This should be an easy fix.

After #3 is merged, it might be nice to investigate and enhancements to the meta-model.

It might be nice to add linters support via pre commit hooks. Also, it might be nice to add isort to python linting process.

It might be nice to change the favicon of the site.

It looks like Junction, Current and Voltage and Ground are missing from the ProjectBase seed, which means the plugin CreateElectricCircuitsMeta will not add them. This means we should add them to the base seed.

It might be nice to remove the old image while deploying this repo. Currently, they are just left untagged.

It would be great to be able to integrate a trained model to assist with model building. There are many different possibilities so I thought I would create this issue to track some of the ideas as well as any discussion about integration.

A couple ideas for analytics integration:

• suggest the next component
• detect errors (or possible anomalies which may be errors)
• search existing models using similarity/dissimilarity

It would be nice to add a checkpoint from GraphGym for recommendations. Currently, the one I am considering has reached about 36% on the hold out set and 50% on the training data.

It might be nice to create a device library using open source Spice device models available. This would mean:

1. Adding a Mixin family of DeviceModels to each nodes in the ComponentBase category.
2. Finding large sources of open source and device models and creating a Library of these Nodes using a Plugin. For example https://github.com/kicad-spice-library/KiCad-Spice-Library library has a huge collection of device models for KiCad.
3. After #44, make these components available for it as well.

Rather than creating a new plugin for each algorithm for recommending the next component, it would be nice to allow users to select the model to use. This would also make it easier to add new models as they won't require creating a whole new plugin.

Error feedback from python plugins is currently not very helpful and you have to view the server logs...

While Testing #25, I found that the NPN and PNP transistors have opposite pin order.

It might be nice to assign positions to these nodes based on category. Thanks to @brollb, I now have an idea on how to do it correctly.

It might be nice to sort out what the reference voltage name in the current controlled voltage sources mean and change the containment if so needed.

Some inspiration can be drawn from circuitlab to create a visualizer for creating and editing circuits. This will also be helpful for generating schematic from the netlist.

• https://www.circuitlab.com/editor/

At this current stage, ConvertCircuitToNetlist is unable to process the following Components.

• VariableResistor
• VariableConductor
• VariableCapacitor
• VariableInductor
• SaturatingInductor
• OpAmp
• OpAmpDetailed
• Gyrator
• OpAmpDetailed
• Potentiometer
• Transformer

It would be nice if the imported circuit (from a netlist) is named after the original filename.

It might be nice to integrate docker deployment of the codebase via github actions. This has already been done and proven to work well for deepforge.

It might be nice to incorporate units support into the components itself.

I noticed that it appears that CreateElectricCircuitsMeta is enabled by default on Circuit nodes which seems a bit counterintuitive - I would expect it to only be enabled on the "language container" (ie, ElectricCircuits in the root node).

It would also be nice if

• the recommendation plugin was enabled by default on circuits
• the export plugin was enabled by default on circuits
• the import plugin was enabled on circuit folders

The tests are are failing because of mongdb requirements.

A decorator should be added that decorates various components of the electric circuits. webgme/logic-gates should serve as a decent starting point.

Currently, there isn't any way to access the plugin configuration from within the python plugin. This is required for the analysis capabilities as I intend on having multiple models available that can be explicitly selected (so different approaches can be compared).

There are three parts to a netlist: 1. Decorations(labels, end, start). 2. Components and SubCircuits 3. Analysis (Transient, Noise, Pole-Zero, Small Signal etc... among others).

1. In any netlist, decorations are always the same, so a code generator can take care of it trivially.
2. According to what I gleaned from the meeting on Monday, the focus should not be simulation. Although, I would argue that for a meaningful metric that can be used for evaluating ML/RL methods we comeup with, simulation will be the part of the pipeline. But, for now we could ignore this.
3. The components aspect is a little bit tricky. Lets look at op-amp for example:
   • There is no direct way to specify an Op-AMP, with different regions of operations. Whether its an ideal op-amp with infinite gain etc.... The first thought here (at least the most st. forward) is to modify the metamodel to add an attribute (Spice-Model/Configurations) for each component that is a dictionary of some sort and could be specified on component drag to the circuit from the PartBrowser to the Model Editor. An alternative is to create different nodes for different configurations , which might be easier.
4. SubCircuits. Handling subcircuits should be trivial because the way we can specify them in the current metamodel.

So, with the current state of the project and decorated metamodel, we need the following steps to be able to generate a SPICE netlist.

1. Parse the PySpice package to get the Spice Component Attributes hence creating any extra metanodes not already in the current metamodel. Once parsed, a domain expert (Peter and Akos) can verify that. (I will work on this first)
2. Add a plugin for spice netlist, using a Python plugin using PySpice. This should be trivial with excluding analysis and model parts.

Also, it might be nice to create some example projects.

The github action to lint code base is not working correctly.

It might nice to integrate CI and Linting using github actions. This should be done once the decorator ( #3 ) is ready.

• https://github.com/github/super-linter
• https://docs.github.com/en/free-pro-team@latest/actions/guides/building-and-testing-nodejs

After #39 is Merged, we should add decorator icons to the sources.

It might be nice to fix these/ remove unnecessary components.

This Library should be integrated with DeepForge. Primarily, after #44. This should be easy considering, after #23, we should be able to create a PySpice Circuit Object/ Some other Pythonic representation of the Circuit.

Uncaught TypeError: Cannot set property '/k/3' of undefined
    at ElectricCircuits.FourTerm.js:65
    at Array.map (<anonymous>)
    at ElectricCircuitsDecorator.FourTerminalComponent.planPins (ElectricCircuits.FourTerm.js:50)
    at ElectricCircuitsDecorator.FourTerminalComponent._updatePorts (ElectricCircuits.FourTerm.js:27)
    at ElectricCircuitsDecoratorCore._update (ElectricCircuitDecorator.Core.js:240)
    at ElectricCircuitsDecoratorCore.update (ElectricCircuitDecorator.Core.js:218)
    at ElectricCircuitsDecoratorCore._renderContent (ElectricCircuitDecorator.Core.js:214)
    at ElectricCircuitsDecorator.on_addTo (ElectricCircuitsDecorator.DiagramDesignerWidget.js:48)
    at DesignerItem._callDecoratorMethod (DesignerItem.js:326)
    at DesignerItem.addToDocFragment (DesignerItem.js:177)

This is causing problems when dragging FourTerminalComponents from PartBrowser to ModelEditor.

As a concrete example of an analytic (referenced generally in #59), it would be nice to add a component recommender/completer to make sure everything is in place so we can experiment with models trained from https://github.com/symbench/spice-completion/.

For the purposes of this issue, the "mock" recommender could simply return the top 3 most commonly used components in the netlist. Concretely, it might be nice to have the mock recommender use the netlist representation (not the GME nodes) and return the top three components with confidences (probably just some hard-coded values for the mock analytic).

It might be nice to parse the PySpice for additional attributes for the components which might be needed to generate netlist for the circuit. See #23 for a related discusion.

While working on #23. I found that the current version of the meta-model is missing all the necessary components to model a circuit in SPICE. There should be some additions to the meta-model such that it incorporates the necessary information presented in the document below:

http://bwrcs.eecs.berkeley.edu/Classes/IcBook/SPICE/UserGuide/elements_fr.html

Here's a non-exhaustive list of decorator enhancements

1. @brollb suggested not to hard-
   code meta-node Concepts with the decorator. So, an alternative should be followed.
2. Improve pin planning for Circuits
3. It might also be useful to display pin names.

Uncaught Error: importScripts failed for AutoRouterActionApplier at /lib/autorouter/action-applier.min.js
http://requirejs.org/docs/errors.html#importscripts
at makeError (require.js:168)
at Function.req.load (require.js:1984)
at Object.load (require.js:1685)
at Module.load (require.js:834)
at Module.fetch (require.js:824)
at Module.check (require.js:856)
at Module.enable (require.js:1176)
at Object.enable (require.js:1557)
at Module. (require.js:1161)

This is happening in webgme.symbench.org

#44 has grown quite large to act as a standalone PR. It might be nice to refactor it into several smaller Pieces such that it all makes sense.

1. Integrate JointJS, Dagre, loadash, GraphLib and a custom circuits domain for JointJS.
2. A base visualizer which serves as a base for #44.
3. Non-Interactive CircuitEditorVisualizer that subclasses visualizer in 2.
4. More Ideas around integration.

The autorouter is throwing the following error when using the decorated components.

assert.AssertionError
actual: false
expected: true
generatedMessage: false
message: "AREdgeList.getEdgeByPointer: edge !== null FAILED!"
name: "AssertionError"
operator: "=="
stack: "AssertionError: AREdgeList.getEdgeByPointer: edge !== null FAILED!↵    at AutoRouterEdgeList.getEdgeByPointer (https://webgme.symbench.org/lib/autorouter/action-applier.min.js:2:12209)↵    at AutoRouterGraph._deleteSamePointsAt (https://webgme.symbench.org/lib/autorouter/action-applier.min.js:4:2888)↵    at AutoRouterGraph._deleteTwoEdgesAt (https://webgme.symbench.org/lib/autorouter/action-applier.min.js:4:1240)↵    at AutoRouterGraph._simplifyPaths (https://webgme.symbench.org/lib/autorouter/action-applier.min.js:4:4138)↵    at AutoRouterGraph._optimize (https://webgme.symbench.org/lib/autorouter/action-applier.min.js:4:18148)↵    at optimizeFn (https://webgme.symbench.org/lib/autorouter/action-applier.min.js:4:16896)"
__proto__: Error

The workflow file needs an edit

Currently its "[0]" for them. But it should be a list of tuples signifying the voltage vs time value.

https://nodejs.org/en/about/releases/

It might be nice to test for node version 12 and 14 instead of 12 and 10

It might be nice to add a basic plugin that takes care of converting Netlist to Circuits. This will open the pathway to create test circuits mined by @brollb in the netlists repository.

It might be nice to remove the unnecessary state added by the self.pyspice_circuit data attribute. It appears this could simply be passed around as a variable making the functions editing it explicitly acting on it rather than as a side-effect.

It might be nice to integrate DeviceModel Pointer for each component that can represent a standard SPICE Model associated with it. This will be more helpful for running simulations with SPICE simulator of choice while generating the netlist.

It might be nice to consolidate SubCircuit Nodes to a Set because it will be a problem when you consider addition of that to a set.