wish_list.md

To Do List for SQuADDS:

Refer to contribution guidelines for more information on how to contribute code.

Bug Fixes:

• Addressing issues on GitHub
• Addressing TODOs in the code
• Fixing any bugs in the code
• Robustly handling caching and environment variables setup for all OS (some Windows users had issues)

Simulations:

• Stress-test and report any bugs for Ansys simulations using the SQuADDS package
• Add guesses for bare resonator frequencies (50 ohm impedance matched) using API calls to https://www.microwaves101.com/calculators/864-coplanar-waveguide-calculator
• Add more resonator property calculator methods via /~https://github.com/ooovector/cpw_coupling/blob/master/Conformal%20mapping%20of%20a%20CPW%20coupler.ipynb integration - may be a GUI? (https://smm.misis.ru/CPW-resonator-coupling/)
• Generalize SQDMetal Palace simulations to work with any HPC and make a PR
• Conduct hyperparameter optimization study of Palace simulations and test for reliability/repeatability
• Integrate SQDMetal (fixed version) as a dependency of SQuADDS
• Add an easy Palace simulation API on SQuADDS (with parallel processing)
• Write a comprehensive tutorial on how to use Palace for simulations
• Provide compute resources to run simulations covering sparse regions of the Hamiltonian space

Core:

• Speed up half-wave cavity operations and decrease memory usage (e.g., speed up both one-time cost methods such as generating the dataset and local operations done at the edge)
• Create a simple API for users to contribute experimental data to SQuADDS_DB
• Handle cases where the user does not wish to specify a resonator_type
• Improve system design for both the SQuADDS package and SQuADDS_DB
• Develop a system to "metalize" any .gds/.dxf file (i.e., generate the corresponding Qiskit Metal file from the CAD file)
• Add support to handle designs generated via other tools (explicitly not Qiskit Metal), such as:
  • KQCircuits
  • DXFWriter
  • gdsfactory
  • pyHFSS
  • PHIDL
• Provide APIs (modules/methods) to easily add more data columns to existing simulation entries in the database (e.g., allowing users to rerun geometries and add participation ratios)
• Transition datasets to SQLite or another format to handle larger-than-memory datasets as we scale
• Refactor code to implement faster methods with lower memory usage for handling DataFrame operations

Contribution:

• Use local LLMs/free secure LLMs to create/update the measured_device dataset using the GitHub repo
• Utilize HuggingFace Hub API for handling contributions to the database in a more streamlined and automated way (e.g., create clone, branch, PR, etc.)
• Implement and deploy an acceptance server for handling contributions to the database (calculates simulation and measured value discrepancies, automatically simulates representative data points for reliability, notifies maintainers for approval) [Not needed in the immediate future]
• Automate integration of CAD files along with their measured Hamiltonian parameters [Not needed in the immediate future]

Machine Learning (ML):

• Develop an architecture/framework for deploying ML models on SQuADDS (via HuggingFace endpoints/spaces, in code, etc.)
• Provide APIs (modules/methods) for incorporating ML interpolation features into SQuADDS
• Utilize HuggingFace Tasks for ML applications
• Identify relevant design space variables for any system given a set of $\hat{H}$ parameters using encoders
• Determine analytical dependence of $\hat{H}$ parameters on design space variables using Kolmogorov-Arnold Networks (KANs)
  • KAN Paper
  • Symbolic Regression Example
• Expand training datasets using cGANs/VAEs/PINNs (i.e., attempt to simulate the simulator)
• Build upon the work by Elie Genois et al., 2021
• system to add measured datasets to the database straight from arxiv/target journal papers

Workflows:

• Add any other workflow that assists developers in contributing
• Implement an automated build check with comprehensive unit tests
• Automated tests upon PR submission

Feature Requests:

• add airbridge generation tool
• add chi to the complete df
• Enable users to add methods with computation and append to merged_df for search in the Analyzer module
• Allow users to pass a circuit from SQCircuits and SQuADDS to provide a first-guess physical layout in Qiskit Metal
  • Helpful resources:
    • SQCircuit
    • Circuit Recognizer
    • CircuitNet
• Incorporate SCILLA and/or its applications

Boring but Necessary:

• add input type error handling to sim code
• automated kink detection + meander smoothing prior to qiskit-metal rendering
• letting users choose the .env file location OR telling them where to find it
• Check for claw dimensions on the Hamiltonian space plot
• Check for breaking changes in the latest version of dependencies and update the package accordingly
• Create unit tests for each feature/file
• Establish proper train/test splits and modify SQuADDS_DB() to always return all data
• Standardize the handling of units for simulated results and implement necessary backend changes
• Add more tutorials on how to use the package and its various applications
• Verify if the precision of design parameters is handled correctly and fix as needed
• Change all instances of NCap to CapNInterdigital
• Separate the documentation from the main package

Fancy/For Fun:

• Implement LLM-based queries for SQuADDS using pandas-ai (support for OpenAI and local LLaMA models)

---