Create your bot
Basics - Essentials to build a bot
Variables in Landbot - Getting Started
Landbot System Variables
Builder's keyboard shortcuts and mechanics! 🏗
Builder Interface
Supported Browsers
Guidelines 🚀 5 best practices for chatbot design in Landbot 🤖
How to "debug" (troubleshoot) your bot's flow to spot possible errors (for non coders)
Build a bot
Blocks and flow operations
AB Test block (randomize flow)
Buttons block
Buttons block
Reply Buttons block (WhatsApp)
Any of the above Output
How to open a new URL in another tab (window)
How to set up Multiple Choice questions
Code Block (Javascript)
Conditions block
Conditions block
Conditions block II (with Dates, Usage and Agents variables)
How to ask a question based on a variable not being set (empty URL params)
Flow operations
Global Keywords 🌍
Keyword Jump
Keyword Options 🔑 Assign keywords to buttons (WhatsApp and Facebook)
Lead Scoring block
Persistent Menu
Jump To block
Formulas block
Get started with the Formulas block
Formulas Blocks Dashboard
Formulas - Regex
Formulas - Date
Formulas - String
Formulas - Logical
Formulas - Math
Formulas - Object
Formulas - Comparison
Formulas - Array
Goals block
Goodbye block
Media Block
Multi-Question block (Create a Form)
Note block
Question blocks
Question: Address block
Question: Autocomplete block
Question: Date block
Question: Email block
Question: File block
Question: Name block
Question: Number block
Question: Phone block
Question: Picture Choice block
Question: Rating block
Question: Scale block
Question: Text block
Question: URL block
Question: Yes/No block
Send a Message block
Send an Email block
Set Variable block
Starting Point
Business Hours block
List Buttons Block (WhatsApp)
Collect Intent block
Bricks
Bot Settings
Bot General Settings
Second Visit Settings
Custom System Messages
Hidden Fields (Get params / UTMs from url and use it as variables)
Landbot native SEO & Tracking tools
Typing Emulation (Message Delay)
Landbot in Wix
Chats - Livechat and Human Takeover
Design (bots for Web)
Share & Embed
Account Settings and Billing
Billing
Account Settings
Account Settings
Account Homepage
Dashboard
Roles & Permissions for Teammates
Agent Status and Log out
Common reasons for not receiving account activation email
Privacy and Security
Landbot's Privacy & Security Policies
Compliance with the GDPR - General Data Protection Regulation
Landbot's Abuse Policy - How to Report an Abusive Use of Landbot
Teammates - Add extra agents (seats) to your Account
Notifications section
Integrations with Landbot
ActiveCampaign
Airtable
Airtable integration block
Get data filtered from Airtable with a Brick- Shop example
20 different ways to GET and filter data from Airtable
How to add/update different field types in Airtable (POST, PATCH & PUT)
How to Create, Update, Retrieve and Delete records in Airtable (POST, PATCH, GET & DELETE)
Get more than 100 items from Airtable
Insert Multiple Records to Airtable with a Loop
How to Get an Airtable Token
Advanced filters formulas Airtable block
Update Multiple Records in Airtable Using a Loop
Calendly
Carrd
DeepL
Dialogflow
Dialogflow & Landbot course
Dialogflow & Landbot intro: What is NLP, Dialogflow and what can you do with it?
Dialogflow & Landbot lesson 1: Create your first agent and intent in Dialogflow
Dialogflow & Landbot lesson 2: Get the JSON Key
Dialogflow & Landbot lesson 3: Setting up of Dialogflow in Landbot
Dialogflow & Landbot lesson 4: Training phrases and responses for a FAQ
Dialogflow & Landbot lesson 5: Entities and Landbot variables
Dialogflow & Landbot lesson 6: Redirect user depending on Dialogflow response parameters (intent, entities and more)
Learn more about Dialogflow - Courses and communities
Integrations > Dialogflow Block
How to extract parameters from Dialogflow response with Formulas
Dialogflow Integration Dashboard
Dialogflow in Unsupported Languages (& Multilingual)
Dialogflow - How to get JSON Key
External API REST
Dynamic Data (Arrays)
Webhook Block Dashboard
Webhook block (External API REST)
How to TEST your Http Request (Webhook block)
Facebook Pixel
Google Analytics
Google Adwords
Google Maps
Google Maps API Key for Address block
Extract Data With Google Maps Geocoding API
Calculate Distances With Google Maps API
Google Sheets
Google Sheets (Insert, Update and Get data)
How to use Google Sheets to create a simple verification system for returning visitors
How to Upload Document & store link
How to give unique Coupon Codes (with Google Spreadsheets)
Google Sheets Integration Dashboard
How to insert a new row with data and formulas in Google Sheets
How to use Google Spreadsheet as a Content Management System for your bot
Google Tag Manager
How Track Google Analytics Events in Landbot with Google Tag Manager (GTM)
Landbot in your web with Google Tag Manager
HubSpot
IFTTT
Integromat
Connecting MySQL with Make.com (formerly Integromat)
How to upload a file to Google Drive using Make.com (formerly Integromat)
Send WhatsApp Message Template from Make (ex Integromat)
Make.com (formerly Integromat) trigger
How to send an email through Sendinblue by Make.com (formerly Integromat)
Get Opt-ins (Subscribers) from Facebook Leads using Make (ex Integromat)
How to extract data from an external source with Make.com and use it in Landbot
Integrately
Intercom
Vonage
Mailchimp
OCR
OpenAI
Create a JSON format response from OpenAI in WhatsApp
Connect OpenAI Assistant with Landbot
How to build a FAQ chatbot with GPT-3
GPT-4 in Landbot
Fine-Tune GPT3 with Postman
OpenAI
Prompt Engineering for GPT-3
Build a Customer Service Bot with ChatGPT and Extract Information
Pabbly
Paragon
PDF Monkey
Pipedream
Salesforce
Segment
Sendgrid
Slack
Stripe
Squarespace
Sharetribe
Shopify
Store Locator Widgets
How to display the near location of your store in WhatsApp (with Store Locator Widgets)
How to display the near location of your store for Web bots (with Store Locator Widgets)
Webflow
Wordpress
Xano
Zendesk
Zapier
Zapier block
Zapier Integration Dashboard
How to insert a row to Google Spreadsheet by Zapier
How to generate a document with PDFMonkey by Zapier
Send WhatsApp Message Templates from Zapier
How to send an email with Gmail by Zapier
Get Opt-ins (Subscribers) from Facebook Leads using Zapier
How to extract data from an external source with Zapier and use it in Landbot
Zapier trigger
How to complete a digital signature flow by Zapier
Make a survey with Landbot and display the results in a Notion table using Zapier
Trigger Automation block
WhatsApp Channel
Getting started with the WhatsApp Integration
WhatsApp Testing
Facebook Business Verification - Best Practices 🇬🇧
Adding a WhatsApp number to your account
WhatsApp Channel Settings
Growth Tools for WhatsApp
Opt-in, Templates and Campaigns
Parent Bot/Linked Bot - Add a main bot to your WhatsApp number
How to get Opt-ins (Subscribers) for your WhatsApp 🚀
WhatsApp Campaigns 💌
WhatsApp's Message Templates
Send WhatsApp Messages with Landbot API
Audience block
Recognise the users input when sending a Message Template with buttons
Audiences
Create Opt-ins (Subscribers) using Landbot API
Opt-In block for WhatsApp 🚀
Opt-in Check Block
WhatsApp Bot Creation - Best Practices and compatibility guide
Build a WhatsApp Bot - Best Practices and User guide
Build a WhatsApp Bot - Best Practices for Developers
WhatsApp bots - Feature Compatibility Guide
Types of Content and Media you can use in WhatsApp 🖼
WhatsApp - How to direct a user through a different bot flow on their second visit
WhatsApp - Get user out of error message loop
Javascript in WhatsApp
How to calculate the number of days between two selected dates (WhatsApp)
Creating a Loop in WhatsApp
AI in WhatsApp
Send Automated Message Templates based on Dates
Different ways to format numbers with JS (WhatsApp)
Notify Teammates of Chat via WhatsApp
Creating a Simple Cart in WhatsApp
1. WhatsApp Article Directory
2. WhatsApp Integration FAQ's
WhatsApp Number Deletion (WA Channel management)
Customize and embed your WhatsApp Widget
Meta processes guide: FBM verification, Official Business Account (OBA) requests, Appeals
Additional Number integration: Limitations and Requirements (Number integration)
Send a Message Template from the Builder
Existing WhatsApp Number Migration (Migration)
Other Channels - Messenger and APIChat
Facebook Messenger
The Facebook Messenger Ultimate Guide
Types of content you can use in Messenger bots 🖼
How to Preview a Messenger bot
Facebook Ad connected to Messenger Bot
API Chat (for Developers)
How to create an API chat bot and set up the Hook Url
Building a Slack APP Bot with Landbot APIChat and Node JS
Build a Telegram Bot with Landbot APIChat and Node JS
Duplicating bots
AI in Landbot
Metrics and Data Management
Metrics Section
How to export the data from your bots
Export data: How to open a CSV file
Bot's Analyze Section
For Developers & Designers
CSS and Design Customizations
Dynamic Customizations
How to change Avatar dynamically (Landbot v3)
Dynamically Change a Bot's Background
Create a Checkbox in the Form Block
Dynamically Change Any Style
Embed Customizations
How to "send" a user to a specific point in the flow with Javascript and with the API
Modifying Embed Size
Load script and display bot on click button
How to launch a Landbot by clicking a button
Open LiveChat bot as soon as page loads (3 ways)
Change Landbot custom CSS dynamically from parent page onload
CSS Customization Examples
CSS Customization Examples: "Back to School" Theme
CSS Customization Examples: Call To Action: WhatsApp
CSS Customization Examples: "Translucid"
CSS Customization Examples: "Minimalist" Theme
CSS Customization Examples: Carrd Embed Beginner
Make Persistent Menu Appear During the Flow
CSS Customization Examples: Video Bubble
CSS Customization Examples: CV Template
Design Customizations
How to use Google Fonts in Landbot
CSS for Typewriter Effect
Proactive Message Customizations with Javascript and CSS
Landbot v3 - Web CSS - RTL
Hide time stamp (under avatar)
How to remove Landbot branding
HTML Template for Emails
Widget/Bubble Customizations with Javascript and CSS
Detect if a visitor is on Mobile/Tablet or Desktop
Customized display in the bot
How to display an HTML Table and a List in Landbot v3 web
Progress Bar Workaround
Add a Chart (with Chart JS library) in your Landbot
How to embed a .gif file inside a message
Different ways to embed Videos in Landbot
Pop up modal to embed third party elements
Display video and hide button to continue until video has ended
Components CSS Library
Components CSS Library Index
Background Class CSS
Identify Blocks CSS
Buttons Class CSS
Header Class CSS
Media Class CSS
Message Bubble Class CSS
Miscellaneous Classes CSS
Recommended Image Sizes
Dynamic Data CSS
Form Block CSS
Multi Questions CSS
Get started guide for CSS Design in Landbot
Landbot APIs
APIs
Create an SMS bot with ClickSend and API Chatbot
reply from Slack: How to create an integration to allow agents reply WhatsApp users from Slack (with Node JS)
MessageHooks - Landbot Webhooks
Javascript and SDK
Workarounds and How To's
Building Tips & How to's
How to build an event registration Landbot (to be used in one screen by many attendees) (web only)
How to redirect user to another url in your site with Livechat open to continue conversation
How to let user select a time of booking (with a minimum 45 minutes notice)
Get the array's index of the user selection and extract information from array
How to redirect visitors to a URL (web only)
Embed Google Maps
How to add a Click-to-Call/Email/WhatsApp button
How to disable a bot
Airtable usecase: Create an event registration bot with limited availability
How to calculate the number of days between two selected dates
Reservation bot with Airtable
Set a timer to get the time spent during the flow
Workarounds & Codes
How to Detect Visitors Browser
Conditional Welcome (Non-embedded bots)
Conditional Welcome: Direct the Flow of Your Bot Based on External Parameters
Two-Step Email Verification
Detect if bot was opened
Different ways to format numbers with JS
Trigger Event if User Abandons Chat
Calculate Distances in WhatsApp
Customized Behavior in Mobile Browsers
Trigger Event if User Abandons Chat Using AWS
Launch Bot On Exit Intent
Display Bot During Business Hours Only (Livechat & Popup)
Delete Customer Data in Flow
Cookie consent banner (full page / full page embed)
Add Captcha Verification (Non-Embedded Bots)
Launching a bot depending on browser language
Trigger a Global Keyword with JS (web v3)
Create a Scale out of Buttons - Workaround
How to set up questions with a countdown
How to pass WordPress logged in user data to Landbot
Create Dynamic Shopping Cart with JS and CSS
Set the flow depending on the url path (for embedded landbots)
Generate a URL that has variables from user answers
Resume flow based on external process with Landbot API (Request, Set, Go)
Creating Masks for User Input (2 examples)
Popup on Exit Intent
Open / Close a Web bot (embedded)
More Topics
- All Categories
- Integrations with Landbot
- OpenAI
- How to build a FAQ chatbot with GPT-3
How to build a FAQ chatbot with GPT-3
Use a FAQ bot to answer customer queries with Landbot’s GPT-3 Integration
Chatbots are commonly used to answer frequently asked questions from customers. This can help reduce the manual workload for human agents by having a chatbot answer FAQs in a conversational, friendly way. In the past, this kind of chatbot experience has not always been positive. Many times, the chatbot would fail to understand the user’s intention, providing generic responses that were simply not relevant. This was a big blocker for the practical use of AI chatbots until the emergence of ChatGPT and other LLMs, but even they have their limitations, still.
GPT-3 Integration and Landbot: How to Build the Ultimate FAQ Bot
We can build a chatbot with GPT-3 and Landbot to answer frequently asked questions and summarize answers, almost like a human. This tutorial will teach you how to build it by going through the following steps:
- Using an existing database of questions and answers, such as articles from a documented FAQ page or Knowledge Base articles on your website. These will serve as a source of information for the FAQ chatbot. In this example, we will use Knowledge Base articles pulled into an Airtable database.
- Configuring our GPT-3 Integration inside Landbot to analyze the users’ questions and search our Knowledge Base articles for the appropriate answer.
- Creating friendly and clear responses on-the-go using GPT-3.
All of this can be accomplished inside Landbot and can also be connected to other integrations, or be included in a more complex chatbot flow.
Let's see how we can do it:
Pre-configuration Requirements
- Make sure you have an OpenAI account and its API key. (You can create your account here.) Have your API key ready to configure it later inside Landbot.
- We will use Airtable our database for this tutorial. We will save our Knowledge Base articles here. You’ll need an Airtable account to create a database for this example.
We’ve created and Airtable template that you can use as an example to save time.
Do you want to create the Airtable database yourself?
- Fill in the content of the columns for each article.
First steps in Landbot’s Builder
- Log in to Landbot and create a new bot.
- Add an open text question (Question text) to the bot, where we will ask the user “what their question is”.
- Save the user's response in a variable called "text". This is the default name when creating a question.
Setting up the questions and answers database using Airtable
As indicated in step 2 of the Pre-configuration Requirements section of this tutorial, you should already have your Knowledge Base content set up in Airtable. Now we are going to connect it to Landbot so that the AI can read it.
- Now the time has come to connect an Airtable block and set up the base and table. This should go right after the Question text block we made, in the previous step.
- Select "Get multiple records" to capture the content from the Knowledge Base articles in a Landbot variable.
- Type “knowledge_base” and save it as the variable name.
- Apply the changes at the bottom.
You now have all the content from the Knowledge Base database in Airtable saved as the "knowledge_base" variable inside Landbot. This is the information GPT-3 will use to answer users’ questions.
Setting up the AI
Next, you are going to set up the AI using our GPT-3 integration so that it can understand the question and use the Airtable database that has our Knowledge Base content stored as a source of information.
What can I do if my Knowledge Base is bigger?
- Use a less expensive, less restrictive LLM (like cohere.ai) to match the user question with the right Knowledge Base article containing the answer, and then use GPT-3 to summarize the answer.
- Or just ask the user which category the question falls under so that you can filter the Knowledge Base articles to reduce their size.
- Or you can fine-tune a GPT-3 model so that you don’t need to send all of your Knowledge Base content to OpenAI. With this approach, you can skip the Airtable section of this article.
- Right after the Airtable block, create three text variables by connecting three set variable blocks in a row.
- In the first set variable block, save your Open AI API key in the variable "openai_apikey".
- In the second "set variable", write a prompt for how you want GPT-3 to understand the question, use the Knowledge Base content and prepare a response. Save it in the variable "prompt".
For the prompt copy the following text:Answer the QUESTION using DATA as the information source, in a clear sentence.
QUESTION: "@text"
DATA:
```
@knowledge_base
``
ANSWER: - It's very important to use the variable names mentioned above to avoid flow errors
- On the third "set variable" block, configure the GPT-3 model that you’ll use for the request. In this case, the variable name will be "model" and its content will be "gpt-3.5-turbo-instruct" *Bonus: You can use other models within the list of available GPT-3 models, or even train one by yourself to reduce costs.
Integrating GPT-3
Finally, set up our GPT-3 integration using Landbot’s webhook block and the configuration values already defined.
- Add the webhook block connected to the final set variable block (the “model” variable name) mentioned above.
- For the URL, set the "post" method paste the URL "https://api.openai.com/v1/completions".
- Set up the "Customize headers" field and add these two parameters:
KEY
VALUE
Content-type
application/json
Authorization
Bearer @openai_apikey
- Personalize the body by copying the following content into the text area:
{
You can customize these advanced options by following OpenAI guides.
"model": "@model",
"prompt": "@prompt",
"suffix": "\\n// Finish",
"temperature": 0.5,
"top_p": 1,
"frequency_penalty": 0,
"presence_penalty": 0,
"stop": ["\\n// Finish"]
} - This step is very important: Add the following test data to send a request to OpenAI and create the variable structure inside Landbot with the response from GPT-3. Copy the following test data into the "Test your request" section (replace the API key value with your real one).
VARIABLE NAME
TEST VALUE
openai_apikey
Replace with your API key
prompt
Answer to the QUESTION using DATA as the information source, in a clear sentence.
QUESTION: "How many days do I have to return a product?"
DATA: You have 10 days to return a product.
ANSWER:
model
gpt-3.5-turbo-instruct
- Once these variables are configured, click on "Test the request". You should see a response similar to this one if there are no server errors. Remember, OpenAI servers get overloaded sometimes.
- Finally, save the GPT-3 response in a new variable called "response" that you will create.
- Click "Save responses as Variables.” Add a new string variable, and call it "response."
- Above the “response” variable, you will see an option called "Entire response body.” Click on it, and you will see a dropdown with several options. Look for where it says "choices.0.text", and select that.
- Activate the "Response routing" option.
- Finally, save the webhook by clicking on "Apply.”
Final details: Display the response and error cases
The webhook block should now display three numeric options: 200, 400, and 500. Option 200 is the success code, indicating that you have received a correct response from GPT-3. 400 and 500 are errors, so your need to prepare flows for those cases.
- Output 200: Connect it to a new message block and add “@response” as its content. This message content now will be a tailored message by GPT-3.
- 400 and 500 outputs: Connect a button block to explain the issue and allow the user to choose whether they want to try the request again. You can also be creative and define more advanced flows to create a better error experience—if you want.
And that's it! Now simply test the flow to make sure there are no errors. We hope you enjoyed this tutorial and can start playing with our GPT-3 integration to answer FAQ from your customers!
Access the FAQ demo bot here –reliability issues curtesy of OpenAI servers may apply.