DocumentationUser GuideDeveloper GuideAPITechnology Partners
GO TO CONSOLECONTACT US !

Gen AI Chatbot

Suggest Edits

The generative AI chatbot is developed using Deep Learning, Machine Learning and Natural Language Processing (NLP) to mimic human dialogue and lead relevant conversations in a personalized and controlled context.

Download ebook | Discover more

indigitall develops ad hoc Gen AI chatbots for each client following their needs and the specific business environment.

Testing

indigitall offers a own space for each client to test the chatbot before integrating on their channels.

First, go to https://webpush.indigitall.com/. Click on the PROD button in the dark blue file below.

prod screen

Then select the CUSTOM environment.

custom screen

Secondly, you must choose the correct platform for your chatbot: AM1 or EU2.

  • AM1: For all customers located in the USA and Latin America.
  • EU2: For European customers and customers from other markets worldwide.
environment screen

Then introduce your email and password and click on the green button at the right to continue. You can reset your password if any time you have forgotten it.

login screen

Before entering, you need to write the code that will arrive in your email account. When you write the code in the corresponding space, please click on the ok button. In this way, indigitall reinforces the security of the system by verifying all the devices that give access to our platform with the double authentication method (2FA).

code screen

Third, choose your project clicking on it: "xxxx PROD"

project screen

A new window appears. Please click on the indigitall logo displayed at the bottom right corner of the screen.

icon screen

Immediately a chat window appears and you can write in the space below your questions and check how the chatbot answers them.

chatbot screen

How is a bot powered?

👍

The quality of the documentation will affect the quality of the responses

Generative AI bots need information in order to generate their responses based on real data, not fabricated. It is important to keep in mind these details:

General requirements

  • Files should be provided in a single folder or file, never in nested folders.
  • The larger a document, the more costly its maintenance will be. There is a 50MB limit per file.
  • If the information source needs to be updated more frequently than daily, it should be done via API access, database, or other methods, but never by reading directly from a URL.
  • The documentation must not be contradictory.
  • All documentation must be in the same language. (There should be no duplicates, such as the same information in two different languages).
  • File types the bot can read:
    • txt: This is the best option.
    • pdf: But it must be text-based. The bot will not read images contained in the pdf, and they may even cause problems in generating the bot.
    • doc: It can read Word documents, but the quality of what has been read needs to be ensured.
    • URLs

Requirements for URLs

For the initial load, a detailed list of URLs in Excel format must be provided, including their update frequency. The URLs must be public or we must have access to them through our IPs (we have a fixed IP).

❗️

Keep in mind

  • The bot does not read images.
  • Websites or main URLs are not accepted.
  • The data that is read are the texts from the HTML, not from the JavaScript.

Updated 9 days ago