enviroCar Voice Model Delivery API ⚙

Voice-Model delivery API created for enviroCar: Voice Command wake-word detection feature, responsible for providing Language Models remotely.

Stack:

• NodeJS
• ExpressJS
• Winston

Installation

• Install dependencies

  npm i

• Create .env (take reference from .env-sample)
• Create assets folder in the project directory and add model files in it. Latest released model file must contain the word latest (Case Insensitive) in it's filename, so that it is served at /latest endpoint. Eg: Vosk-Model-Latest.7z

Deploy

• Start server

  npm run start

OR

• Build docker image

  docker build . -t <TAG>

• Run container

  docker run -d --name <CONTAINER_NAME> -p <HOST_PORT>:8000 --mount type=bind,source=$(pwd)/assets,target=/usr/src/app/assets <IMAGE_TAG>

Endpoints

Endpoint	Request Type	Description
/latest	GET	Returns the latest model file
/models	GET	Returns a JSON list of available models
/models/:name	GET	Returns model file with the filename name if it exists else returns a 404. Eg: <address>/models/Vosk-Model-Latest.7z returns Vosk-Model-Latest.7z model file
/refresh	GET	Refreshes the models list and returns the operation result (Useful when new models are added)

Logs

Logs are collected for every request and saved in two files in the project root dirctory:

• app.log
• error.log