How zayedu/polycare works
Project Type and Purpose
• PolyCare is a full-stack health application combining a React Native/Expo mobile client (Android/iOS) with a Python backend.
• Purpose: aggregate three independent analyses (symptoms via LLM chat, blood‐test algorithm, ultrasound image ML) under a blackboard architecture to compute and display a PCOS likelihood score and recommendations.
File Structure
• Top-level folders:
• backend – Python microservices and modules
‑ UltrasoundAnalyzer (ML model, data folders, training script)
‑ account_system (CSV data, auth, CRUD account operations)
‑ calculateResults (modular calculators + blackboard logic)
‑ uploadResults (services to formulate discussion/narratives)
‑ api.py (entry point/REST API server)
‑ results_viewer_service.py
‑ data (Excel DB, CSV, prompt templates)
• pcos-app – React Native/Expo mobile application
‑ app (UI screens: Login, Home, data‐upload screens, results screen)
‑ android/ios folders (native project files)
‑ package.json, tsconfig.json
README for Architectural Insights
• Confirms use of a blackboard architectural pattern:
• Agents (LLM symptom chat, blood test analyzer, ultrasound analyzer) post partial results to a central “blackboard.”
• A controller/aggregator (overall_calculation) computes a weighted average.
• Key services: Account Management, Symptom & Result Upload, Likelihood Calculation, Result Viewer.
• External dependencies: TensorFlow/Keras for ML, a large language model (LLM) API (e.g., OpenAI), openpyxl for Excel, Expo/React Native.