What shipped
- A dependency-free local Node.js dashboard named
lead-leak-recovery-desk. - CSV importers for calls, forms, calendar records, estimates, and manual notes.
- File-backed JSON workspace storage under
data/. - Deterministic normalization, leak classification, prioritization, and template generation.
- A browser dashboard with metrics, queue filters, lead detail review, status updates, draft actions, leak-cause reporting, booked revenue ledger, and CSV import controls.
- Markdown report and CSV queue exports under
exports/current/. - Sample data for a fake Nashville HVAC business.
- Service collateral for a
$2,500 Lead Leak Diagnostic. - A smoke test covering reset, classification, booking update, report generation, and export verification.
Architecture
- Used Node.js standard-library HTTP instead of Express to avoid install/network dependency risk while preserving the requested API surface.
- Kept business logic in dedicated modules: importer, normalizer, classifier, templates, reports, and store.
- Stored all workspace state in
data/workspace.jsonso the demo is inspectable and portable. - Used deterministic templates instead of live LLM calls to keep the recovery loop runnable without credentials.
- Made rule explanations visible on every queue item so owners can audit why a lead was flagged.
Trimmed scope
- No auth, billing, multi-tenant model, database, or live third-party integrations.
- No automated SMS or email sending.
- No PDF export.
- No LLM-assisted transcript analysis.
- No real customer data.
Limitations
- CSV schemas are intentionally narrow and match the committed samples.
- Duplicate handling is contact-key based and basic.
- Classification rules are transparent but not tuned on real-world client exports.
- Compliance guardrails are operational checks only and are not legal advice.
- The frontend is a local single-page dashboard, not a production SaaS shell.
Suggested next steps
- Add schema validation and import error reporting per CSV row.
- Add richer duplicate resolution across phone, email, and names.
- Add a reviewer note field and durable status history UI.
- Add adapter modules for CallRail, Twilio, Google Calendar, and common home-service CRMs.
- Add optional PDF rendering for the owner report.
- Add a fixture-based test suite for classifier edge cases.