- NodeJS 18
- Express.js
- Docker / Docker-compose
- Webhook
- Twillio account with a whatsapp number (https://console.twilio.com/?frameUrl=/console)
- Open AI API KEY (https://beta.openai.com/account/api-keys)
- Whatsapp with twillio (https://www.twilio.com/docs/whatsapp)
- Integration NodeJS with twillio (https://www.twilio.com/docs/whatsapp/quickstart/node)
- Integration NodeJS with Open AI (https://beta.openai.com/docs/libraries/node-js-library)
- Twillio Integration
- Open AI integration
- Receive and send whatsapp message from twillio number
- Send response from Open AI for the question back to the number who sent the message to twillio number
First step : clone this repository with
git clone https://github.com/marciowinicius/chat-gpt-whatsapp.git
Second step : copy .env.example to .env on the project root dir
Third step : change the env vars on .env file that you already clone. Explain every env var: PORT is the port that you want docker-compose expose to your machine, by default the value is 3000 TARGET is the target of multi-stage of dockerfile. If you are on localhost you use dev. If you are on production server you use production OPENAI_API_KEY is the API KEY that you got from https://beta.openai.com/account/api-keys TWILLIO_FROM_NUMBER is the number that you got from https://console.twilio.com/us1/develop/phone-numbers/manage/active?frameUrl=%2Fconsole%2Fphone-numbers%2Fincoming%3Fx-target-region%3Dus1 TWILIO_ACCOUNT_SID is the account sid that you got here https://console.twilio.com/?frameUrl=%2Fconsole%3Fx-target-region%3Dus1 at the end of the page TWILIO_AUTH_TOKEN is the account token that you got here https://console.twilio.com/?frameUrl=%2Fconsole%3Fx-target-region%3Dus1 at the end of the page
Fourth step : run in the terminal
docker-compose up -d
Fifth step : Now you have to send a message to the number of the twillio that you got so for this example I got the number and you can send a message with this link on whatsapp https://api.whatsapp.com/send/?phone=16692607219&text=Hi. Just change the number on parameter phone of this link and you will send a message to the phone number that you got on twillio. Now twillio can send a message without a message template of whatsapp. If you don't make the contact before and try to send a message, you will not receive the message, because twillio can't send messages if you don't make the first interaction with the number.
Sixth step : Now you can you the webhook on your own machine The webhook on your machine is
curl --location --request POST 'http://localhost:3000/message' \
--header 'Content-Type: application/x-www-form-urlencoded' \
--data-urlencode 'To=whatsapp:+14155238886' \
--data-urlencode 'From=whatsapp:+556293163846' \
--data-urlencode 'Body=Quem é akon no mundo da música?'You can deploy the application and them config the webhook on twillio. For this you can choose any host that you want.
I chose a free NodeJS hosting here is the website : https://render.com/.
You can fork this project to your github account and use the free tier connected to github account. Don't forget to put the env vars
Since you deployed the app on render they will give you an URL to access your web application.
The URL will be like https://something-here.onrender.com.
And the webhook will be https://something-here.onrender.com/message With this webhook on your hands you will access the menu to config the webhook on the sender of twillio whatsapp here https://console.twilio.com/us1/develop/sms/senders/whatsapp-senders?frameUrl=%2Fconsole%2Fsms%2Fwhatsapp%2Fsenders%3Fx-target-region%3Dus1 Edit your whatsapp sender webhook url to https://something-here.onrender.com/message
Now you can interact with the number of twillio without simulate the webhook on your machine.