LeadDesk AI: White-Label Chatbot and Lead Capture

PROJECT OVERVIEW LeadDesk AI is a production-ready, white-label website assistant for local service businesses. I built it to answer repetitive customer questions from an approved knowledge base and convert high-intent visitors into structured quote requests.

THE PROBLEM Small service teams lose ready-to-buy visitors after hours and spend valuable time repeating the same answers. Generic chatbots can also invent details, expose unclear sources, or collect enquiries that are too incomplete to follow up.

WHAT I BUILT • A branded customer-facing demo and embeddable website widget • Approved FAQ retrieval with visible source links and safe fallback behavior • Optional OpenAI responses with deterministic failover when the provider is unavailable • A structured quote form with validation, lead scoring, and SQLite storage • A private token-protected dashboard with lead statuses and formula-safe CSV export • Optional webhooks for Zapier, Make, n8n, Slack, or a CRM

PRODUCT WORKFLOW Visitor question → approved-answer retrieval → sourced response → intent detection → structured quote request → validation → dashboard, export, or webhook.

TECHNOLOGY AND DELIVERY The product uses Python, FastAPI, SQLite, HTML, CSS, JavaScript, Docker, and an optional server-side OpenAI integration. It is prepared for Railway deployment with persistent storage and a health endpoint. Buyer delivery includes a questionnaire, white-label configuration, knowledge-base setup, QA checklist, deployment guide, and operating documentation.

SAFETY AND PRODUCT DECISIONS Business facts come from buyer-approved content. Unknown questions receive a safe fallback. API keys and admin tokens remain server-side. AI response storage is disabled by default, and the application works without a paid AI API. The public demo uses fictional BrightHome Cleaning data and does not represent client performance.

VALIDATION Automated tests cover the application, retrieval behavior, AI fallback paths, and deployment configuration. The product is intentionally scoped for one small business deployment; PostgreSQL and multi-user authentication are recommended before high-volume or multi-tenant use.

LIVE PRODUCT Try the working demo at https://leaddesk-ai-production.up.railway.app/