Project provides activity and outfit recommendations based on the weather in the city. City is provided as an input to the route.
- In any directory give: pip install virtualenv
- Create virtual environment by giving command: virtualenv <env_name> [env_name is the name of the virtual environment]
- To activate virtual environment run command: <env_name>\Scripts\activate
- Inside virtual environment: Go to project folder (where requirements.txt is present) and run command: pip install -r requirements.txt
- Update the OpenAI api key in the file at location: fastapi_server\src\main.py Here, update the parameter os.environ['OPENAI_API_KEY'] value. To fetch Openai API key go to, https://auth0.openai.com, login with valid credentials and then under the profile logo you should see option to ‘View API keys’. Here, you need to create new secret key and use this value for os.environ['OPENAI_API_KEY']
- After all the requirements are installed, give following command to run the server: Uvicorn main:app –reload
- The application will start running at: http://localhost:8000
- To see weather based recommendations, go to : http://localhost:8000/weather/{city}
The entry point for the project is the main.py where app() has been initialized. The route for getting weather, outfit and activity summary is /weather/{city}, get_weather_city() is called for this route which is defined in weatherapis.py. This makes call to OpenWeatherAPI and passes city as parameter.
OpenAI API is called for generating funny summary. Model used is text-davinci-003 from GPT-3.
From weatherapi.py -> outfit_recommendation_call() is called, which return the outfit recommended. For getting recommendations DecisionTreeClassifier is used. Input are provided through excel sheet, which can be uploaded using API endpoint /upload(define in src> routes> crud_excel.py). There is a template defined for this in src> templates> uploadfile.html. For uploading excel the file name should be outfit_table.xlsx Features are defined from the excel on basis of which the output i.e. recommended outfit is returned based on weather parameters.
For activity recommendation, get_activity_recommendation() is called from weatherapi.py whih is defined in src > sql_app> core.py This fetches recommendation from database. Sqlite is used as the database. Based on weather condition temprature, humidity and wind are classified and saved in database as high, moderate, low etc. When recommendation call comes, then based on the params the value is fetched from database.
There are API endpoint for adding database entries for activity recommendation i.e. /add_activity and to view all the entries for activity recommendations use /get_all_activity, API endpoint. These API endpoints can be used via SwaggerUI.
In src> sql_app folder:
- core.py: defines all API endpoint for adding activity recommendation and viewing them
- db.py: create session and db object
- models.py: Has the database model for the Activity data table.
- schemas.py: Contains Pydantic schema for APIs
src> cache.py: File that defines functions for caching openweather api response, which expires after every 2 hours
Unit test are defined in src> test_app.py Go to project folder and then to src, then run python -m unittest test_app.py
- /weather/{city} - To view weather summar, outfit and activity recommendation
- /upload - To upload excel for outfit recommendation, this is used by DecisionTreeClassifier. Name of excel should be outfit_table.xlsx
- /add_activity - To add activities for activity recommendation, can be accessed through http://localhost:8000/docs
- /get_all_activity - To view all entries for activity recommendation, can be accessed through http://localhost:8000/docs