flask-api-test

Flask API running on AWS

This simple test application allows user to extract linguistic features from a piece of text. The app uses AWS SQS messages to keep a queue of requested jobs; it also sends output reports to GCS in JSON format.

All of this was done for practicing purposes.

Build and Run

Flask

Useful for debugging. The Flask application is app.api.py

python -m app.process &
export FLASK_APP=app/api.py
flask run

This execution will expose the API through localhost:5000.

Docker

Run the script build.sh to execute the application using Docker and Gunicorn in production-like environments.

The application expects receiving several environment variables from the Docker environment. When using build.sh, you can define those variables from the host using the same names.

Environment variables

• PORT: The exposed port from the Docker container and to perform the binding with the host. Already defined in build.sh as the port 5000 through the host environment variable PORT.
• AWS_ACCESS_KEY_ID: AWS access key for managing the SQS queue.
• AWS_SECRET_ACCESS_KEY: AWS secret key for managing the SQS queue.
• SQS_REGION_NAME: SQS needs the region_name parameter at construction time of the client (boto3).
• SQS_QUEUE_URL: SQS URL access.
• SPACY_MODEL_NAME: spaCy model name of the model used to perform the text parsing. Already defined in build.sh as en_core_web_sm, which the only one available from the Docker environment.
• GCP_CREDENTIALS: GCP service account info (service account credentials JSON file content). Preferred to avoid storing the file in repo when building on the cloud.
• GCP_CREDENTIALS_FILE: GCP service account JSON file (useful when working in local). Not used in build.sh.
• GCS_OUTPUT_BUCKET: GCS bucket name for the output report.
• GCS_OUTPUT_PREFIX: GCS object prefix for the output report.
• LOG_LEVEL: Logging level of the app.process task. Default is ERROR (this default is also hardcoded in build.sh).

Endpoints

• /submit

  • method: POST
    • description: Creates a job to process a piece of text, maximum size is 256 KB.
    • content-type: x-www-form-urlencoded or through query string parameter.
    • data:
      • text: Piece of text to process.
        • type: string
    • return:
      • type: JSON
      • fields:
        • submitted: true if the job was successfully placed in queue.
          • type: boolean
        • id: job ID, which is also the same as the SQS message in queue.
          • type: string
      • status: 201 if success.

• /status

  • method: GET
    • description: Checks API status. Doesn't expect parameters.
    • returns:
      • type: JSON
      • fields:
        • processing_jobs: true if there are messages in the queue or in flight. type: boolean
        • queue_attributes: SQS attributes values now ApproximateNumberOfMessages and ApproximateNumberOfMessagesNotVisible casted to number. type: object
      • status: 200 if success.

Testing endpoints

• /message:
  • method: GET
    • description: Gets a message from the SQS queue.
    • returns:
      • type: JSON
      • fields:
        • message_received: true if message received.
          • type: boolean
        • sent_timestamp: time epoch when the SQS message was sent.
          • type: string
        • body: SQS message body, includes sent text and a timestamp taken when starting the submit process.
          • type: object
        • id: SQS message ID.
          • type: string
      • cookies:
        • receiptHandle: receipt handle ID to delete the message form the queue.
          • Path: /message
          • Max-Age: 600
    • status: 200 if OK, 204 if no messages (response empty).
  • method: DELETE
    • description: Deletes a message received from the SQS queue.
    • returns:
      • type: JSON
      • fields:
        • message_deleted: true if message deletion happened successfully.
          • type: boolean

Tests

• test_text_processing.py checks if the parsing process is working the same way for the parsed_text_sample.json. This test aims to work as a verification step on a CD workflow using GitHub Actions.

TODO

• Continuous Deployment to AWS ECS using GitHub Actions and AWS CloudFormation.