decoder_agent

This decoder agent is to be used in conjunction with a decoder.

The agent's responsibilities are:

• Retrieving tasks from external Task Controller through http
• Downloading audio files to local storage for decoder to detect and begin decoding
  • Audio files will be downloaded to input/ directory
  • For Docker/Kubernetes deployments, Agent and Decoder will both need to access this same directory
• Receive decoding status updates from decoder and passes onward to Task Controller through http
• Uploading of transcriptions files to either AWS S3 or Azure Blob Storage
  • Transcription files are assumed to be in output/ directory
  • If audio file has the name abcdef.wav, the resulting output files are assumed to be in output/abcdef/ directory
  • The entire output/abcdef/ directory will be uploaded to the chosen Cloud Storage Providor
  • For Docker/Kubernetes Deployments, Agent and Decoder will both need to access this same directory
• For Docker images built with the Dockerfile, input/ and output/ directories' full paths are /usr/src/app/input/ and /usr/src/app/output/ respectively

Expected functionality of Decoder:

• Send status updates to agent via POST /status
  • Media type application/json
  • Request body should have these fields
    • { "status": string, "filename": string }
    • If filename provided by Decoder is different from the original filename obtained from the TaskController, Agent will take the new filename value when searching for the output transcription files
  • status: "DONE" will signal to the Agent that decoding process is complete and will begin uploading of transcription files to the chosen Cloud Storage Providor (Azure/AWS)
  • For all other statuses, Agent will simply forward them to TaskController
• Send error update to agent via POST /error
  • Media type application/json
  • Request body should have these fields
    • { "status": string }
  • Agent will forward error status to TaskController and begin cleanup
  • This signals to the Agent that decoding process for the current file has stoppped and is waiting for a new file to begin deocding process again

Environment variables:

AWS_ACCESS_KEY_ID=
AWS_SECRET_ACCESS_KEY=
AWS_BUCKET=

AZURE_ACCOUNT=
AZURE_ACCOUNT_KEY=
AZURE_CONTAINER=

# which cloud storage providor to use
# USE_STORAGE=azure OR USE_STORAGE=aws
USE_STORAGE=

PORT=                           # port number that the Agent will run HTTP server
TASKCONTROLLER_URL=             # URL of TaskController to receive tasks from
WORKER_QUEUE=                   # task's queue to filter for
WORKER_LANGUAGE=                # task's language to filter for

# OPTIONAL
DECODER_START_WAIT_TIME=        # time to wait for decoder to send "STARTING" status update, in milliseconds, defaults to '30000'
WORKER_SAMPLING_RATE=           # task's sampling rate to filter for, defaults to '16khz'
POLLING_PERIOD=                 # how often it polls TaskController for new tasks in milliseconds, defaults to '15000'

npm install to install needed dependencies to run locally.

node src/agent.js to run agent.

Build into docker image with Dockerfile to run in container.