Skip to content

Build a Telegram chatbot enhanced by OpenAI's GPT-3.5 to interact with clients and use BookingAPI to suggest suitable hotels.

License

Notifications You must be signed in to change notification settings

Lafifi-24/HotelFinderAssistant

Repository files navigation

Introduction

This project is a chatbot assistant on Telegram using OpenAI API and langchain. It collects information from users to suggest hotels based on their preferences.

Application schema

There are two versions. In the first one, I used a Flask server. In the second one, I deployed the bot using AWS Lambda, AWS ECR, and AWS DynamoDB.

First version

Alt text

Second version

Alt text

Demonstration

In the video, you will notice the chatbot starting by welcoming the user with their Telegram username.Then it begins to ask about user preferences, such as city, check-in, and checkout... . Once it has collected the required information, the system will utilize the Booking API to present hotel options.

demo.1.mp4

Run the app

  • First step create your telegram Bot
  • Second Create the backend:
    • falsk server :

      • Fill api_keys.sh with your APIs tokens
      • Build docker image docker build --no-cache -f Dockerfile.flask -t BookingMate .
      • Run the docker image docker run -p 5000:5000 BookingMate
      • Congrats!
    • aws lambda serverless:

      • Build docker image docker build --no-cache -f Dockerfile.lambda -t BookingMate .
      • follow instructions in deployment part from aws documentation
      • Create lambda function using docker image
      • Add AmazonDynamoDBFullAccess policy to the role.
      • Add Environment variables from api_keys with your tokens
      • Change Memory=1024 MB and Timeout = 1 min
      • Open this link in your browser https://api.telegram.org/bot<Bot_token>/setWebhook?url=<API_GateWay_link>, be sur that you get the message "webhook was set"

Tests

After installing requirments and fill and execute api_keys.sh, you can try to excute this tests from tests folder.

- automated tests:
    contains tests created using pytest to ensure each component from the application works as expected 
    To run it : `python -W ignore -m pytest tests/automated_tests/`

- manual tests : it has two files 
    - test_chatbot_manager : is a test for the bots perfomence with different scenarios, it helps to improve the prompt for Bots 
    to run it : `python -W ignore tests/manual_tests/test_chatbot_manager.py`
    - test_telegram_bot : is a test for TelegramBot class if it has the ability to interact with user side 
    to use it : 
        - first : execute `python tests/manual_tests/test_telegram_bot.py`
        - second :  use telegram account to communicate with your bot

About

Build a Telegram chatbot enhanced by OpenAI's GPT-3.5 to interact with clients and use BookingAPI to suggest suitable hotels.

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages