Mendly

Your personal health platform
Understand health topics, explore medicines, find nearby care, and track your health.

📋 Contents

🌟 Vision
✨ Features
🖥️ Live Demo
📊 Repository Activity
🏗️ System Architecture
🧠 AI Model
💾 Memory Architecture
🗺️ Roadmap
📁 Project Structure
🚀 Quick Start
📖 Usage Guide
🔧 Development
❓ FAQ
🤝 Contributing
📄 License

🌟 Vision

"Bridging the gap between health information and understanding — through conversation, memory, and care."

Most health apps either dump raw data on you or lock you into a single feature. Mendly bridges the gap — combining AI conversation, drug databases, location-based care discovery, and personalized health tracking into one seamless experience.

The Problem	How Mendly Solves It
❓ Hard to find reliable health info fast	🤖 AI chatbot answers instantly
💊 No easy way to check drug interactions	⚡ Interaction checker with FDA data
🏥 Can't find nearby care in emergencies	📍 Location-based hospital & pharmacy finder
📚 Scattered health bookmarks	🔖 Save & organize medicines + conditions
🌍 Different emergency numbers per country	🆘 Country-wise contacts with tap-to-call

✨ Features

Feature	Description
🤖 Elix AI Chatbot	Conversational AI that answers questions about diseases, symptoms, medicines & interactions
💊 Medicine Search	Browse FDA-approved drugs — uses, dosage, side effects & precautions
🩺 Medical Conditions	Disease profiles with symptoms, causes, treatment & prevention
⚡ Drug Interaction Checker	Check two medicines for conflicts & adverse reactions
🏥 Nearby Hospitals	Find healthcare facilities by search or geolocation
💊 Nearby Pharmacies	Locate pharmacies & medical stores near you
🆘 Emergency Contacts	Country-wise emergency numbers with tap-to-call
🔖 Saved Items	Bookmark medicines & conditions for quick reference
🔐 Auth System	Secure signup / login with JWT
📱 Works Everywhere	Responsive design that works on desktop, tablet, and mobile

🖥️ Live Demo

Service	URL
🌐 Frontend	mendlyapp.web.app
⚙️ Backend API	mendly-backend-0vyg.onrender.com
📖 API Docs	mendly-backend-0vyg.onrender.com/docs

📊 Repository Activity

🏗️ System Architecture

graph TB
    subgraph Client["📱 Client Layer"]
        User([👤 User])
        FE["Vanilla JS SPA<br/>(Firebase Hosting)"]
    end

    subgraph API["⚙️ API Layer"]
        BE["FastAPI Backend<br/>(Render)"]
        Auth[🔐 JWT Auth]
        Chat[🤖 AI Chatbot Engine]
    end

    subgraph Services["🧩 Service Layer"]
        FDA[📡 openFDA Client]
        KB[📚 Knowledge Base]
        NIM[🧠 NVIDIA NIM API]
        Mem[💾 Memory Manager]
    end

    subgraph Data["🗄️ Data Layer"]
        DB[(MongoDB Atlas)]
    end

    User -->|API Calls| FE
    FE --> BE
    BE --> Auth
    BE --> Chat
    BE --> FDA
    BE --> KB
    Auth --> DB
    Chat --> Mem
    Mem --> NIM
    Mem --> DB
    KB --> DB
    FDA -->|REST| OpenFDA[🏛️ openFDA]

    style FE fill:#0d2137,color:#fff
    style BE fill:#0d2137,color:#fff
    style Mem fill:#00bcd4,color:#fff
    style DB fill:#0d2137,color:#fff
    style NIM fill:#0d2137,color:#fff

🧠 AI Model

Attribute	Detail
Provider	NVIDIA NIM
Models	Llama 3.3 Nemotron, DeepSeek variants
Role	Health Q&A, symptom guidance, medicine info, emotional support
Prompt Strategy	System-prompted with medical disclaimer, context-aware memory injection
Fallback	Local knowledge base for offline / common queries
Temperature	0.3 (factual) — 0.7 (conversational)

Inference Flow

sequenceDiagram
    actor U as User
    participant FE as Frontend
    participant BE as Backend
    participant MEM as Memory
    participant NIM as NVIDIA NIM
    participant DB as MongoDB

    U->>FE: Types a message
    FE->>BE: POST /chat
    BE->>MEM: Fetch conversation history
    MEM->>DB: Retrieve past sessions
    DB-->>MEM: Session summaries + context
    MEM-->>BE: Assembled memory context
    BE->>BE: Construct system prompt
    BE->>NIM: API call with context
    NIM-->>BE: Streamed response
    BE->>MEM: Save response to history
    BE-->>FE: Stream tokens
    FE-->>U: Display response

💾 Memory Architecture

Mendly uses a hybrid memory system that balances conversation continuity with token efficiency.

Memory Layers

Layer	Scope	Storage	Retention
🟢 Working Memory	Current session messages	In-memory (Python dict)	Session lifetime
🔵 Episodic Memory	Recent conversations	MongoDB — `sessions` collection	30 days
🟡 Summarized Memory	Compressed long-term history	MongoDB — `memory_summaries` collection	Indefinite
🔴 Profile Memory	User preferences + saved items	MongoDB — `users` collection	Until changed

How Memory Works

graph LR
    subgraph Online["🟢 Working (In-Memory)"]
        WM[Session Messages]
    end

    subgraph Recent["🔵 Episodic (MongoDB)"]
        EM[Recent Sessions<br/>24h window]
    end

    subgraph Long["🟡 Summarized (MongoDB)"]
        SM[Compressed Summaries<br/>Key topics + mood]
    end

    subgraph Profile["🔴 Profile (MongoDB)"]
        PM[User Profile<br/>Saved Items + Prefs]
    end

    WM -->|Flush on session end| EM
    EM -->|Summarize every N sessions| SM
    PM -->|Inject on login| WM

    style WM fill:#00bcd4,color:#fff
    style EM fill:#0d2137,color:#fff
    style SM fill:#0d2137,color:#fff
    style PM fill:#0d2137,color:#fff

Context Assembly

When a user sends a message, the backend assembles context in this priority:

System prompt — Role, boundaries, medical disclaimer
User profile — Name, saved items, preferences
Recent memory — Last 10-20 messages from current session
Summarized history — Compressed key topics from past sessions
Knowledge base — Relevant disease/drug info if detected
Current message — The user's latest input

This keeps responses context-aware without exceeding the model's token window.

🗺️ Roadmap

gantt
    title Mendly Development Roadmap
    dateFormat  YYYY-MM-DD
    section Core
    AI Chatbot v1           :done, 2025-01-01, 2025-03-01
    Medicine Search         :done, 2025-02-01, 2025-04-01
    Drug Interaction Checker:done, 2025-03-01, 2025-05-01
    Nearby Hospitals        :done, 2025-04-01, 2025-06-01

    section Current
    Firebase Hosting        :active, 2025-06-01, 2025-08-01
    Memory System v2        :active, 2025-07-01, 2025-09-01

    section Upcoming
    Long-term Summaries     :2025-08-01, 2025-10-01
    Habit & Mood Tracking   :2025-09-01, 2025-11-01
    Community Discussions   :2025-11-01, 2026-01-01
    Crisis Detection        :2025-12-01, 2026-02-01

✅ Completed

AI chatbot with NVIDIA NIM integration
Medicine search via openFDA API
Drug interaction checker
Nearby hospitals & pharmacies (geolocation)
Emergency contacts database
Saved items / bookmarks
JWT authentication
Firebase Hosting migration
Responsive SPA (desktop + mobile)

🔄 In Progress

Memory system v2 (long-term summarization)
UI / UX polish

📅 Planned

📁 Project Structure

mediguide/
├── frontend/                    # Vanilla HTML/CSS/JS SPA
│   ├── index.html               # Single HTML entry point
│   ├── js/
│   │   └── app.js               # SPA router, all views, features
│   ├── css/
│   │   └── style.css            # Full design system, responsive
│   ├── assets/
│   │   └── icon/                # Logo variants (SVG)
│   ├── manifest.json            # PWA manifest
│   └── sw.js                    # Service worker
│
├── backend/                     # FastAPI backend
│   ├── app/
│   │   ├── main.py              # Routes, CORS, security headers
│   │   ├── auth.py              # JWT, bcrypt, user dependencies
│   │   ├── chatbot.py           # AI chat engine
│   │   ├── knowledge_base.py    # Disease/medicine data
│   │   └── openfda_client.py    # FDA API client
│   ├── requirements.txt
│   └── .env
│
└── README.md

🚀 Quick Start

Prerequisites

Requirement	Version	Link
Node.js	18+	nodejs.org
Python	3.12+	python.org
MongoDB	Atlas (free tier)	mongodb.com/atlas
NVIDIA API Key	—	build.nvidia.com

Setup

# 1. Clone
git clone https://github.com/Satendra90390/Mendly.git
cd Mendly

# 2. Backend
cd backend
python -m venv .venv
# Windows: .venv\Scripts\activate
# macOS/Linux: source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env  # add your keys
uvicorn app.main:app --reload --port 8002

# 3. Frontend (new terminal)
cd frontend
npx serve . -p 3000

📖 API docs at http://localhost:8002/docs 🌐 Open http://localhost:3000 in your browser

📖 Usage Guide

flowchart LR
    A[🔐 Create Account] --> B[📊 Dashboard]
    B --> C[💬 Chat with Elix AI]
    B --> D[💊 Search Medicines]
    B --> E[📍 Find Nearby Care]
    C --> F["Ask health<br/>questions"]
    D --> G[⚡ Check Interactions]
    E --> H[🏥 Hospitals / 💊 Pharmacies]

Feature	Location	How To Use
💬 AI Chat	Sidebar → Chat	Type your health question
💊 Medicines	Sidebar → Medicines	Search by name or condition
⚡ Interactions	Medicines → Checker tab	Select two drugs to compare
🏥 Hospitals	Sidebar → Hospitals	Allow location or search city
🆘 Emergency	Sidebar → Emergency	Pick country → tap to call
🔖 Saved	Sidebar → Saved	Bookmark from any medicine page

🔧 Development

Command	What It Does
`npx serve . -p 3000`	Start local dev server
`uvicorn app.main:app --reload --port 8002`	Start backend with hot reload

Conventions

Area	Convention
Frontend	Vanilla HTML, CSS, JavaScript — single-file SPA
Backend	Python FastAPI, async/await, Pydantic validation
HTTP	Native `fetch()` — no Axios
Auth	JWT in `Authorization: Bearer <token>` header
API	JSON, RESTful routes
Styling	CSS custom properties, mobile-first, design tokens

❓ FAQ

Question	Answer
Is Mendly a real medical service?	No. Mendly is an experimental software project for informational purposes. It does not provide diagnosis, treatment, or professional medical advice.
Can I use Mendly in an emergency?	No. If you are experiencing a medical emergency, call your local emergency services immediately. Mendly's emergency section provides contact numbers only.
Is my data private?	Conversation data is stored in MongoDB Atlas. We do not share or sell your data.
Do I need an API key?	To run the backend locally, yes — you need an NVIDIA NIM API key (free tier available). The live demo is pre-configured.
What AI model powers the chatbot?	Mendly uses NVIDIA NIM with Llama 3.3 Nemotron / DeepSeek models, fine-tuned via system prompts for health information.
Why the name "Mendly"?	Mend (to heal/fix) + -ly (friendly/serene) — a friendly companion for your health journey.

🤝 Contributing

flowchart LR
    A[Fork] --> B[Branch]
    B --> C[Code]
    C --> D[Commit]
    D --> E[Push]
    E --> F[Open PR]
    F --> G[Review]
    G --> H[Merge 🎉]

Principle	Guideline
🔒 Security	Never commit secrets or API keys
🧪 Testing	Verify changes locally before PR
📚 Docs	Update README for new features
🎨 UI/UX	Follow existing component patterns

📄 License

MIT License — Free to use, modify, and distribute.

Built with care using vanilla JS, FastAPI, and MongoDB

🐛 Report Bug · 💡 Request Feature

Name		Name	Last commit message	Last commit date
.github/workflows		.github/workflows
.wrangler/cache		.wrangler/cache
assets		assets
backend		backend
frontend		frontend
scripts		scripts
.gitignore		.gitignore
LINKEDIN.md		LINKEDIN.md
PATREON.md		PATREON.md
README.md		README.md
package-lock.json		package-lock.json
package.json		package.json

Navigation Menu

Repository files navigation

Mendly

📋 Contents

🌟 Vision

✨ Features

🖥️ Live Demo

📊 Repository Activity

🏗️ System Architecture

🧠 AI Model

Inference Flow

💾 Memory Architecture

Memory Layers

How Memory Works

Context Assembly

🗺️ Roadmap

✅ Completed

🔄 In Progress

📅 Planned

📁 Project Structure

🚀 Quick Start

Prerequisites

Setup

📖 Usage Guide

🔧 Development

Conventions

❓ FAQ

🤝 Contributing

📄 License

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

Mendly

📋 Contents

🌟 Vision

✨ Features

🖥️ Live Demo

📊 Repository Activity

🏗️ System Architecture

🧠 AI Model

Inference Flow

💾 Memory Architecture

Memory Layers

How Memory Works

Context Assembly

🗺️ Roadmap

✅ Completed

🔄 In Progress

📅 Planned

📁 Project Structure

🚀 Quick Start

Prerequisites

Setup

📖 Usage Guide

🔧 Development

Conventions

❓ FAQ

🤝 Contributing

📄 License

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages