feat: BoTTube parasocial hooks for audience-aware agents (issue #2286)

[HuiNeng6]

Summary

Implements parasocial hooks that let agents notice and adapt to audience context in BoTTube flows.

Features

1. Audience Tracking (�udience_tracker.py)

• Track viewer/commenter history per agent
• Classify viewers by status (NEW, REGULAR, SUPERFAN, CRITIC, RETURNING)
• Calculate comment sentiment
• Persistent storage in JSON format

2. Personalized Responses (comment_responder.py)

• Status-based response templates
• Boundary enforcement (no creepy/desperate language)
• Rate-limited personalization (~30%)
• Batch response generation

3. Community Shoutouts (description_generator.py)

• Top commenter mentions in video descriptions
• Inspiration acknowledgments
• Weekly community summaries
• Milestone messages

Viewer Statuses

Status	Criteria
NEW	First comment within 7 days
REGULAR	3+ videos commented
SUPERFAN	5+ videos + positive sentiment
CRITIC	3+ videos + negative sentiment
RETURNING	Back after 30+ day absence

Boundaries Enforced

• No creepy phrases (tracking viewing times, specific patterns)
• No desperate language (pleading for return visits)
• Natural frequency - not every comment gets personalized

Testing

Run: \python -m pytest tests/test_parasocial_hooks.py -v\

Note: Test file being added via separate commit due to upload size.

RTC Payout Wallet

9dRRMiHiJwjF3VW8pXtKDtpmmxAPFy3zWgV2JY5H6eeT

Closes

#2286