PROJECT: AI-POWERED CONVERSATIONAL DOCUMENT CLOUD ACCESSIBLE VIA WHATSAPP 1. Overview The project consists of creating an intelligent document cloud, accessible primarily through WhatsApp, where users can ask anything related to the company (machines, documentation, spare parts, regulations, internal data, etc.), and the AI automatically returns the correct information, either as a text response or by delivering the exact PDF document required. This is not a traditional app and not a simple chatbot. It is the living memory of the company, organized in the cloud and accessed conversationally. 2. Entry Point: WhatsApp WhatsApp is the only access channel. The phone number identifies the user. There are no usernames or passwords. The system automatically recognizes: The phone number The associated company The assigned role Users can send: Text Photos (plates, parts, documents) PDFs Audio messages Videos 3. Roles and Permissions Each phone number has an assigned role, which defines the type of documentation it can access. Example roles Operator: operator manuals, technical datasheets Technician: technical manuals, exploded views, schematics Administration: insurance, registration documents, legal documentation Management: full access External / Client: limited access Permissions are applied by document type, not by folder. 4. Cloud Structure There is no folder per machine. Folders are global and organized by document type: /FOTOS_PLACAS /FOTOS_MAQUINAS /CE /MANUALES_OPERARIO /MANUALES_TECNICOS /DESPIECES /ELECTRICOS /HIDRAULICOS /FICHAS_TECNICAS /SEGUROS /PERMISOS_CIRCULACION /EXCEL All documents for all machines coexist in these folders. 5. Internal Entity: “Machine” Even without a machine-specific folder, the system creates an internal Machine entity to relate all information. Main fields Brand Model (normalized) Serial number / chassis number / internal ID Name variants Serial ranges (if applicable) This entity is used to filter and relate documents precisely. 6. Intelligent Indexing of the Cloud All content is automatically indexed. PDFs Internal text OCR per page (for scanned documents) Tables References Page numbers Images OCR of visible text Automatic classification (plate, part, schematic, general photo) Excel Sheet reading Columns Cells Relationships between data Each document is linked to one or more machines with: Confidence level Evidence (page or text where model/serial appears) 7. Machine Identification via Plate Photo This is the primary identification method. Flow User sends a photo of the identification plate OCR extracts brand, model, and serial Data normalization High-confidence machine identification Automatic filtering of all relevant documents If the serial number is unclear, the system works with brand + model and may request additional confirmation. 8. Document Search (Example: CE Certificate) Ideal case User: “I want the CE certificate for this machine” + plate photo System: Identifies the machine Searches in /CE Selects the correct PDF by model and/or serial Sends the CE PDF directly via WhatsApp If multiple options exist, the system asks for the serial number or a new plate photo. 9. Spare Parts Module via Photo Key rules A plate photo is always mandatory Without plate identification, no exact reference is returned Alternatively, the user may manually enter model and serial Full flow User sends plate photo System identifies the machine (MachineKey) User sends part photo (recommended: one loose part + one mounted photo) System searches only compatible manuals and exploded views Part identification using: OCR on the part Visual similarity (shape) Mounting context Comparison with exploded diagrams Result Exact part reference Exploded view PDF Exact page Item number (if available) Visual evidence If confidence is low, the system requests an additional photo. 10. Excel Search via Natural Language Users can ask questions such as: “How many hours can a bus driver work per day?” “What insurance does machine 1696 have?” The AI: Searches relevant Excel files Extracts the required data Responds in natural language No file opening required. 11. Company Phones and Contacts From WhatsApp, users can ask: “Company phone numbers” “I want to call insurance” The system returns: Correct person Role Direct phone number 12. Languages and Translation Automatic language detection Responses in the same language as the user Audio and video Automatic transcription Translation Response in configured language 13. Document Delivery Documents are delivered: As a PDF attachment via WhatsApp, or As a secure, time-limited link Optionally: Summary Relevant page Key extracted data 14. Development Phases Phase 1 – Internal Use (MVP) WhatsApp interface Roles and permissions PDFs and Excel Plate-based identification Document delivery Translation Phase 2 – Advanced Spare parts via photo Fault diagnostics ERP integration Multi-company (SaaS) 15. Final Definition An AI-powered intelligent document cloud, accessible via WhatsApp, where the system understands what the user needs, searches across all company documents, and delivers the exact file or correct information, with full permission control and traceability. 16. Development Philosophy: Scalable and Evolutive Project This project is not conceived as a finished product, but as a living platform, designed from day one to grow, improve, and integrate new functionalities continuously. Core principles Modular development: each function (documents, spare parts, Excel, translation, ERP, etc.) is an independent module Scalable architecture: ready to grow in documents, users, and companies without rebuilding the system Continuous improvement: the system learns from real usage Future integrations considered from the initial design This document defines the BEGINNING The scope described here represents: A solid functional starting point A minimum viable foundation for internal validation A first operational product that already delivers real value It is not the final state of the system. 17. Evolution Roadmap (Mid-Term Vision) Once internally validated, the system is prepared to progressively incorporate: New document types Improved image-based part recognition Advanced fault diagnostics ERP and external system integrations Automated supplier ordering Usage analytics and performance optimization Multi-company SaaS model Each phase will be driven by real usage and detected needs, not by a closed development plan. 18. Key Message for the Developer The goal is a flexible, well-structured, and scalable development, where the initial objective is not to build everything, but to build the foundation properly, knowing the system will grow day by day. The developer must clearly understand that: This project will evolve continuously New data sources and integrations will be added The architecture must enable change, not block it