Voice Assistant for handling restaurant related requests, developed as a student project.
Handles chosen restaurants in the area of Tricity, Poland. List od restaurants and their ? is available in the ./restaurants_data/all_restaurants_info.csv file.
The bot should allow for 3 main actions:
- Recommending the best suited restaurant based on at least city and cuisine
- Cities available to choose from: "Gdańsk", "Gdynia", "Sopot",
- Cuisines available to choose from: "Italian", "Turkish", "Indian", "Chinese", "sushi", "pizza", "Mexican", "Japanese", "French"
- Reserving a table in the restaurant (fictional reservation in a project database)
- Informing about already made reservations
Assistant is intented to use as localy deployed voicebot. The setup scripts are prepared for a UNIX based systems. To run the assistant having a working speakers and microphone is necessary! Also, the assistant is using Google Text-To-Speech API. In order to use there there is a need to:
- Have an account on Google Could Platform
- Have Cloud Text To Speech API enabled - you can check it here
- Create a Service Account - the instruction can be found here
- Generate Service Account API key in JSON format - the instruction to do this can be found here
- Save generated json file in the main application folder with the name tts_sa.json
- Create a python3.10 environment
- Run a
setup.shscript to install system dependencies - Install requirements from
requirements.txtfilepip install -r requirements.txt - Run
python -m spacy download en_core_web_mdin order to download entity recognition model required for the chatbot. - Run
docker compose upin the main project directory - Run
python va_main.pyin the main project directory in separate terminal window.
Currently the project was not testes on the Windows environment, due to not having access to such operating system at the moment. However, from my research it seems that the steps should be similar. The only change is in step 2 - there is no need for running setup.sh file. Insted interested user should make sure to have ffmpeg library installed
Docker componse creates the environment for rasa chatbot. It consists of 4 services:
- duckling server for entity extraction - from rasa/duckling
- actions server that hosts rasa actions - build by dockerfile in /rasa_chatbot/actions, stored on aszych/iu_rasa_actions
- rasa server that hosts rasa chatbot model - build by dockerfile in /rasa_chatbot, stored on aszych/iu_rasa_chatbot
- postgres server that host PostgreSQL database with restaurant and reservatios - build by dockerfile in /rasa_chatbot/actions/database, stored on aszych/iu_rasa_database
Those services allow main assistant loop to use chatbot model developed in Rasa software via API.
Since voice assistant is based on rasa chatbot, there is a possibility to run and "talk" with the chatbot itself, skiping the voice component. To do this, one can either:
- run docker compose up
- query rasa API using
curl -XPOST http://localhost:5005/webhooks/rest/webhook -H "Content-type: application/json" -d '{"sender": "test", "message": "Hi. What can you do?"}'
OR
- in the /rasa_chatbot/config.yaml file, comment line 31 and uncomment line 32. It should set duckling endpoint to localhost
- start duckling server separately, running
docker run -p 8000:8000 rasa/duckling - in the /rasa_chatbot/actions/database/db_config.py change to 'host': 'localhost'. It should set connection to postgres on localhost
- start postgres server, by running
docker run -d --name shared_postgres_container -e POSTGRES_USER=rasa_user -e POSTGRES_PASSWORD=rasa_password -e POSTGRES_DB=rasa_db -p 5432:5432 -v rasa_postgres_data:/var/lib/postgresql/data aszych/iu_rasa_database:0.0.x - in the /rasa_chatbot/endpoints.yaml file uncomment lines 13-14 and comment lines 16-17. It should set actions endpont to localhost
- start action server separately
docker run -p 5055:5055 aszych/iu_rasa_actions:0.0.x - start rasa chatbot in the shell, from the /rasa_chatbot folder run
rasa shell