Harden gemini_brain: env keys, retry/backoff, tests & CI

[Claude]

Removes hardcoded API keys, adds exponential backoff for rate limits, and establishes test infrastructure with CI.

Security

• Environment variables for secrets: All API keys now read from GOOGLE_API_KEY and PEXELS_API_KEY environment variables
• Files updated: gemini_brain.py, imagen_brain.py, scan_weapons.py, veo_brain.py, veo_proof.py, pexels_brain.py
• Configuration: Added .env.example template with required keys

Reliability

• Exponential backoff: gemini_brain.get_viral_script() retries 429 rate-limit errors with increasing delays (5s → 10s → 20s)
• Configurable: retries and backoff_factor parameters allow tuning

# Before: hardcoded key, no retry logic
API_KEY = "AIzaSy..."
response = model.generate_content(prompt)

# After: env-based key, automatic retry on rate limits
api_key = os.environ.get("GOOGLE_API_KEY")
for attempt in range(retries):
    try:
        return model.generate_content(prompt).text
    except Exception as e:
        if "429" in str(e):
            time.sleep((backoff_factor ** attempt) * 5)

Testing

• Mocked unit tests: tests/test_gemini_brain.py and tests/test_backoff.py validate API key handling and retry behavior
• CI workflow: .github/workflows/python-tests.yml runs pytest on push/PR
• Dev dependencies: requirements-dev.txt added with pytest

Build artifacts

• Updated .gitignore: Excludes generated media files (temp_*.mp3, VISUAL_*.png, etc.) and standard Python artifacts