🌈 Rainbow bank

An offline & mobile first neo-bank for underserved population.

Tech

Backend: NodeJS
IAAS:
1. Db: AWS Dynamo
2. Compute: AWS Lambda
3. Gateway: AWS API Gateway
4. Cache: AWS Elasticache
5. Broker: AWS SQS
CPAAS:
- Expensive, difficult, scalable (evaluated, but not used in this project):
  - Receiving:
    - textlocal +919220592205 NLLG7 (prefix)
    - ifttt (sms → webhook)
  - Sending:
    - twilio
    - gupshup
- Inexpensive, easy, very small scale (used in this project):
  - Android phone connected to internet, is installed with IFTTT & Pushbullet
  - Users send sms to this phone, an ifttt recipe triggers on incoming sms containing a keyword RAIN
  - Trigger reads sms content and POST's to Webhook which is hosted in AWS
  - Webhook processes the sms content & calls pushbullet api to send sms via same android phone
  - Inexpensive, since only sms sending charge as per carrier tariff
  - No need to dealing with CPAAS service providers
  - Caveats:
    - Android phone must be online all the time with IFTTT and Pushbullet running in the background

Infra

Serverless architecture, primarily depends on lambda functions for running backend workflows and sqs for message broking
User interfaces with SMS, offline & mobile first

Db schema

table: userAccountIdMapping

phone (pk)
id

table: userAccount

id (pk)
name
phone
loc
pan
verification
balance
currentActive
createdAt

table: userTxn

txnId (pk)
firstParty
secondParty
requestType
money
status
createdAt
currentActive

table: floatingCashRequest indicates latest ask for floating cash request id

phone(pk)
id

table: userRequest

id (pk)
phone
requestType
where
money
otherAccount
status
extraInfo
currentActive
createdAt

table: userAccountLedger

id (pk)
phone
op
note
money
openingBalance
createdAt

table userBucket

phoneWithBucketName (pk)
balance

enums

const requestType = {
  register: 'register',
  collect: 'collect',
  pay: 'pay',
  transfer: 'transfer',
  seeSaved: 'seeSaved',
  sip: 'sip',
  bucket: 'bucket',
  findDeposit: 'findDeposit',
  findWithdraw: 'findWithdraw'
}

const txnStatus = {
  created: 'created',
  success: 'success',
  failed: 'failed',
}

const op = {
    debit: 'debit',
    credit: 'credit'
}

event schema

Message received from SQS (textlocal)

{
    "Records": [
        {
            "messageId": "315f83aa-e7f5-4cab-ab7a-08ebed382643",
            "receiptHandle": "AQEB9RDP/7wzhKAyAvKcPsj4ZIiP0lT2v7rUEHS3NdY9NUcl3c/MYrUxx/A+G95BCMxSbbvQlHr5/aqREd2fR7jlAljH+Tco+9yJpWD5m+2zlJL5eZ0lthD/Iu2fXTS7CoibGe9OXygVjtpprk5Rw6npGDSLm2g7XFnH9WQUS7Ff1arQF3QlAu300XTv8nvJVEn7ME2eUrp/y1lIQYQUamULE2tj1gEoC2aXKJx6D/nlV0UR3a9LxFgWQ43AGenzKjWCWw+XwT7aaqys1Rl93NlQeEx+uHOfTdCBjlb8+Ef7Xb0=",
            "body": "{\"sender\":\"919439831236\",\"content\":\"NLLG7 FIND WITHDRAW 1250 DELHI\",\"inNumber\":\"919220592205\",\"submit\":\"Submit\",\"network\":\"\",\"email\":\"none\",\"keyword\":\"NLLG7\",\"comments\":\"FIND WITHDRAW 1250 DELHI\",\"credits\":\"10\",\"msgId\":\"8831460984\",\"rcvd\":\"2022-04-26 17:27:08\",\"firstname\":\"\",\"lastname\":\"\",\"custom1\":\"\",\"custom2\":\"\",\"custom3\":\"\"}",
            "attributes": {
                "ApproximateReceiveCount": "1",
                "SentTimestamp": "1650974230025",
                "SequenceNumber": "18869393476595952384",
                "MessageGroupId": "Group1",
                "SenderId": "AIDAXYDVU3NFEQC345T7T",
                "MessageDeduplicationId": "3df828ca-1c61-40ef-aa43-f09343c13b3b",
                "ApproximateFirstReceiveTimestamp": "1650974232040"
            },
            "messageAttributes": {},
            "md5OfBody": "d4cb092b27df4d5c194720c09f47e11c",
            "eventSource": "aws:sqs",
            "eventSourceARN": "arn:aws:sqs:ap-south-1:532822678346:smsReceivedQ.fifo",
            "awsRegion": "ap-south-1"
        }
    ]
}

Message body published to sqs

original payload parsed from textlocal

{
        "sender": "919439831236",
        "content": "NLLG7 TRANSFER 3000 9876543210",
        "inNumber": "919220592205",
        "submit": "Submit",
        "network": "",
        "email": "none",
        "keyword": "NLLG7",
        "comments": "TRANSFER 3000 9876543210",
        "credits": "10",
        "msgId": "8831346667",
        "rcvd": "2022-04-26 16:25:34",
        "firstname": "",
        "lastname": "",
        "custom1": "",
        "custom2": "",
        "custom3": ""
    }

Note: payload actually sent to sqs

{ 
  "sender": "919439831236", 
  "content": "rain balance"
}

Docs

Dynamodb document client
Setup awscli for deploy script deploy.sh:

python3 -m venv .venv
.venv/bin/pip install awscli

To build & deploy:

sink (receive incoming sms)

./deploy.sh sink

worker (process sms)

./deploy.sh worker

worker (process sms sending tasks)

./deploy.sh sender

deploy all 3 functions:

./deploy.sh sink && ./deploy.sh worker && ./deploy.sh sender

To build lambda layer

create directory structure:

mkdir -p aws-sdk-layer/nodejs
cd aws-sdk-layer/nodejs

Use amazon linux 2 compatible environment to install required packages:

docker run --entrypoint "" -v "$PWD":/var/task "public.ecr.aws/lambda/nodejs:14" /bin/sh -c "npm install aws-sdk redis node-fetch@2.6.7; exit"

zip it

zip -r ../package.zip ../

create layer

Name: <anything>
Upload zip
Runtimes: Node.js 14.x

Use the layer in lambda function. This prevents bundling node_modules with zip that is deployed to lambda. Reduces the lambda zip package size and prevents version conflicts with lambda preinstalled packages.

Note: If elasticache is in private subnet

Lambda function needs to be inside vpc and must be associated with privates subnets same as elasticache
AWS VPC endpoints needs to be setup for - SQS & DynamoDB, since requests for these services travels through internet
Environment variables (Key: Value), needs to be set for lambda functions:

ACCESS_KEY_ID:	---
QUEUE_URL:	https://---.fifo
REDIS_ENDPOINT:	redis://---.aps1.cache.amazonaws.com:6379
REGION:	ap-south-1
SECRET_ACCESS_KEY:	---

Modify config: timeout to 30seconds at least, Memory to 256mb
Add follwing policies to the role (not ideal setup, just a shotgun approach):

AWSLambdaBasicExecutionRole-64...ef	(already exist, add the remaining 👇)
AmazonSQSFullAccess
AmazonElastiCacheFullAccess
AmazonDynamoDBFullAccess
AWSLambdaDynamoDBExecutionRole
AdministratorAccess
AWSLambdaSQSQueueExecutionRole
AWSLambdaInvocation-DynamoDB
AWSLambdaVPCAccessExecutionRole