Integrate Hugging Face summarization pipeline

[justinmadison]

Summary

Add an abstractive summarization step so each article gets a short “TL;DR.”

Motivation

• Helps end-users absorb long articles quickly.
• Demonstrates multi-document summarization (e.g. daily digest).

Scope

None

Acceptance Criteria

• summarize(text) produces a concise summary (<200 chars)
• summarize_task(article_id) saves "summary" to the article record
• CLI summarize command runs without errors and prints confirmation
• Tests pass in CI and README clearly describes both execution paths

Additional Context

Details

• Category: nlp
• Priority: P1
• Estimate: 2d
• Dependencies:
  • Database connection module (nlp/db.py) in place
  • Articles already normalized and persisted

Tasks

1. Add dependencies
   • Add transformers and torch to /nlp/requirements.txt.
2. Core function signature
   • Define in /nlp/core.py:

     def summarize(text: str) -> str
     

3. Celery task hook
   • In /nlp/tasks.py, register:

     @app.task
     def summarize_task(article_id: str) -> str
     

4. CLI entrypoint
   • In /nlp/cli.py, expose:

     python -m nlp.cli summarize --article-id=<id>
     

5. Tests & documentation
   • Unit test that summarize() returns a non-empty string under 200 chars.
   • Test that summarize_task() updates the DB with a "summary" field.
   • Update /nlp/README.md with:
     • Installation steps
     • How to run the Celery task
     • How to invoke the CLI command