SculptFlowApp

MySculptFlow (SculptFlow)

Multi-tenant SaaS for plastic surgery clinics. A lead messages a clinic on WhatsApp or Telegram, an AI agent (orchestrated by n8n) answers from the clinic’s Knowledge Base, qualifies the lead and books consultations, and staff can take over any conversation from the dashboard Inbox. Clinics also run WhatsApp template campaigns (especially old-lead reactivation) against their existing leads.

For the full, current state of the project read PROJECT_HANDOFF.md first. It is the single source of truth for architecture, tables, endpoints, config keys, migrations, decisions and open TODOs. This README is only a short orientation.

Stack

• .NET 10, ASP.NET Core Razor Pages (staff dashboard) + Web API controllers (dashboard JS, n8n, Meta, Telegram). Project/assembly name is PlasticSurgery; only user-visible text says MySculptFlow.
• EF Core / Npgsql on Supabase PostgreSQL with pgvector. The schema is hand-written in Database/schema.sql (idempotent, appended blocks), not EF migrations; ApplicationDbContext is kept in sync by hand.
• ASP.NET Core Identity (role-free); every user belongs to exactly one clinic via clinic_users.
• SignalR (/hubs/inbox) for live Inbox, calendar and notification updates; PostgreSQL stays the source of truth.
• Hand-rolled CSS and small vanilla-JS files in wwwroot/ (no frontend framework).
• Deployed as a Docker image on Render (https://sculptflowapp.onrender.com) from main.

Main features

• Inbox (WhatsApp + Telegram) with AI / human modes, take over / return to AI, unread tracking, 24h WhatsApp window
• Leads, procedures, appointments month calendar, structured clinic availability, AI booking tools (get_available_slots, schedule_consultation, cancel_consultation, get_my_appointments)
• WhatsApp Embedded Signup, templates, health monitoring; Telegram bot channel
• Campaigns with audience picker (old leads who never booked, all contactable, custom filters, manual)
• Knowledge Base: manual entries, PDF/DOCX/TXT upload, website crawling, semantic search (its retrieval benchmark is run from the admin portal)
• Staff notification bell; Calendar Integrations (one-way Google/Outlook sync through a separate n8n workflow)
• Self-service registration (new user → new isolated clinic), Staff list, Clinic Info settings
• Subscriptions & prepaid usage billing (plans + entitlements, wallet, rate cards, billable events, ledger) — off until Billing:Enabled; see docs/billing.md

Local setup

1. Database. Run against a Postgres instance with pgvector (Supabase or local), in order:

   psql "<connection string>" -f Database/schema.sql
   psql "<connection string>" -f Database/seed-config.sql   # every setting with its default (never overwrites)
   psql "<connection string>" -f Database/seed.sql   # optional demo clinic, procedures and leads
   psql "<connection string>" -f Database/seed-demo.sql   # optional fully populated demo clinic (re-runnable, dummy data; sign-in and placeholder password in the file header, change it)
   

   Both are idempotent. Tables live in one Postgres schema per area (core, crm, billing, …); running schema.sql on a database created before 2026-10-07 moves its tables out of public (back it up first).

2. Secrets. From the PlasticSurgery project folder, use user-secrets (never commit real values):

   dotnet user-secrets set "ConnectionStrings:Postgres" "Host=...;Port=5432;Database=...;Username=...;Password=..."
   

   Use key=value format (not a postgresql:// URI) and, for Supabase, the session pooler host. The other secrets (Meta:*, N8n:*; the embeddings API key is set on the admin portal’s Configuration page) are listed in PROJECT_HANDOFF.md §20.

3. Run.

   dotnet build
   dotnet run --project PlasticSurgery --launch-profile https
   

   Open https://localhost:7276, register an account (this creates a new clinic) and you land on /dashboard. Swagger is at /swagger in Development only. There is no Razor runtime compilation, so .cshtml/.cs changes need a stop, build and run.

4. Tests. dotnet test PlasticSurgery.Tests. The billing database tests need SCULPTFLOW_TEST_DB set to a throwaway local PostgreSQL server (e.g. Host=localhost;Port=5432;Username=postgres;Password=...); each run creates and drops its own database. Without it those tests are skipped. Never point it at Supabase.

Known gaps

See PROJECT_HANDOFF.md §17 (pending work) and §22 (TODOs). The most important: rotate the secrets that were once in git history, derive clinicId server-side for the AI endpoints, verify Meta webhook signatures, add forwarded-headers handling and persistent DataProtection keys, and protect the anonymous POST /api/leads. Only the billing module has automated tests so far.