DEV Community

Arnold Samuel Chan
Arnold Samuel Chan

Posted on • Updated on • Originally published at

Penniless Chatbot Architecture

Everybody loves free stuff, programmers as well!
Running an advanced cloud service without paying a penny sounds not possible. By utilizing AWS free-tier offers, I've managed to create a simple cloud-powered chatbot for LINE Messenger

** Please note that this architecture mostly suitable for a chatbot in general use to keep the bill low **

** check the source code on github**

AWS free tier

If we carefully check on always free AWS Free-Tier offers, there are a complete service set to create a simple chatbot:

  • Script executor (AWS Lambda) free 1 million requests per month
  • Storage (AWS DynamoDB : NoSQL-styled storage) free 25 GB storage
  • Message webhook handler (AWS API Gateway) free 1 million calls per month, for 12 months


Simple Chatbot Structure

Basically, there are 4 steps to process the user's incoming message in LINE messenger:

  1. LINE messenger server receive and transfer the message to registered webhook (door)
  2. The door finds the correct function (gears) that responsible to handle the message
  3. The function sends a reply message to LINE Messenger, and stores necessary log to the database (memory card)
  4. LINE messenger transmits the reply message to the user

Pretty simple, isn't it? πŸ˜„
There are 3 AWS resources:

  • AWS API Gateway serves as the door to receive and executes the function
  • AWS Lambda serves as the function to execute the code to process the message
  • AWS DynamoDB handles message logs and records the logs as JSON-styled information. This resource is portrayed as memory card in the picture

Why Lambda?

Cheaper for low traffic usage.

Unlike the common way of computing information, it's not necessary to rent one computing resource the whole 24/7. AWS Lambda lets us rent the computing resource only at the necessary event, in this case: processing the incoming message.

AWS suggests the use of AWS Lambda only if the computing resource time short (below 15 minutes) and the service has a low-traffic, periodic activity, otherwise, there'll be a hole in your wallet πŸ˜….
Therefore, AWS Lambda is suitable in general-use chatbots and side-projects computing needs.

Enough the theory. Let's dive in the technical stuff!

Directory Structures

β”œβ”€β”€ template.yml
└── requirements.txt
Enter fullscreen mode Exit fullscreen mode


from linebot import (
    LineBotApi, WebhookHandler
from linebot.exceptions import InvalidSignatureError
from linebot.models import TextSendMessage
import json
import requests
from dynamodb import DynamoDB

access_token = <<PUT YOUR ACCESS TOKEN HERE>>  #replace with your access token
channel_secret= <<PUT YOUR CHANNEL SECRET HERE>> #replace with your channel secret
table_name = 'echobot' #DynamoDB table name
key_name = 'userID' #database's primary key
sec_key_name = 'timestamp' #databse's secondary key

def lambda_handler(event,context):

    # create BotEcho object with parameters
    line_bot = BotEcho(
        access_token = access_token ,
        channel_secret = channel_secret)

    # check the signature (originality of source)
    if line_bot.signature_check != 200:
        print("fail to signature check")
        return {'statusCode': line_bot.signature_check}

    # send the message back to the user
    response = line_bot.send_reply(line_bot.text_message)
    if response != 200:
        print('fail to send reply')
        return {'statusCode': response}

    return {'statusCode':200}

Enter fullscreen mode Exit fullscreen mode

The main file is it's the first script that is executed each time there's a received message by AWS Lambda. AWS Lambda will call lambda_handler function with 2 parameters as input: event and context. Both contain message information sent by LINE Messenger.
There are 3 steps in echo back the message:

  1. Create the BotEcho class with the LINE messenger credentials and event
  2. Validate the message signature
  3. Echo the message back to the sender

In the same file (, BotEcho class is also defined:

class BotEcho:
    def __init__(self,event,access_token,channel_secret):
        self.event = event = LineBotApi(access_token)
        self.handler = WebhookHandler(channel_secret)
        self.body = json.loads(event['body'])['events'][0]
        self.sender_id = self.body['source']['userId']
        self.text_message = self.body['message']['text']

    def signature_check(self):
        signature = self.event['headers']['X-Line-Signature']
            self.handler.handle(self.event['body'], signature)
            return 200
        except InvalidSignatureError:
            return 400

    def send_reply(self,message):
        response =['replyToken'], \
        return response

    def log(self,sender,to,message):
        log_head = ['to','message']
        log_value = [to, message]
        response = DynamoDB(table_name, key_name, sec_key_name)\
        return response
Enter fullscreen mode Exit fullscreen mode

BotEcho receives event,access_token, and channel_secret. This class utilizes a LINE Bot API python package.
There are 2 functions:

Of course, you can add the bot's ability rather than echoing back the message. Just use your creativity in it!


  1. Create a new virtual environment
  2. Install LINE Bot API package
  3. pip freeze > requirements.txt
  4. Voila!


Serverless Application Model offered by AWS.
I'll cover this part in detail in another post.
Basically it's a convenient way to deploy your app to AWS services in a script configuration. It's easy and traceable.

You need to:

  1. Have an AWS account (obviously)
  2. Install aws-cli and login to your AWS account
  3. Install SAM
  4. Create a template.yml file in your folder.

Here's the template.yml contents:

AWSTemplateFormatVersion: '2010-09-09'
Transform: AWS::Serverless-2016-10-31
Description: >
  Chatbot Echo demo
# More info about Globals:
Globals: #shared configuration
    Timeout: 3 
    OpenApiVersion: 3.0.1 ### to remove the stage name deployment bug
Parameters: #shared parameters
  StageName: # version staging config
      Default: 'prod'
      Type: String
Resources: #AWS services list
  ChatbotAPI: # API Gateway
    Type: AWS::Serverless::Api
      EndpointConfiguration: REGIONAL
      StageName: !Ref StageName
        swagger: "2.0"
          title: "ChatbotAPI"
        - "https"
              - "application/json"
                  description: "200 response"
                    $ref: "#/definitions/Empty"
                  Fn::Sub: arn:${AWS::Partition}:apigateway:${AWS::Region}:lambda:path/2015-03-31/functions/${ChatbotFunction.Arn}/invocations 
                    statusCode: "200"
                passthroughBehavior: "when_no_match"
                httpMethod: "POST"
                contentHandling: "CONVERT_TO_TEXT"
                type: "aws_proxy"
            type: "object"
            title: "Empty Schema"
  DynamoDBTable: #DynamoDB Table
    Type: AWS::DynamoDB::Table
      TableName: echobot #change to your table name
        - AttributeName: userID #change to db primary key and type
          AttributeType: S
        - AttributeName: timestamp #change to db secondary key and type
          AttributeType: S
        - AttributeName: userID #change to db primary key
          KeyType: HASH
        - AttributeName: timestamp #change to db primary key
          KeyType: RANGE
        ReadCapacityUnits: 1 
        WriteCapacityUnits: 1

  ChatbotFunction: # Lambda Function
    Type: AWS::Serverless::Function 
      Handler: app.lambda_handler
      Runtime: python3.7
          Type: Api
          Properties: #connect to our API Gateway
            RestApiId: !Ref ChatbotAPI
            Path: /
            Method: Post
        - Version: '2012-10-17'
            - Effect: Allow
              Action: #only allowed to put the record
                - dynamodb:Put
              Resource: !GetAtt DynamoDBTable.Arn

Enter fullscreen mode Exit fullscreen mode


Simple commands:

  1. sam build
  2. sam deploy --guided
  3. Attach your AWS API Gateway invoke URL to LINEWebhook URL
  4. And your EchoBot is good to go!

All SAM's deployed app can be checked on AWS Cloudformation


Refer to our AWS Services usage:

The chatbot most likely free in the first 12 months and costs some penny for the AWS API Gateway usage. love it <3


With the wonderful AWS Free-Tier offers, developers can manage to create a chatbot that'll cost very few or even free.

Serverless Application Model or SAM makes developers' life easy with a simple deployment configuration.

Finally, you can modify the chatbot's ability to do more than echo back the message! check the source code on github
Feel free to drop me a message to improve my work.


"Whatever you do, work heartily, as for the Lord and not for men" - Colossians 3:23

Top comments (0)