zale144/instagram-bot

Microservices with Go, Docker and Kubernetes

Instagram Bot

Instagram Bot is a concept app that demonstrates how you can promote your website/platform on Instagram.

Initially I have written it as a monolithic web application, but I've rewritten it to follow the Microservice architecture.

Features

• Send invites to your website to Instagram users
• Batch process users based on a hashtag
• Follow Instagram users
• No account creation, login with your Instagram credentials
• Uses face detection to select a suitable background image

Microservices

• web - Go
• api - Go
• session - Go
• htmlToImage - Go
• faceDetect - Python

Technologies/libraries used

• Golang
• Micro
• gRPC
• Kubernetes
• Docker
• My fork of Goinsta
• Open Graph
• Labstack Echo
• Python
• OpenCV
• wkhtmltopdf
• PostgreSQL
• Consul (before moving to Kubernetes)
• React.js
• Gorm
• dep
• Bootstrap

Send invites to your website to Instagram users

Once you login you will see a list of all Instagram users that you follow. You can use the search bar to filter them. After clicking on a specific user, they will receive a Direct Message on Instagram with a link to your (TODO) website registration page, and in the link an OG tag will be embedded, containing an image of what their profile on your website would potentially look like. It will show their pictures and basic info as it is on their Instagram profile.

To achieve this,

• A request is sent to the api microservice to process the user by username
• The api microservice sends a request to htmltoimage microservice to create an image of your HTML template with that user's pictures and info applied to it.
• htmltoimage does this by executing the wkhtmltoimage binary and passing the image parameters and the URL to the fake HTML profile as input
• wkhtmltoimage consumes the URL, runs a detached browser to render the fake HTML profile, and generates an image out of it
• The fake HTML profile will be served from the web microservice.
• When requested, web composes that page by applying user info it fetches from session to an HTML template (in the future this template should be customizable)
• session gets basic info from the user's Instagram profile, like the profile image, biography, username, full name, media feed, from Instagram API, via goinsta.
• session selects the background image according to criteria: it's landscape, it's a picture of the user, or at least it's a picture of a person
• To detect if it's a picture of a person, it calls the facedetect microservice with a link to the image.
• The image created by htmltoimage is returned to api and saved as a file.
• Then, api continues by calling session to send an Instagram Direct message to the user with the link to a page that should (TODO) redirect to your registration page, with that generated image of the fake profile as an OG tag, which will show up nicely in the message.
• This page will again be served from the web service.

Batch process users based on a hashtag

By navigating to the Jobs tab, you will be able to create and view batch processing jobs. Issuing a job requires you to provide an Instagram #hashtag, limit and a message. All users associated with the hashtag, within limit provided, will then receive a Direct message on Instagram with the invitation to your website, with a custom message. Meaning, they will be processed just like this.

To achieve this,

• A request is sent to api, to create a Job.
• The job will be saved to the database and run in the background, so you don't have to wait around.
• The job will:
  • Request session to get all users by hashtag within provided limit.
  • Process them each by each if they fulfill two conditions: they're not processed yet, the user is not you.
  • Save the processed users the the database

Follow Instagram users

The search bar can be used to find Instagram users that are not in your followed list, and if found, you can follow them by clicking on the Follow button.

Login with your Instagram credentials

In order to login, simply provide your Instagram account credentials (they're not secretly saved to a database). The login request happens like this,

• A request is sent to web service to create a cookie for you, and you will gain access to the app's main page.
• web sends a call to session to create a session for you
• session calls the Instagram API to log you in and cache the session
• web calls api to create a JWT token for you
• The JWT token is then saved to your browser's local storage. This is how you will have authorization to access the api as a web service
• If successful, you will be redirected to the app's main page

Background image selected by face detection

Besides the username, full name, profile picture, biography, session gets also user's media from the Instagram API. A background image for the fake profile will be chosen from those media. To choose a suitable background image, criteria will be respected, such as:

• It's a landscape image,
• It's a picture of the user (if user is tagged in the picture)
• It's a picture of a person.

To detect faces in the image,

• session calls the facedetect microservice by RPC with a URL to the image as a parameter.
• facedetect runs a simple RPC server written in Python and uses machine learning with OpenCV for detecting faces in images.
• facedetect returns the number of faces it found in the image from the provided URL.