Release Notes#

Version 0.4.5#

Bug Fixes

  • Capped pandas below 3.0 because pyEPR still uses the removed read_csv(delim_whitespace=...) argument when parsing Q3D exported capacitance matrices. Fresh installs could otherwise solve successfully in Ansys, fail while loading the matrix, and surface as a confusing NoneType simulation result.

  • AnsysSimulator.simulate now raises a clear RuntimeError if a low-level simulation path returns no result payload, instead of printing Simulation Completed Successfully! and returning None.

Version 0.4.4 (2026-04-19)#

  • Alpha Version 0.4.4

Driven-Modal HFSS Simulations

  • Added a new squadds.simulations.drivenmodal subpackage that provides a stable, production-ready API for HFSS driven-modal workflows on three SQuADDS geometry families: NCap interdigital couplers, qubit-claw structures, and quarter/half-wave qubit-cavity-feedline coupled systems.

  • Added typed request/result/spec models (CapacitanceExtractionRequest, CapacitanceExtractionResult, CoupledSystemDrivenModalRequest, CoupledSystemDrivenModalResult, DrivenModalSetupSpec, DrivenModalSweepSpec, DrivenModalLayerStackSpec, DrivenModalArtifactPolicy, DrivenModalPortSpec, DrivenModalRunManifest) so driven-modal jobs can be specified, serialized, and resumed without ad-hoc dictionaries.

  • Added an explicit, user-visible Qiskit-Metal layer-stack profile (squadds_hfss_v1) with thickness-only overrides; substrate/metal materials remain fixed by preset. Layer stacks are emitted as CSV alongside every run and re-attached on restart.

  • Added an AnsysSimulator.run_drivenmodal(request, *, checkpoint_dir=None, export_artifacts=True) entrypoint that initializes a checkpointed driven-modal run from a typed request without bypassing the existing AnsysSimulator lifecycle.

  • Added resumable artifact/checkpoint manifests (squadds.simulations.drivenmodal.artifacts) so a crash or restart does not require re-solving sweep points whose Touchstone/parquet artifacts are already on disk.

  • Added port-spec builders (build_capacitance_port_specs, build_coupled_system_port_specs, split_rendered_ports) that produce Qiskit-Metal-compatible port_list / jj_to_port payloads from declarative port mappings, with strict validation of component / pin / metadata inputs.

  • Added pure post-processing helpers for coupled systems (calculate_chi_hz, calculate_g_from_chi, calculate_kappa_hz, calculate_loaded_q, y_to_s, terminate_port_y) that operate on raw Y-parameter tensors and return SQuADDS-native Hamiltonian quantities.

  • Added admittance-to-capacitance helpers (capacitance_matrix_from_y, capacitance_dataframe_from_y_sweep, maxwell_capacitance_dataframe) for NCap and qubit-claw extraction.

  • Added a JJ-port admittance toolkit (squadds.simulations.drivenmodal.qubit_admittance) with parallel R || L || C JJ models, terminated-port admittance reduction (Schur-complement based, with explicit short-circuit handling), zero-crossing extraction of the linear qubit pole, and scqubits-backed transmon f_q / alpha extraction from saved Y33 data.

  • Added HFSS-data plumbing (parameter_dataframe_to_tensor, network_from_parameter_dataframe, write_touchstone_from_dataframe) that converts pyEPR parameter tables into dense complex tensors and exports first-class Touchstone artifacts via scikit-rf.

  • Added live-design helpers (create_multiplanar_design, write_qiskit_layer_stack_csv, apply_cryo_silicon_material_properties) that align driven-modal HFSS with the existing eigenmode/Q3D convention of cryogenic silicon (epsilon_r = 11.45, loss_tangent = 1e-7) instead of the AEDT room-temperature default.

  • Exposed the new request/result/spec models as top-level squadds.simulations exports so users can compose driven-modal runs without importing private modules.

AnsysSimulator Improvements

  • Hardened AnsysSimulator.update_simulation_setup against None, empty, and non-dict setup payloads in both the unknown-parameter discovery loop and the final update loop, so malformed entries are skipped with a warning instead of raising.

  • Made update_simulation_setup non-interactive friendly: when input() raises EOFError (e.g. CI, scripts, MCP), unknown parameters are auto-accepted with a clear console message instead of crashing.

  • Replaced non-ASCII glyphs in the AnsysSimulator console output (simulation plan banners, status checks, parameter prompts) with ASCII so Windows cp1252 consoles no longer mangle the logs.

  • Refactored setMaterialProperties to prefer the live HFSS renderer’s already-open project/design when patching cryogenic silicon, with safe ansys.aedt.core fallbacks; added DI-friendly materials_factory / hfss_factory hooks for testability.

  • Hardened material lookups against PyAEDT versions where Materials.exists_material(...) returns a boolean rather than the material object: helpers now try checkifmaterialexists first and only mutate values that actually expose permittivity / dielectric_loss_tangent.

MCP Server

  • Added a built-in Model Context Protocol (MCP) server that lets AI coding assistants (Claude, Cursor, VS Code Copilot, Gemini, Codex) interact with the SQuADDS database through the standardized MCP protocol.

  • 16 MCP Tools: Database browsing, design search (find_closest_designs), physics-based interpolation (interpolate_design), Qiskit-Metal code snippet generation (get_qiskit_metal_snippet), contributor info, and more.

  • 6 MCP Resources: Version info, citation, component lists, dataset summaries, and a comprehensive CPW layout guide (squadds://layout-guide) specifying impedance matching, feedline topology, and airbridge generation.

  • 3 MCP Prompts: Guided workflow templates for designing fab-ready qubit-cavity chips (design_fab_ready_chip), exploring the database, and finding optimal designs.

  • Run with uv run squadds-mcp (stdio) or SQUADDS_MCP_TRANSPORT=streamable-http uv run squadds-mcp (HTTP).

  • Works with Claude Desktop, Claude Code, Cursor, VS Code, Antigravity, Gemini CLI, and OpenAI Codex.

  • Full documentation: MCP_README.md and MCP_DEVELOPER_GUIDE.md.

  • Added CI workflows for MCP testing (nightly compatibility checks against latest PyPI SQuADDS) and automated sync notifications on new releases.

Refactor and Maintainability

  • Refactored large internal modules into smaller helper modules while preserving the public SQuADDS APIs.

  • Kept squadds.core.*, squadds.simulations.*, and contributor/database entrypoints as compatibility facades over the refactored internals.

  • Added characterization tests around the extracted helper modules to make future cleanup safer.

Testing and CI

  • Added a dedicated integration-smoke workflow job that runs tests/mvp_test.py on GitHub Actions with live Hugging Face access enabled.

  • Restored visible MVP smoke-test output in CI logs so collaborators can inspect the merged dataframe, closest-match result, interpolated design, and setup payloads directly.

  • Improved test logging for integration debugging by using explicit pytest tracebacks/output settings in CI.

Documentation

  • Added runnable driven-modal tutorial scripts to tutorials/:

    • Tutorial-10_DrivenModal_Capacitance_Extraction.py (NCap and qubit-claw capacitance extraction from driven-modal Y-parameters).

    • Tutorial-12_DrivenModal_Qubit_Port_Admittance.py (qubit Y33 extraction with model-sweep / feedline-termination sensitivity analysis).

    • Tutorial-13_DrivenModal_Combined_Hamiltonian_Extraction.py (single render, segmented qubit/bridge/resonator sweeps, end-to-end Hamiltonian comparison table).

  • Added a tutorials/export_docsite_notebooks.py utility that converts the # %% script tutorials into .ipynb notebooks for both the tutorials/ directory and the public docsite.

  • Published the two production-ready notebooks to the docsite tutorial index:

    • Tutorial-10_DrivenModal_Capacitance_Extraction.ipynb.

    • Tutorial-11_DrivenModal_Combined_Hamiltonian_Extraction.ipynb (docsite-facing renumbering of Tutorial 13).

  • Added a regression test (tests/test_drivenmodal_tutorial_docs.py) that fails if the shipped notebook artifacts disappear or lose the Tutorial 5-style title/license/footer cell structure.

  • Added MCP server unit tests (43 tests covering schemas, utilities, and server creation).

Dependencies

  • Added scikit-rf>=1.2 as a runtime dependency to support Touchstone export and Network-based post-processing for driven-modal coupled systems.

  • Updated the datasets floor and lockfile so live Hugging Face metadata with Json feature types remains compatible during MVP smoke tests.

  • Added optional mcp dependency group: mcp[cli]>=1.9, pydantic>=2.0.

Bug Fixes

  • SQuADDS_DB.create_system_df (and therefore any Analyzer-derived dataframe for the qubit+cavity coupled system) now inflates JSON-string sub-payloads in design_options_qubit / design_options_cavity_claw / design_options into real dicts. Previously, datasets that stored cplr_opts / lead / meander as JSON strings (current HuggingFace dataset schema) broke downstream workflows that mutate those fields via nested dict access (e.g. pred_df.design_options_cavity_claw.iloc[0]["cplr_opts"][key] = value).

  • Analyzer.find_closest now recomputes when any of the requested target_params columns are absent from the cached dataframe instead of raising KeyError mid-search.

  • QubitCavity.make_cpws now preserves caller-provided lead and meander overrides (using setdefault instead of unconditional assignment), and actually applies the previously-discarded min(fillet, spacing/2.1) clamp so meander geometries stop self-intersecting on the Windows/Ansys validation machine.

  • QubitCavity.make_qubit / make_cavity / make_cpws now use explicit key checks for cpw_opts / cpw_options and coupler_options / cplr_opts instead of bare except blocks, so the underlying KeyError is raised cleanly when neither alias is present.

  • Tutorial driven-modal flows now work around a pyEPR.load_ansys_project(...) path-duplication bug when QHFSSRenderer is initialized with both project_path and project_name.

  • Fixed a ScalingInterpolator regression where NumPy was not imported during interpolation scaling.

  • Fixed half-wave parquet output writing so reruns overwrite expected artifacts even when the destination directory already exists.

  • Fixed helper edge cases found during refactor review, including option-key detection, sweep chunk key preservation, optional merger-term handling, contributor token fallback behavior, and live dataset setup/design payload normalization.

Backwards-Compatibility Notes

  • QubitCavity.make_cpws default lead.start_straight changed from "75um" to "150um" (carried over from the QubitCavity geometry fix). Users who previously did not specify lead.start_straight in cavity_claw_options.cpw_opts.left_options will see a longer initial straight section in the meandered CPW. Pass an explicit lead.start_straight override to restore the prior geometry.

  • The fillet on the meandered CPW is now actually clamped to min(provided_fillet, spacing/2.1); previously this calculation was performed but its result was discarded. Geometries that relied on the unclamped fillet may render slightly differently.

—

Version 0.4.3 (2026-01-28)#

  • Alpha Version 0.4.3

Performance Improvements

  • Great speedup of halfwave cavity workflows (seconds now instead of minutes*!)

  • Replaced joblib with NumPy vectorization in Analyzer.find_closest, making database queries instant and eliminating overhead.

  • Added numba.prange support for true multi-core CPU utilization during parameter extraction.

New Features

  • AnsysSimulator is now stateful (stores device_dict).

  • Added update_simulation_setup(target, **kwargs) for granular hyperparameter updates:

    • Supports target=”qubit”, “cavity_claw”, “coupler”, “generic”, or “all”.

    • Intelligently maps targets to correct setup dictionaries based on system type.

    • Interactive confirmation for unknown parameters to prevent typos and for specifying more hyperparameters.

  • Added get_simulation_setup(target) to view current setup parameters in formatted tables.

  • simulate() now uses valid internal state if no argument is provided.

  • Transparency: Prints simulation hyperparameters securely before execution.

  • All three setups (setup_qubit, setup_cavity_claw, setup_coupler) now properly included in half-wave cavity dataframes.

  • Added update_design_parameters(**kwargs) for direct geometry modification.

Bug Fixes

  • Corrected N=4 (Quarter-Wave) hardcoding to N=2 (Half-Wave) in objects.py.

  • Resolved TerminatedWorkerError by removing process-based parallelism in favor of vectorization.

  • Fixed KeyError in run_eigenmode for NCap simulations.

  • Fixed SettingWithCopyWarning and linting errors in analysis.py.

Improvements

  • Cleaned up unused parallel processing methods and dependencies.

  • Improved code stability across operating systems (macOS, Windows, Linux).

  • Added rich for beautiful, colored terminal status outputs during simulations.

—

Version 0.4.2 (2026-01-27)#

  • Alpha Version 0.4.2

Bug Fixes

  • Fixed a critical Numba compilation error in half-wave cavity calculations by replacing incompatible Convert utility calls with JIT-compatible implementations.

—

Version 0.4.1 (2026-01-23)#

  • Alpha Version 0.4.1

Improvements

Bug Fixes

  • Fixed a critical issue in Ansys simulations by pinning the pyaedt dependency to a compatible version (0.23).

—

Version 0.4.0 (2025-01-09)#

  • Alpha Version 0.4.0 [MAJOR INFRASTRUCTURE RELEASE]

Breaking Changes

  • Migrated from setup.py to modern pyproject.toml (PEP 621)

  • Switched from qiskit-metal to quantum-metal>=0.5.0 (ARM64 compatible)

  • Removed Conda-based installation; now uses uv for package management

  • Python 3.10+ required (dropped 3.9 support)

  • NumPy pinned to <2.0 for quantum-metal compatibility

New Features

  • Full native Apple Silicon (ARM64) support via quantum-metal

  • Modern uv package management for faster, more reliable installs

  • Added ruff linting to development dependencies

  • Added PySide6 for modern Qt GUI support

  • Cross-platform CI/CD testing on Python 3.10, 3.11, 3.12

Performance Improvements

  • ~7x speedup for half-wave cavity Hamiltonian parameter calculations

  • Added LRU caching for transmon E01/anharmonicity calculations

  • Pre-compute unique EC values before parallel processing to avoid redundant matrix diagonalizations

  • Vectorized EC calculation using numba for array processing

  • Optimized value mapping using numpy searchsorted (O(log n) instead of O(n))

  • Reduced transmon calculations from ~16.5M to ~1934 unique values for typical datasets

Infrastructure

  • Replaced all CI/CD workflows with astral-sh/setup-uv action

  • Added uv.lock for reproducible builds

  • Updated Docker image to use uv instead of Conda

  • Modernized PyPI publishing with trusted publishing (OIDC)

  • Added ruff configuration in pyproject.toml

Dependencies Removed (Bloatware)

  • memory_profiler (unused)

  • addict (unused)

  • dask (transitive)

  • pyarrow (transitive)

  • cython (build-time only)

  • qutip (transitive via scqubits)

Dependencies Added

  • matplotlib (was missing)

  • shapely (was missing)

  • pyside6 (for quantum-metal)

  • ruff (dev dependency)

Documentation

  • Completely rewrote installation instructions for uv workflow

  • Updated developer notes with new development setup

  • Simplified getting started guide

  • Removed all Conda/environment.yml references

Bug Fixes

  • Fixed deprecated HfFolder import (replaced with get_token())

  • Fixed version consistency checks in prepare-release workflow

—

Version 0.3.7 (2024-03-19)#

  • Alpha Version 0.3.7

New Features

  • Added automated version bumping workflow

  • Added Palace tutorial with EPR calculations

  • Added ML interpolation notebook

  • Added SQuADDS WebUI tutorial

  • Added GDS post-processing files for DVK foundry

  • Added support for airbridges

  • Added DeepWiki integration for codebase chat

  • Added contribution portal integration

Documentation

  • Updated Tutorial 7 with EPR calculations and improved formatting

  • Updated Tutorial 8 with ML interpolation examples

  • Added Apple Silicon installation instructions

  • Fixed broken links in tutorials

  • Added Qiskit Metal Fall Fest ML interpolation tutorial link

  • Updated contribution guidelines and documentation

  • Added issue templates and improved issue management

Bug Fixes

  • Fixed branch name inconsistency in Docker workflow

  • Updated outdated GitHub Actions versions

  • Fixed GUI/design bug in Tutorial 7 leading to 4 ports on TL

  • Fixed KLayout version compatibility issues

  • Fixed broken links in documentation

Infrastructure

  • Added automated contributor list updates

  • Improved CI/CD pipeline with version bumping

  • Added automated release notes generation

—

Version 0.3.6 (2024-10-13)#

  • Alpha Version 0.3.6

Bug Fixes

  • Fixed bugs in the AnsysSimulator code

  • Added non local rst dependencies for contributor list

  • Added some more db access methods

—

Version 0.3.5 (2024-10-11)#

  • Alpha Version 0.3.5 [MAJOR RELEASE]

New Features

  • Added methods to Analysis to retrieve design parameters easily from any dataframe with design_options column

  • Added custom TransmonCross and JJ elements

  • Added methods for getting design_df from ML interpolations

  • Added some gds post-processing methods for gds files

Bug Fixes

  • Fixed bugs in the AnsysSimulator code

  • Minor bug fixes in string inconsistencies and better error messaging

Change Log

https://github.com/LFL-Lab/SQuADDS/compare/v0.3.4…v0.3.5

—

Version 0.3.4 (2024-09-27)#

  • Alpha Version 0.3.4

New Features

  • Fixed unicode error in local builds on Windows

  • Added measured device data API

  • Added more data columns to experimental device

—

Version 0.3.3 (2024-09-20)#

  • Alpha Version 0.3.3

New Features

  • Patched some bugs in simulation and contribution info query

—

Version 0.3.2 (2024-08-30)#

  • Alpha Version 0.3.2

New Features

  • Patched a bug in simulation of some half-wave cavity resonator systems

—

Version 0.3.1 (2024-08-16)#

  • Alpha Version 0.3.1

New Features

  • Added added methods for showing better contributions and measured device information (includes fabublox recipes)

  • Updated tutorials (1,3 and 3p5)

—

Version 0.3.0 (2024-08-12)#

  • Alpha Version 0.3.0

New Features

  • Added support for half-wave cavity resonators

—

Version 0.2.36 (2024-07-06)#

  • Alpha Version 0.2.36

Breaking Changes

  • support for NCap and CapNInterdigitalTee strings everywhere in HuggingFace and the codebase (`Ncap` will be dropped in the future)

  • select_coupler would no longer be supported. Use select_resonator_type instead

  • Analyzer object no longer requires SQuADDS_DB() in its constructor

New Features

  • Added the option to create lambda/2 resonators

  • Interpolation support for lambda/2 resonators

Improvements

  • Added chi as a query parameter

  • Better and more intuitive API

  • “hot reload” of Analyzer object

  • Updated documentation and tutorials

  • Added release-drafter for automated release notes

Bug Fixes

  • Fixed bugs in Simulation Code (simulation of CapNInterdigitalTee and half feature-half_wave_cavity)

—

Version 0.2.35 (2024-07-04)#

  • Alpha Version 0.2.35

Bug Fixes

  • Bug in accessing simulation data from HuggingFace has been patched

Documentation

  • Added FAQ for issue with accessing HF data on SQuADDS releases < 0.2.35

Version 0.2.34 (2024-07-04)#

  • Alpha Version 0.2.34

Bug Fixes

  • Addressed API to only show sim data configs

  • Fixed a bug in AnsysSimulator for setup_dict for qubit-cavity systems

  • Fixed bugs for geometries extracted from interpolator (and utils.py)

Documentation

  • Added tutorial3p5 files

  • Added README.md, wish_list.md, and docs/source/developer/index.rst from feature-half_wave_cavity (latest)

—

Version 0.2.33 (2024-03-14)#

  • Alpha Version 0.2.33

  • Bug fixes in ansys_simulator code for whole device sweep functionality

  • Added multiple helper/utility methods for ansys simulations

  • Methods added to clulate chi, full dispersive shift of the cavity

  • Updated requirements.txt and documentation

  • Added method to set GITHUB_TOKEN

Version 0.2.32 (2024-02-02)#

  • Alpha Version 0.2.32

  • Bug fix in ansys_simulator code

  • Fixed hyperlinks

Version 0.2.31 (2024-01-17)#

  • Alpha Version 0.2.31

  • Bug fix in contributor validation function

Version 0.2.3 (2024-01-17)#

  • Alpha Version 0.2.3

  • Bug fixes in simulator engine

  • Added sweep functionality to simulator

  • Updated Tutorial 2 to reflect changes in simulator code

  • Added functionality for adding to existing configurations

  • Completed Tutorial 3

Version 0.2.2 (2024-01-10)#

  • Alpha Version 0.2.2

  • Documentation added to the entire codebase

Version 0.2.1 (2024-01-10)#

  • Alpha Version 0.2.1

  • Bug fixes:

    • change “c” to “readout” in both code and database entries

  • Handled Warnings from pyaedt

Version 0.2 (2023-12-24)#

  • Alpha Version 0.2

  • Simulator functionalities added

  • Tutorial-2_Simulate_interpolated_designs added

  • Issues with automated docsite generator persist

Version 0.1.7 (2023-12-23)#

  • Alpha Version 0.1.7

  • Fixed issues with automated docsite generator

  • Standardized path imports in all files

  • Version to merge with simulator functionalities

Version 0.1.6 (2023-12-20)#

  • Alpha Version

    • Database hosted on HuggingFace

    • Pre-simulated data on TransmonCross, Cavity with Claw and Couplers only.

    • Closest pre-simulated design and interpolated design retrieval implemented

    • Interpolation logic based on our paper

    • Tutorials on basic usage, contribution, and simulation added

    • pypi package created