Cranial is a Framework and Toolkit for building distributed applications and microservices in Python, with a "streaming-first" approach to data pipelines, and built especially for services delivering predictions from online learning models, with a hope to be useful to many kinds of applications.
The machine learning components do not provide algorithms or models like SciKitLearn or Tensorflow or Spark or H2O, but instead provide wrappers so that models and pipelines created by these tools can be deployed and combined in standardized ways.
A slide deck with detailed diagrams of Cranial architecture can be found here: https://docs.google.com/presentation/d/131RK79w-Ls7uKuQocDcyEBXWDWABv6fXpaK_1THBG2Y/edit?usp=sharing
Learn about Enterprise Integration Patterns here: https://www.enterpriseintegrationpatterns.com/patterns/messaging/Chapter1.html
The Cranial Ontology is now formalized in OWL. Canonical: http://ld.chapmanmedia.com/cranial Github: https://github.com/tribune/cranial-messaging/blob/master/ontology/cranial
The PyPI packages are possibly out-of-date. Install from Github to be able to run latest examples.
What is it? A Library...
Modules for reading and writing data from various sources, built with a "streaming-first" approach.
What is it? A universal pipe application...
Provides a utility that listens for messages at some URI and relays them to some target. Currently optimized for user convenience and development speed over run-time performance.
$ cd cranial-messaging/bin/ $ echo "hello world" | ./cranial stdin:// file://./out.txt $ ./cranial pipe file://./out.txt stdout:// $ ./cranial pipe --response --echo file://./out.txt http://httpbin.org/anything $ echo "- also means stdin" | ./cranial pipe --response - httppost://httpbin.org/post $./cranial pipe kafka://broker.a,broker.b/topic # stdout is the default $ ./cranial pipe postgresql://your.host:5439/name/table?last_id=0 \ ssh://email@example.com:22022/file.json.bzip2`` $ ./cranial pipe db://your.host/name/table?driver=mysql \ hdfs://example.com/path/to/file.json.gz $ ./cranial pipe tweets://yourname:password@#someTag \ fb://yourname:password@ # Doesn't exist yet, but Easy to implement. $ ./cranial pipe --response out.txt http://httpbin.org/anything \ | ./cranial pipe - s3://bucket/piping-to-myself/responses.txt.gz $ ./cranial pipe --list # Get supported protocols $ ./cranial pipe --help
Distributed Application Tools
- "Messengers" (a.k.a Publishers) "Notifiers" (a.k.a. Transports) and "Listeners" (a.k.a. Subscribers) for asynchronous remote message passing, suitable for implementing Actor & Enterprise Integration Patterns.
- Pluggable Service Discovery, initially implemented for Marathon, and a a desire to implement peer-to-peer gossip as a default mechanism.
Wrappers & Adapters for common services and protocols
- Amazon Kinesis Firehose
- Python DBAPI2 Databases
- Celery (Incomplete & Deprecated in favor of Kafka)
- Apache Mesos and Marathon
Some Candidates for Future Notifier sub-modules?
Contributions welcome! #. Logstash #. Redis
Adapters: Wrappers to standardize interfaces to datastores.
Fetchers: Utilities for transferring Bytes or Records between different datastores, or between datastores and services. The current implementation does not match the OWL spec. Lots @TODO here.
KeyValue: A Dict-like interface to DBAPI2 and other datastores.
Questions, Suggestions, Support requests, trouble reports, and of course, Pull Requests, are all welcome in the Github issue queue.
# Clone the source code in editable form $ git clone firstname.lastname@example.org/tribune/cranial-messaging # Create & activate a virtual environment wiht the tool of your choice, for example:: $ mkdir cranial-dev && cd cranial-dev $ virtualenv -p python3 venv && source venv/bin/activate # You may need to install poetry manually: $ pip install poetry # ...and then run $ poetry install