15 things all Dialogflow bot makers should know

Here are a list of things that I would like to tell anyone who is getting started with Dialogflow.

Things all Dialogflow bot makers should know

Note: These are not in any particular order 🙂

Scripted

NLU powered chatbots are different from scripted bots. This means a Chatfuel bot can never do what a Dialogflow bot can.

Regex

Regex entities have already been added into Dialogflow. Please start using them and don’t reinvent the wheel.

Initiate conversation

You cannot really have Dialogflow initiate the conversation. Even in those cases it seems possible, something much more complex is going on under the hood.

Website chatbot

Dialogflow has a 1-click integration which lets you add a chatbot to your website.

Unfortunately, the 1-click integration is also very limited, and you need to create your own integration if you need any of these things: clickable hyperlinks with anchor text, images, buttons, bold face, other text formatting, and any other type of advanced control such as list boxes.

Slot filling

Slot filling is a very tricky feature to implement and I recommend avoiding it

Context lifespan

The lifespan you choose for your contexts will decide your candidate intents at each step in the conversation, so I recommend using a lifespan of 1

Flowcharts

You can jump into agent creation without creating a flowchart first, but you will find that your agent becomes hard to manage very quickly. 

Agent ZIP file

Dialogflow provides an agent ZIP file import/restore feature. This means you can define all your intent training phrases in a spreadsheet, create a tool which converts it into the ZIP file format, and import it, and you will have a working Dialogflow agent

Export agent ZIP file

You can also EXPORT the agent ZIP file. This means, instead of wasting time clicking into each and every single intent to find a specific training phrase or response, you can create a tool which lets you search across your intents

Training

You can break your Dialogflow agent just by doing some innocuous thing, like “approving” a phrase in your Training tab

Open source

Dialogflow is not open source, and it is possible they might never release the source code as open source.

Machine learning

Almost no one (outside of the Dialogflow team) knows what machine learning algorithm is used in Dialogflow

(But there is an article which came out recently which might provide some hints)

Webhooks

You don’t need to write code just to understand how webhooks work under the hood

Wildcard entity

The wildcard entity should be used carefully and sparingly

NLU

Learning a little bit of NLU (Natural Language Understanding) can be very helpful when it comes to building Dialogflow agents.

About this website

BotFlo1 was created by Aravind Mohanoor as a website which provided training and tools for non-programmers who were2 building Dialogflow chatbots.

This website has now expanded into other topics in Natural Language Processing, including the recent Large Language Models (GPT etc.) with a special focus on helping non-programmers identify and use the right tool for their specific NLP task. 

For example, when not to use GPT

1 BotFlo was previously called MiningBusinessData. That is why you see that name in many videos

2 And still are building Dialogflow chatbots. Dialogflow ES first evolved into Dialogflow CX, and Dialogflow CX itself evolved to add Generative AI features in mid-2023