DrChatPatin

A Decentralized AI-Powered Medical Assistant for Rare Disease Diagnosis on the Internet Computer Protocol

Abstract

DrChatPatin is a decentralized application (dApp) designed to assist healthcare professionals and patients in the preliminary diagnosis of rare diseases, following diagnostic criteria aligned with the U.S. Food and Drug Administration (FDA) standards. The system operates entirely on the Internet Computer Protocol (ICP), ensuring data immutability, censorship resistance, and persistent conversation storage through on-chain canister smart contracts. The frontend and backend components are co-deployed on the ICP network, eliminating reliance on centralized cloud infrastructure.

Overview

Rare diseases affect a disproportionately small patient population, which historically has led to limited diagnostic tooling and clinical decision support. DrChatPatin addresses this gap by providing an AI-assisted conversational interface capable of guiding diagnostic reasoning according to established FDA rare disease classifications.

Key properties of the system:

• Decentralized execution: Both the user interface and conversation storage logic are deployed as ICP canisters, ensuring no single point of failure.
• Persistent conversation history: Patient-provider interactions are encrypted, stored immutably on-chain via the backend canister.
• External AI integration: A dedicated API layer connects the application to an external AI model, enabling intelligent diagnostic responses.
• Voice interaction support: The application supports speech-to-text input, improving accessibility in clinical environments.
• Secure rendering: AI responses are rendered as sanitized Markdown to prevent injection vulnerabilities.
• Internet Identity integration via NFID: User authentication is handled by NFID (Identity Provider), providing pseudonymous, cryptographically secure login.

System Architecture

The application follows a three-layer architecture deployed on the Internet Computer mainnet:

The frontend canister serves the React-based user interface as static assets. The backend canister, written in Motoko, manages conversation state and persistent storage. The API layer handles communication with an external AI model responsible for generating diagnostic responses.

API

DrChatPatin includes a dedicated API layer that serves as the bridge between the ICP Canisters and the external AI diagnostic engine. This module is responsible for forwarding user queries, receiving model responses, and returning structured data to the frontend for rendering.

Responsibilities

• Receive conversation context
• Extract information from RAG (if enabled)
• Process model response (local or third party API uses)
• Parse and return AI-generated responses to the frontend

Configuration

API credentials and endpoint URLs can be managed via environment variables. Before running the project locally, ensure the following variables are defined in your .env file:

# AI model endpoint
AI_API_URL=<your_api_endpoint>

# API authentication key
AI_API_KEY=<your_api_key>

Technology Stack

Web3 / Frontend

Web3 / Blockchain

Web2 / API

AI / ML Stack

Runtime & Tooling

Deployed Canisters

The application is live on the ICP mainnet. The following canister identifiers correspond to the production deployment:

The live frontend is accessible at:

https://zu6ru-giaaa-aaaad-qhqba-cai.icp0.io

Repository Structure

DrChatPatin/
├── src/
│   ├── DrChatPatin_backend/
│   │   └── main.mo                  # Motoko backend canister (conversation logic & storage)
│   └── DrChatPatin_frontend/
│       └── dist/                    # Compiled frontend assets (served by asset canister)
├── canister_ids.json                # Mainnet canister IDs
├── dfx.json                         # DFX canister configuration
├── mops.toml                        # Motoko package dependencies
├── package.json                     # Node.js workspace configuration
├── package-lock.json
├── tsconfig.json                    
├── README.md
API/
├── API_DrChatPatin.py
└── README.md

Prerequisites

Before setting up the project locally, ensure the following tools are installed:

• Node.js >= 16.0.0 — Download
• npm >= 7.0.0 — Included with Node.js
• DFX SDK — The DFINITY Canister SDK. Installation Guide
• MOPS — Motoko package manager. Installation Guide

Local Development Setup

1. Clone the repository

git clone https://github.com/DvdRivas/DrChatPatin.git
cd DrChatPatin

2. Install Node.js dependencies

npm install

3. Configure environment variables

Create a .env file in the project root and add the required API credentials:

cp .env.example .env
# Edit .env with your API endpoint and key

4. Start the local ICP replica

dfx start --background

5. Deploy canisters to the local replica

dfx deploy

Once deployment completes, the application will be available at:

http://localhost:4943?canisterId={asset_canister_id}

6. Start the frontend development server (optional)

# Regenerate Candid declarations (recommended before starting)
npm run generate

# Start the development server
npm start

The development server runs at http://localhost:8080 and proxies all canister API requests to the local replica at port 4943.

Production Deployment

To deploy to the ICP mainnet:

dfx deploy --network ic

Available Scripts

from DrChatPatin_frontend/src/ directory

Environment Configuration

API credentials and endpoint URLs can be managed via environment variables. Before running the project locally, ensure the following variables are defined in your .env file:

# AI model endpoint
AI_API_URL=<your_api_endpoint>

# API authentication key
AI_API_KEY=<your_api_key>

Motoko Dependencies

Backend canister dependencies are managed via MOPS and declared in mops.toml:

References

• Internet Computer Protocol Documentation: https://internetcomputer.org/docs
• DFINITY Developer Quick Start: https://internetcomputer.org/docs/current/developer-docs/setup/deploy-locally
• Motoko Programming Language Guide: https://internetcomputer.org/docs/current/motoko/main/motoko
• Motoko Language Reference: https://internetcomputer.org/docs/current/motoko/main/language-manual
• MOPS Package Registry: https://mops.one
• NORD Rare Diseases Resources: https://www.fda.gov/patients/rare-diseases-fda
• Database Used to Create RAG: https://github.com/DvdRivas/NDB
• POE API: https://poe.com/api

Repository maintained by DvdRivas. Deployed on the Internet Computer mainnet.