Skip to content

Data Point Viewer

The viewer (benchmark/viewer/index.html) shows each data point interactively (the input, the oracle answers, metadata and the parameters of every component). Ideal for exploring the dataset and understanding a plant. It runs offline in the browser, with no server.

Data point viewer with plant graph and sidebar

What the viewer shows

  • Plant graph: components as nodes, connections as edges.
  • Component details: a table of the simulated parameters, acceptance band and notes.
  • Sidebar: the input, the expected follow-up questions, the oracle
    answers
    , the "must not invent" prohibitions, the dropped (not simulated) parts and the metadata.

Legend (graph)

Element Meaning
Node color component type (Digester, GasStorage, CHP, Flare, …)
Edge color / style connection type: liquid · gas · heat
Node border the component's obligation (given / derivable / may-be-asked / automatic)

The legend in the lower-left of the viewer summarises the color coding.

Launching

You can run the viewer offline, but new changes to the dataset will not be shown. Alternatively, refresh the index file or start a local server to see changes directly in the viewer.

Offline

Double-clicking benchmark/viewer/index.html opens the viewer directly (file://). It then shows an embedded copy of the data points. Use Load files… to open arbitrary data point JSONs manually.

Live mode

Always shows the current dataset; changes appear after a reload.

python -m http.server 8000              # simple local server

Then open in the browser: http://localhost:8000/benchmark/viewer/

Controls

Action Control
Previous / next data point ◀ / ▶ buttons or arrow keys ←/→
Show component details click a node in the graph
Zoom + / − / ⤢ buttons or mouse wheel
Open your own data points Load files… (one or more .json)

After changing the dataset

If you added or changed data points, regenerate the index:

python benchmark/eval/make_index.py

More about the dataset and the evaluation scripts is in Using the Dataset.