The goal is to build a lean, web-based AI application that helps craftsmen and service providers in Switzerland create professional business texts and PDF documents quickly and easily. The product focuses on: • Speed & Simplicity: "One-click" document generation. • Mobile-First: The application must be fully optimized for use on smartphones (on the road/construction site). • Swiss Business Context: The AI must generate text in professional Swiss High German (e.g., correct terminology, polite tone, "ss" instead of "ß"). Important: This is a Minimum Viable Product (MVP). It focuses on immediate utility and usability. It is not a full accounting system or complex CRM. 2. Target Users • Profile: Craftsmen & service providers (1–20 employees). • Market: Switzerland (German-speaking). • Tech Affinity: Low. The UI must be extremely intuitive. • Use Case: Creating documents while in the van or quickly in the evening. 3. Core Functionality (The "Brain") 3.1 AI Chat Interface • UI: Simple chat interface (WhatsApp/ChatGPT style). • Input: User types instructions or speaks via microphone. • Speech-to-Text Engine (CRITICAL): • Must use OpenAI Whisper API. • Reason: It is the only commercially available engine that reliably understands Swiss German dialects and transcribes them into standard High German. • AI Logic: 1. Detects user intent (e.g., "Create offer", "Remind customer"). 2. Extracts data (Name, Date, Amount). 3. Selects the correct template. 4. Generates the content. 4. Supported Document Templates (MVP) The system must support specific workflows to protect the user's revenue and save time. Financial & Sales Documents: 1. Offer / Quotation 2. Order Confirmation 3. Invoice (Generation only) 4. Down Payment Request (Akonto) (Essential for material costs) 5. Confirmation of Additional Work (Nachtrag/Regie) (Crucial to avoid disputes over extra work) Communication & Issues: 6. Payment Reminder (Friendly tone) 7. Payment Reminder (Firm tone) 8. Appointment Rescheduling 9. Complaint Response (De-escalating tone) 10. Customer Rejection (Polite decline of an order) Marketing & Retention: 11. Project Completion & Review Request (Asking for a Google review) The "Joker" (Dynamic): 12. General Customer Letter (Smart Context) * Usage: For any situation not listed above (e.g., delivery delays, holiday greetings). * AI Logic: The AI analyzes the user's request, invents a professional Subject Line (Betreff), and writes the body text automatically. 5. User Profile & Global Settings (Crucial) To ensure efficiency, the user configures their "Sender Identity" once. These details are automatically applied to the Header/Footer of every generated PDF. Fields to store: • Company Name • Logo (Image Upload) • Owner Name / Contact Person • Company Address • IBAN / Bank Details • VAT Number (MWST-Nr) – Optional • Default Payment Terms (e.g., "30 Days net") The AI should never ask for this information in the chat; it must pull it from the database. 6. Invoice Logic • Content: Auto-fills Sender Data (from Settings) + Customer Data + Service Description + Amount + VAT (if set) + Bank Details. • Invoice Numbering: • Default logic: auto-incrementing. • Manual Override: The user must be able to edit the invoice number manually before saving/downloading (to prevent gaps in bookkeeping). • Disclaimer: Visual note on the screen that this is a document generator, not a certified accounting tool. 7. Document Output • Formats: • Plain text (for copy/pasting into Email/WhatsApp). • PDF Download: Clean, professional, minimal layout. • PDF Engine: Use a robust library (e.g., react-pdf or puppeteer) that handles page breaks correctly. Text must not be cut off at the bottom of a page. 8. Document History • Simple list of previously created documents. • Display: Document Title, Customer Name, Date. • Actions: • Re-open / View details. • Re-download PDF. • Delete. • No complex folder structures. 9. File Upload (Data Extraction) • Input: PDF, JPG, PNG. • Use Case: User takes a photo of handwritten notes or an old invoice. • Process: System uses AI (GPT-4 Vision) to extract readable text -> Shows data to user -> Generates new document based on it. 10. Customer Storage (Light Address Book) Simple storage to reuse customer details. • Fields: Company/Name, Contact Person, Email, Address, Tone Preference (Friendly/Formal). • Function: When generating a document, the user can select an existing customer to auto-fill the recipient address. 11. Authentication & Subscription • Auth: Email/Password or Magic Link (via Supabase Auth). • Payment: Stripe Checkout integration (Monthly Subscription). • Gate: Application access is restricted to active subscribers only. 12. Data Protection & Security • RLS (Row Level Security): Strict database rules ensuring users can only access their own data. • File Limits: Max 10MB per upload. • Deletion: Users must be able to delete their account and data. 13. Technical Expectations • Mobile-First Design Strategy: The UI must be optimized for touch and small screens (iPhone/Android) first, desktop second. • Suggested Stack: • Frontend: Next.js / React • Styling: Tailwind CSS (for speed and clean UI) • Backend/DB: Supabase (Postgres, Auth, Storage) • AI: OpenAI API (GPT-4o & Whisper) • Deployment: Vercel • Code Quality: Clean, maintainable code. 14. Deliverables 1. Fully functional Web App (deployed on Vercel). 2. Source Code (GitHub/GitLab repository). 3. Setup Documentation (How to run locally, environment variables). 15. Timeline • MVP delivery within 14 days. 16. Expectations from the Developer • Strong MVP mindset (Simplicity > Complexity). • Experience with OpenAI API and Prompt Engineering. • Experience with Mobile Responsive Design. • No "feature creep" – stick to this scope. 17. Definition of Success The project is accepted when: 1. A user can subscribe via Stripe and set up their profile (Logo/IBAN). 2. A user can speak a request in Swiss German, and the system outputs a correct High German PDF. 3. The system correctly generates the 12 defined templates, including the "Dynamic Letter" with an auto-generated subject line. 4. Invoices have a sequential number that can be manually edited. 5. The PDF layout is professional and handles page breaks correctly. Final Note This project prioritizes practical value for "non-tech" users. Usability on a smartphone and the quality of the "Swiss Business Tone" are the most important success factors