Skip to content

ALchemist v0.3.2

Choose a tag to compare

View all tags
@calebcoatney calebcoatney released this 05 Feb 21:46
· 67 commits to main since this release
v0.3.2
a817e73

ALchemist v0.3.2 Release Notes

Release Date: February 5, 2026

Highlights

This release significantly expands the visualization and plotting capabilities available through the OptimizationSession API, introduces a staged experiments workflow for autonomous optimization, and adds real-time WebSocket events to the web frontend.

New Features

Visualization & Plotting (Core)

  • Comprehensive visualization module (alchemist_core/visualization/) — pure plotting functions decoupled from UI, usable in notebooks, scripts, and the web API
  • 3D voxel plot for model predictions across the search space
  • Acquisition function evaluation & visualization — plot EI, PI, UCB landscapes alongside model predictions
  • 3x3 visualization grid with dedicated acquisition styling (green color scheme)
  • User-configurable uncertainty calibration for regret/probability-of-improvement plots
  • Pareto frontier & hypervolume computation (Phase 1.1 of multi-objective Bayesian optimization)
  • Enhanced OptimizationSession plotting methods: contour, slice, regret, suggested-next, and more

Staged Experiments API

  • New /api/v1/sessions/{id}/experiments/staged endpoints for managing a workflow queue of pending experiments (stage, batch stage, retrieve, clear, complete with outputs)
  • Enables autonomous reactor-in-the-loop workflows where experiments are queued, executed, and completed programmatically

Web App: Real-time Events & Session Management

  • WebSocket-based session events replace polling — lock status changes, experiment additions, and model training are pushed to the frontend in real time
  • Session reconnection fix — URL ?session= parameter now takes priority over stale recovery backups
  • Join Session UI — paste a session ID on the landing page to connect to an existing session (useful for reactor clients)
  • Initial design panel can now stage DoE points directly into the workflow queue

Target Column Selection

  • CSV uploads now support selecting which column(s) to treat as target/objective variables
  • Flexible target column selection for multi-objective preparation workflows

Improvements

  • load_session() now supports both static and instance usage patterns
  • BoTorchModel updated to use float64 tensors for improved numerical stability
  • Corrected PI (Probability of Improvement) computation in OptimizationSession
  • Corrected sklearn acquisition functions for maximization objectives
  • Fixed best_f calculation and added batch acquisition support for plotting
  • Recovery-based session management refactored for reliability
  • Build hook project root path corrected

Infrastructure

  • Repository migrated from NREL to NatLabRockies GitHub organization

Full Changelog

v0.3.1...v0.3.2 — 21 commits

Assets 2
Loading