Obsei Workflow Execution via GitHub Action

This repo Showcase Obsei workflow execution via GitHub actions. We have created this for DataTalks.Club's demo.

This repo represent following workflow -

• GoogleNews Observer search and fetch google news article with given query
• Then TextSplitter split article in small chunks so Classification Analyzer can process it easily
• Then Classification Analyzer classify chunks in given labels using HuggingFace zero shot models
• Then Inference Aggregator aggregate classification output for each chunks and compute final inference based on given aggregation function
• And finally Slack Informer send result to given slack channel using API

Screenshot of final result -

• Full article text
• Final article classification result

How to use it?

• Fork this repo
• Update workflow.yml based on your need (Note: Never add you credentials in yml file, instead use environment variable and fetch them via github secret)
• [Optional] Update workflow.py if you would like to add some custom code
• Add required credentials in the secret to prevent their exposure to public
• Update github action.yml along with required environment variables (For example scheduling at regular time or based on some event refer link for more detail)
• Spread Obsei love with the world :)

This link might help you to regarding how collect credetials of observers (Facebook, Twitter etc) and informers (Slack, Zendesk etc).