myGenAssist Studio

Internal AI assistant for Bayer employees - a desktop companion app for myGenAssist, Bayer's internal AI platform.

Features

• Azure AD SSO - Seamless single sign-on with your Bayer credentials
• Enterprise Security - Secure authentication and data handling
• Multi-agent Workflows - Visual workflow builder with parallel agent execution
• Local File Access - Direct access to local files and documents
• Terminal Integration - Built-in terminal for development tasks
• MCP Tools - Extensible tool system via Model Context Protocol
• Offline Capable - Work with AI models even without constant network

Quick Start

Prerequisites

• Docker Desktop - installed and running
• Node.js >= 18.0.0
• npm or yarn

1. Start Local Backend

The local backend (PostgreSQL + FastAPI) is required for development.

cd server
cp .env.example .env  # If .env doesn't exist
docker compose up -d

• API server: http://localhost:3001
• API docs: http://localhost:3001/docs

2. Configure Environment

In .env.development, ensure these values are set:

VITE_USE_LOCAL_PROXY=true
VITE_PROXY_URL=http://localhost:3001

3. Start Frontend

npm install
npm run dev

4. Configure Model Provider

1. Sign in with your Bayer Azure AD credentials
2. Go to Settings > Models
3. In OpenAI Compatible section:
   • API Host: Auto-configured based on environment (dev/prod)
   • API Key: Auto-filled from SSO token
   • Model Type: claude-sonnet-4.5 (or other available models)
4. Click Save to validate and store

Architecture

Component	Technology	Purpose
Frontend	React + Vite + Electron	User interface and desktop app
Local Backend	FastAPI + PostgreSQL (Docker)	Provider config storage and API proxy
Authentication	Azure AD SSO (MSAL)	Enterprise single sign-on
Model API	OpenAI-compatible endpoint	Configured via VITE_MYGENASSIST_API_BASE

Development

Commands

# Start development server
npm run dev

# Run tests
npm run test
npm run test:watch      # Watch mode
npm run test:coverage   # With coverage

# Type checking
npm run type-check

# Storybook for component development
npm run storybook

Key Environment Variables

Variable	Description
VITE_USE_LOCAL_PROXY	Set to true for local development
VITE_PROXY_URL	Local backend URL (http://localhost:3001)
VITE_AZURE_CLIENT_ID	Azure AD application client ID
VITE_AZURE_TENANT_ID	Azure AD tenant ID
VITE_MYGENASSIST_API_BASE	myGenAssist API base URL (dev: https://dev.chat.int.bayer.com, prod: https://chat.int.bayer.com)

Building for Production

# Build for macOS
npm run build:mac

# Build for Windows
npm run build:win

# Build for Linux
npm run build:linux

# Build for all platforms
npm run build:all

Distribution Files

• macOS: myGenAssist-Studio.dmg
• Windows: myGenAssist-Studio.Setup.exe
• Linux: myGenAssist-Studio.AppImage

Support

• Email: mygenassist@bayer.com
• Documentation: https://docs.int.bayer.com/baychatgpt

Upstream Source

This project is built on Eigent, an open-source multi-agent AI desktop application. We maintain a thin wrapper approach to enable periodic upstream merges while keeping Bayer-specific customizations isolated.

License

Apache License 2.0 - See LICENSE for details.