It is important to note that the platform will crawl, starting with the URL you enter, and find all the URLs below that one on the site. It will scrape the content from all the URLs and use that information to populate the knowledgebase.

You will want to choose the right starting URL to keep the crawling focused on the pages you want to use for training. If you use just the top-level domain, you could end up with a lot more data than you really want. In the example below, the top URL was specified as chatnode.ai/blog and it returned a total of 10 URLs. Using “chatnode.ai” as the starting URL returns 18 URLs.



Once you click the “Train Bot” button, the training starts. The time required is, obviously, dependent on the volume of data being used for the training, but it is still very quick. It will rarely take longer than five (5) minutes and often takes less than one (1) minute.

Once the training is complete, you CAN deploy the bot. However, it is best to do some customization to make it match your brand more closely.

Updated on: 19/12/2023

Was this article helpful?

Share your feedback

Cancel

Thank you!