What shipped
Built a standard-library Python CLI package named local_ai_visibility_watchboard for generating local AI visibility packets from local business evidence. The CLI supports:
samplewatchboardvalidate-watchboard
The seeded demo creates three business folders: east-nashville-dentist, brentwood-hvac, and germantown-cafe. The generator writes the required Markdown, CSV, and JSON watchboard outputs under watchboard/.
Architecture
- Reused the parent watchboard pattern: local folder input contract, deterministic scoring, generated packet artifacts, JSON export, and strict validation.
- Kept the stack Python 3.11+ standard library only.
- Kept all evidence local and human-inspectable.
- Used simple scoring rules instead of opaque heuristics so agencies can explain urgency bands to owners.
- Put compliance warnings directly into JSON and review-reply outputs.
Trimmed scope
- No hosted UI, database, scheduler, auth, billing, CRM, email sending, posting, scraping, OAuth, browser automation, or live integrations.
- No Google Business Profile, Gemini, OpenAI, Claude, Anthropic, OpenRouter, GitHub, or other provider/API calls.
- No ranking guarantees or automated profile publishing.
Limitations
- Input parsing expects the documented JSON and CSV shapes.
- Theme extraction uses supplied
themesfields rather than NLP. - AI/search visibility is based only on manually supplied snapshots and prompt metadata.
- Review replies are conservative drafts and still need human review for tone, facts, and platform rules.
Suggested next steps
- Add more robust import adapters for common review export formats.
- Add a diff mode for weekly or monthly refreshes.
- Add a white-label agency cover sheet and client checklist.
- Add optional HTML rendering after the packet proves useful in buyer tests.