Release notes *3.0.8*

x86/64

docker run -it \
  -v ~/wljs:"/home/wljs/WLJS Notebooks" \
  -v ~/wljs/Licensing:/home/wljs/.WolframEngine/Licensing \
  -v ~/wljs/tmp:/tmp \
  -p 8000:3000 \
  --name wljs \
  ghcr.io/wljsteam/wljs-notebook:main

brew install --cask wljs-notebook

Workers API

WLJS now provides asynchronous access to parallel kernels. The freeware Wolfram Engine is not limited to a single parallel kernel or subkernel, so you can use this API to build truly parallel Wolfram Language workflows.

Why not ParallelSubmit?

The default Wolfram parallel API is blocking, which makes it hard to stream data in and out without pausing the main kernel. The Workers API avoids that bottleneck. Here is a basic message loopback example:

w = WorkerLaunch[Module[{},
  EventHandler[Null, {
    _ -> (EventFire[Null, #]&)
  }]
]]

WorkerObject[(*VB[*) LinkObject["'/Applications/Wolfram Engine.app/Contents/Resources/Wolfram Player.app/Contents/MacOS/WolframKernel' -noicon -wstp -noinit -subkernel", 819, 11], "dfaaf4aa-19b6-484b-8ce0-ef7d3dc13233" (*,*)(*"1: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"*)(*]VB*)]

You can also start a worker from a .wl file:

w = WorkerLaunch["path_to_worker.wl"];

Each worker object runs in a separate OS thread and communicates with the main kernel through the shared-memory WSTP protocol.

This design mirrors the JavaScript Workers API used across frontend and backend systems. In the example below, we subscribe to messages from a worker, then submit one to it:

EventHandler[w, Print];
EventFire[w, "Hey!"];

Terminate the worker when it is no longer needed:

WorkerClose[w]

Check the documentation for more examples.

Typed `Shallow` / Expression Summarization

With help from @Hachann, we extended Shallow with typed expression summarization. You can tune both the depth and the number of arguments shown at each level. It now hides the internals of opaque objects such as ByteArray and TimeSeries, whose input forms are usually less useful than their overall structure.

We also expanded the basic skeleton view from the Wolfram Standard Library. Instead of showing only <<N>>, WLJS can now display types and expression patterns such as:

Table[{i,j}, {i,1000}, {j,10}]//Shallow

{{{1, 1}, {1, 2}, {1, 3}, {1, 4}, {1, 5}, <<5, {_Integer, _Integer}>>}, {{2, 1}, {2, 2}, {2, 3}, {2, 4}, {2, 5}, <<5, {_Integer, _Integer}>>}, {{3, 1}, {3, 2}, {3, 3}, {3, 4}, {3, 5}, <<5, {_Integer, _Integer}>>}, {{4, 1}, {4, 2}, {4, 3}, {4, 4}, {4, 5}, <<5, {_Integer, _Integer}>>}, {{5, 1}, {5, 2}, {5, 3}, {5, 4}, {5, 5}, <<5, {_Integer, _Integer}>>}, <<995, {{_Integer, _Integer}..}>>}

Byte arrays are summarized by structure rather than full contents:

{{{{ByteArray[{1,2,3,4}]}, Red, I}}, ByteArray[RandomInteger[{0,255}, 10000]]} // Shallow

{{{{ByteArray[...]}, RGBColor[1, 0, 0], I}}, ByteArray[...]}

A more complex example:

Shallow[Plot[x, {x,0,1}], {2,1}]

Graphics[{Annotation[<<3>>]}, <<1, {_Rule | _RuleDelayed..}>>]

UI Improvements

The notebook now autoscrolls the focused cell into view.
Fixed race conditions during multiple Rasterize calls.

MCP Server

Improved summarization, with contributions from @Hachann.
Added more options for cell evaluation.
Improved error messages.
Improved the CLI.

Wolfram Engine 15.0

For now, WLJS Notebook will stay on Wolfram Engine 14.3. Early testing with Wolfram Engine 15.0 showed a 15-20% performance regression, including on the official WolframMark benchmark on MacOS/ARM64 machines. WLJS still runs on 15.0 in shallow tests, but we cannot guarantee that every new feature works reliably there yet.

Misc