apf-base 1.0.2

ALeRCE Alert Processing Framework.

Alert Processing Framework (apf)

apf is a framework developed to create a dockerized pipeline to process an alert stream, that can be easily be deployed in a local machine or distributed using Kubernetes.

First developed to process ZTF data it is capable to be used for any stream/static data processing pipeline.

Installing apf

apf installation can be done with pip

pip install apf_base

This will install the apf python package and apf command line script.

apf design

apf is based on steps conected through Apache Kafka topics.

Each step is composed by a consumer and is isolated from other steps inside a docker container.

When running, the step calls the execute() method for each message consumed. A step can have multiple producers and databases back-ends plugins that can be accessed inside the execute method to have a more complex logic.

This generic step greatly reduce the development of each component of the pipeline and make it easier to test each component separately.

apf features

• Automatic Metric Sender (KafkaMetrics)
• Automatic Code Generation (apf new-step <step_name>)
• Multiple Consumer Plugins:
  • Kafka
  • AVRO
  • CSV
  • JSON
• Producers:
  • Kafka
  • CSV
• Metrics:
  • Kafka

apf Quickstart

A quick-start guide to create a new step can be found here.