Bot(skeleton) for your Facebook page messages

• pip install Flask

• pip install requests

Need https for using facebook api. So if you are using local development server, install ngrok from Here

• Run ngrok : ngrok http 5000

Then note down the address. It will look something like this "https://ac433506.ngrok.io"

• Create a Facebook app with category as "Apps for pages".

• Then click "Add products" and add "Messenger".

• Select your page from dropdown and generate the access token. Note down the access token. You will need to provide it in "server.py".

• Click "Setup webhooks".

• But before setting up webhooks, run temp_server.py. Facebook will send a GET request to the callback URL we provide. The request will contain a custom secret we can add (while setting up the webhook) and a challenge code from Facebook. They expect us to output the challenge code to verify ourselves. To do so, we have written a quick GET handler using Flask in temp_server.py

• Add address obtained from ngrok as Callback url. Input something as Secret, our code doesn't care much about it. Tick message_deliveries, messages, messaging_optins and messaging_postbacks. Click Verify and Save. You will be back to previous page. In webhooks section, select page and click subscribe.

• Stop the temp_server.py

• Run server : python server.py

The code here accepts a message, retrieves the user id and the message content. It reverses the message and sends back to the user. For this we use the ACCESS_TOKEN we generated before hand. The incoming request must be responded with a status code 200 to acknowledge the message. Otherwise Facebook will try the message a few more times and then disable the webhook. So sending a http status code 200 is important. We just output “ok” to do so.

Try messaging to your page from facebook messenger. Good luck. 👍

This is just a skeleton. You can do a lot and improve a lot in this. Sky is the limit.

What I did : I downloaded (around 3000)tweets of bollywood actor Shah Rukh Khan using my GetTweets script. Then I used Markovify(a simple Markov chain generator) to generate random sentences from the tweets I have downloaded, and used those sentences as reply messages

@app.route('/', methods=['POST'])
def handle_incoming_message():
	data = request.json
	sender = data['entry'][0]['messaging'][0]['sender']['id']
	f = open("tweets.csv")
	text = f.read()
	f.close()
	model = markovify.Text(text)
	message = model.make_short_sentence(100)
	reply(sender, message)
	
	return "ok"

Hooked this into a page called "Talk to Shah Rukh Khan". Whenever user sends a message, page will reply back with random sentences imitating Shah Rukh Khan.