Getting Started¶
ALchemist provides both a web application (browser-based) and desktop application (native GUI). Both interfaces offer the same functionality with nearly identical layouts and workflows. Choose whichever fits your needs - sessions are fully interoperable between both.
Launching ALchemist¶
Web Application¶
alchemist-web
Open your browser to http://localhost:8000
Desktop Application¶
alchemist
The desktop GUI launches directly.
Basic Workflow¶
All Bayesian optimization projects in ALchemist follow the same basic steps:
- Create or load a session - Your optimization project container
- Define variables - Set up your search space (see Variable Space)
- Add initial data - Import experiments or generate space-filling designs
- Train a model - Fit Gaussian Process to your data (see Modeling)
- View diagnostics - Check model quality (see Visualizations)
- Get suggestions - Run acquisition functions for next experiments (see Acquisition)
- Repeat - Add new data and iterate steps 4-6
Key Features¶
Session Management: Create, save, and load optimization projects. Sessions include your variable definitions, experimental data, trained models, and full audit logs.
Flexible Variable Types: Define continuous (bounded ranges), discrete (specific values), or categorical (named options) variables.
Data Import/Export: Add experiments manually or import from CSV. Export suggestions for lab execution.
Model Training: Choose between BoTorch (advanced, GPU-ready) or scikit-learn (lightweight) backends with automatic hyperparameter optimization.
Visualizations: Parity plots, Q-Q plots, calibration curves, and metrics evolution help you understand model performance.
Acquisition Strategies: Expected Improvement, Upper Confidence Bound, Probability of Improvement, and Thompson Sampling.
What's Different Between Web and Desktop?¶
Very little - they share the same core functionality:
| Feature | Web App | Desktop App |
|---|---|---|
| User Interface | Browser-based | Native GUI |
| Functionality | Identical | Identical |
| Sessions | Fully compatible | Fully compatible |
| Offline Use | Requires backend server | Works fully offline |
Choose the web app for remote access or team collaboration. Choose the desktop app for offline work or if you prefer native applications.
Where to Go Next¶
This guide covers launching ALchemist and the basic workflow. For detailed information on each step:
-
Variable Space Setup - Define your optimization problem
-
Model Training - Configure and train Gaussian Processes
-
Visualizations - Interpret model diagnostics
-
Acquisition Functions - Choose how to explore your space
-
Session Management - Save, load, and share your work
-
REST API - Programmatic access for automation