Back to home

🧾 TECHNICAL SPECIFICATION (TS)

Three AI-powered projects in the medical domain

for medical institutions

IMPORTANT: This is the project's primary source of truth — all development must strictly conform to this TS!

📌 Source: GitHub Issue #5 - Customer technical specification

🏗️ Unified architecture across all projects

Frontend stack:

  • - Web: React/Next.js for web interfaces
  • - Mobile: Flutter for mobile apps (iOS/Android)
  • - Terminal/Kiosk: React-based interface for kiosks

Backend stack:

  • - API Framework: FastAPI (Python) — single standard across all projects
  • - Database: PostgreSQL for all data
  • - Cache & Sessions: Redis for caching and sessions
  • - Message Queue: Redis/RabbitMQ for async operations

Security and compliance:

  • - OAuth 2.0 / OpenID Connect for authorization
  • - End-to-end encryption for all medical data
  • - Personal data under Kazakhstan law
  • - Medical audit logs with full traceability

Infrastructure:

  • - Docker containerization for all services
  • - API Gateway as a single entry point
  • - Real-time monitoring systems
  • - Webhook system for real-time notifications

AI & Core system:

  • - VARGATES AI CORE — centralized system for:
  • • Unified LLM Gateway (OpenAI / local medical models)
  • • Medical AI Models orchestration
  • • Prompt management and versioning
  • • AI performance monitoring and audit
  • • Medical data processing

1. Project: AI chat — Pre-doctor consultation

Project goal:

Build an intelligent chatbot to collect complaints, symptoms, and patient history, then generate preliminary findings and route the patient to the right specialist.

Core features:

  • - Dialog system to collect symptoms and medical history
  • - Complaint analysis using AI via VARGATES AI CORE
  • - Generation of a structured document (PDF/Word) with preliminary diagnoses
  • - Recommendations on which specialists to consult
  • - Integration with online appointment booking

Technical requirements:

  • - Interface: web chat on the institution's website (React/Next.js)
  • - AI model: integration via VARGATES AI CORE with medical models
  • - Security: personal data under Kazakhstan law
  • - Logging system for medical audit and fine-tuning
  • - Language support: Russian and Kazakh at minimum

Expected outcome:

Reduced load on therapists, intake, and front desk; faster initial patient routing.

System architecture:

Frontend:

  • - Web interface (React/Next.js) with mobile responsiveness
  • - Real-time chat interface with media file support

Backend:

  • - FastAPI server for chat processing
  • - Integration with VARGATES AI CORE for AI processing
  • - Redis for session caching and fast responses

Database:

  • - PostgreSQL for chat history, user profiles, and statistics
  • - FHIR HL7 compatible data structure

Integrations:

  • - API to the medical information system (MIS) via FHIR
  • - Document workflow system (PDF generation)

User Flow (user scenario):

  1. 1. User opens the website and starts the chat
  2. 2. AI via VARGATES AI CORE greets the user and clarifies the purpose of the visit
  3. 3. The bot sequentially collects data: complaints, symptoms, history
  4. 4. On completion the system analyzes the data and forms preliminary hypotheses
  5. 5. A report is generated and an appointment with a specialist is suggested
  6. 6. The user receives a link / QR code to download the document and book the appointment
  7. 7. Collected information is sent to the MIS via API

2. Project: AI — Virtual help desk

Project goal:

Develop an interactive AI-powered help desk to provide patients and visitors with information about the institution, services, schedules, and rules; print tickets and referrals; print patient routes; and provide application templates.

Core features:

  • - Answers to frequently asked questions (FAQ) about services, hours, prices, and routes
  • - Visualization — a virtual avatar of a medical worker (2D/3D)
  • - Voice and text interaction
  • - Site navigation and help with online booking
  • - Support on terminals / info kiosks

Technical requirements:

  • - Scenario-based and LLM-based approach
  • - Web interface (React) and a kiosk version
  • - Integration with the services and schedule database
  • - Automatic updates of reference information
  • - Voice synthesis and recognition (in Kazakh, Russian, and English)

Expected outcome:

Better-informed patients and fewer repetitive requests to staff.

System architecture:

Frontend:

  • - Web interface (React/Next.js) with a 2D/3D character
  • - Kiosk version for terminals
  • - Voice and text input support

Backend:

  • - FastAPI server for request processing
  • - Integration with VARGATES AI CORE
  • - Redis for caching frequently requested information

Knowledge base:

  • - PostgreSQL to store reference information
  • - Admin panel for content updates by administrators

Additional:

  • - Text-to-Speech and Speech-to-Text module
  • - Real-time updates of reference information

User Flow:

  1. 1. User reaches the help desk via screen or terminal
  2. 2. Asks a question by voice or types text
  3. 3. AI via VARGATES AI CORE analyzes the request and provides the answer
  4. 4. If needed, the user is redirected to the relevant page or to online booking

3. Project: AI — Personal doctor

Project goal:

Build an AI-powered personal assistant to monitor the user's health, analyze biometric data, and provide recommendations for lifestyle, prevention, and visiting a doctor.

Core features:

  • - Interpretation of received lab results in PDF
  • - Change analysis and detection of deviations
  • - Notifications and reminders (medication, doctor visit, etc.)
  • - Personalized advice based on the user's profile
  • - Ability to keep a wellbeing diary

Technical requirements:

  • - Interface: integration module inside a mobile app (iOS/Android)
  • - Cloud-based data processing
  • - Multilingual interface support

Expected outcome:

Better self-monitoring of health, prevention of chronic diseases, and fewer non-essential clinic visits.

System architecture:

Frontend:

  • - integration module for the iOS/Android mobile app
  • - Diary, vitals monitoring, and recommendations interface

Backend:

  • - FastAPI server with REST API
  • - Integration with VARGATES AI CORE for data analysis
  • - Redis for caching user data

Integrations:

  • - API to the MIS
  • - Push notifications for mobile devices

Data storage:

  • - PostgreSQL with end-to-end encryption
  • - Compliance with medical standards

User Flow:

  1. 1. User opens the Flutter mobile app
  2. 2. The chatbot connects to the MIS and patient data via FHIR API
  3. 3. AI via VARGATES AI CORE analyzes deviations and issues recommendations
  4. 4. The user receives personalized reminders and keeps a wellbeing diary

🔗 Artifacts

  • - TS web page: /technical-requirements
  • - Related EPIC: #1
  • - Customer: Medical institutions

📋 Status

  • - Status: Approved by the customer ✅
  • - Priority: Critical — foundation for all development
  • - Source: Original technical specification
  • - Updated: Tech stack standardization

© 2026 DreamLight Labs. Lonevi technical specification.

Source: GitHub Issue #5 - The project's primary source of truth

IMPORTANT: This is the project's primary source of truth. All development must strictly conform to this technical specification!