Integrate Your Knowledge-base into a Chatbot

A guide to Relevance AI's "Ask Relevance" which enables you to create a QA engine (i.e. chatbot) on your knowledge base in a matter of minutes.

Problem Statement

Ever considered integrating a chatbot into your website to provide instant answers to clients' queries? An accurate model that understands the knowledge base, as well as clients' questions, can save us millions of minutes dealing with repetitive queries. Relevance AI provides you with a zero-code platform to apply AI to your data, a quick-to-set-up and scalable Chat GPT and a strong infrastructure to even improve Chat GPT. All the above under your free account!

Who Are the Typical Editors and Viewers of This Use Case?

Editors: Business owners, Customer service managers
Viewers: Anyone who can benefit from a certain knowledge-base

How long does it take to set up?

It is a very quick process, for a medium size dataset, it is a matter of minutes. The time-consuming step is where AI reads your data, meaning the time is heavily affected by the size of your knowledge base. However, the good news is that once the data is loaded to your account and prepared for the Ask Relevance widget there is no wait time for further use it.

How does it work?

Here are the three steps you need to take after signing up to Relevance AI.

Upload your data
Set up Ask Relevance for your data
Add the widget to the chatbot on your website

How To Get Started: Ask Relevance

1. Upload your data

After signing up to Relevance, go to the homepage and click on "Ask relevance". The wizard will guide your through the steps.

Next, you will be uploading your data. Select the appropriate option based on your data type. For instance, if your knowledge base is a PDF, select the "PDF Files" option.

2. Set up Ask Relevance for your data

If using the onboarding flow, the relevant workflow for processing your knowledge base will open automatically. For instance, Enable Ask Relevance on Text or PDF if your source data is a CSV or a PDF. If you are not using the onboarding flow for Ask Relevance, you can access the relevant workflows under the workflow page as shown in the image below.

Follow the workflow wizard and in a few minutes you will be provided with your chatbot widget and sandbox to try the chatbot out.

3. Ask Relevance Widget

The Ask Relevance widget and the preview box to test the chatbot are shown in the image below.

Type in your question and hit enter to try it out
Use the copy button and add the snippet to your website to access the chatbot directly from your platform.

Configuration Section
Even though automatically set up, this section allows to make some modifications. The text field is the main content against which queries are run. You need to select the corresponding vector under "Vector field" and the model used to generate the vector.
Note1: Make sure the vector field generated by the selected Model is used. For instance, if you use the Open AI model under Enable Ask Relevance on Text or PDF, you will need to select the corresponding vector field on this page.
Note2: By toggling on the next options, you can select to display the references or a list of other possible responses when a search result is provided.

📘
Unable to answer
This normally happens when the set up is incorrect.

Embed on your website Section
Copy what is under this section and embed it to your website for the QA widget to appear on the page.

Tip

When testing the chatbot, ask questions about the content of your data, not questions dealing with matters such as counting the number of times something is found in the data.

Common questions

What if my knowledge base is online?

It depends on the source. One very common approach is to crawl the sitemap.xml. Also, it is possible to access public repos such as GitHub.

What version of Chat GPT are you using under Ask Relevance?

It is currently 3.5. If you wish to use your API key for version 4, it is best to use our Chaining functionality, which lets you create custom prompts. Under chaining can take your own API key.

What plan to use if my main use-case is Ask Relevance?

Our pricing starts from $82/month - this is typically enough for most use-cases of Ask Relevance. If your usage grows, we charge based on the credits used which depend on the workflow you run.

Do you handle versioning if my source contains multiple versions?

Not yet! But it is on our roadmap.