Your Knowledge Base

Structuring Your Knowledge Base

While AI advancements often negate the need for extensive data restructuring, optimizing your knowledge base can significantly enhance outcomes. Here's how to refine your data for murai:

1. Clarity & Grammar: Ensure language is clear and grammatically correct.
2. Simplicity: Distill complex details for straightforward understanding.
3. Link Descriptions: Clearly articulate the content of embedded links.
4. Logical Order: Start with broad concepts, then detail specifics.
5. Format: Use comma-separated lists and plain text for ease of reading.
6. Direct Answers: Provide straightforward responses, omitting Q&A formats.
7. Contextual Independence: Make each piece of information understandable on its own.

For the best AI performance, a well-organized knowledge base is key. Start by consolidating FAQ content into a singular, unambiguous document. Clear, well-structured data ensures more precise AI interactions.

For detailed guidelines, click here.

Uploading Your Knowledge Base

Uploading your knowledge base is a pivotal step, setting your chatbot apart from just a regular GPT chat.

Tip: The amount of documentation you can add depends on your plan. If you're still on a trial plan, consider consolidating all information into one text file or contact us to discuss your use case.

To provide the chatbot with your knowledge base you can:

1. Go to the desired chatbot.
2. Expand the chatbot's dropdown by clicking on it.
3. Click the Train Your Chatbot button.
4. Choose one of the options below depending on the format of your data.

Upload Full Documents

If you already have prepared documents with information, you can upload them directly as plain text.

Tip: If you have a lot of PDF documents check out our PRO plan. It enables users to OCR scan their PDF files and convert them to plain text directly in our application.

Uploading Plain Text

1. Drag the text file into the Upload Documentation box or paste the entire text via the Paste Text option.
2. Click on Add New Document and wait for the uploading and analyzing process to finish.

Tip: If you have a lot of PDF documents check out our PRO plan. It enables users to OCR scan their PDF files and convert them to plain text directly in our application.

Uploading PDFs

1. Upload the desired PDF file.
2. Depending on the PDF size, wait for the OCR process to finish.
3. Double-check the information in the created text file for accuracy.
4. Upload and save the data.

Import Website Data

The simplest way to incorporate data into your chatbot's knowledge is by supplying your website's URL. The murai will automatically gather content from your website, which will then be utilized to train your chatbot.

You can easily add your own page by:

1. Clicking the Add Website button.
2. Enter the full page URL, such as: https://example.com, and confirm it.
3. Provide any additional subpages you want our app to check and read, following the above example.
4. The app will display successfully scanned webpages and their content. Ensure everything is correct and none of the pages are missing.

Provide Data Through Functions

Note: This feature is still in beta.

The murai functions provide another excellent method of adding information to the chatbot. The main advantage of using functions is that information is gathered live and processed automatically. This is done by connecting to your desired service, which you or your company are already using, and connecting to it via API calls. However, there are some limitations; for example, you can only send POST requests through the API.

Adding a new function may require some technical knowledge, depending on the API you are calling. If you get stuck, you can always contact us.

1. Click the Add Function button.
2. Fill out the function form with the correct data, obtained from the service's documentation. You can find more about the fields in the below table.
3. Test the function by clicking Test Function to see how it responds to your input.

Field	Description
Function Name	Name under which the function will be stored. Provides context for the chatbot.
Function URL	This URL should serve function-specific data. When the function call is executed, the chatbot will utilize the HTTP protocol and employ the POST method to call this URL. Function parameters are sent as a URL query string.
Function Description	Provides additional context for the chatbot. Describe what your function does. Chatbot will select which function to call based on the provided description and the user's query.
Enable Authentication	You can enable authentication to secure your function's data. Function authentication is added to HTTP function call headers.
Authentication Header	Name of the HTTP header used for function authentication.
Authentication Token	Authentication token used for function authentication. Authentication token is sent with every function call in the HTTP header specified by the Authentication header field.
Function Parameters	Add parameters that can be used to call your function. Chatbot will select the parameters automatically based on the user's query and parameter description. Name: Obtained from chosen service documentation. Has to match. Type: Conveys to chatbot what kind of data to expect. Description: The provided description provides additional context for chatbot. Describe what the filed is for. Required: Checkbox that marks if the parameter is required or not.

Success! You have now enriched the chatbot's data and can begin testing it.