A lightweight Streamlit interface for setting up, running, and exploring GenX.jl capacity-expansion models — so you can stop hand-editing the CSVs that GenX reads and produces.
Started as a single-zone learning tool. It now handles the standard GenX example systems, including multi-zone / DC-OPF networks, and reads GenX output directly (no Julia re-run to look at results).
| Page | What it does |
|---|---|
Runner (app.py) |
Pick a case, see a Run Preview of what the solve will contain (timesteps, zones, resources, policies, solver), and a System Resources table showing which capacities are optimized vs. fixed. Launch Julia and stream the terminal output live; a failed run gets a plain-language diagnosis instead of a wall of stack trace. Keep a copy of any run with Archive this run. |
| Cases | Create a case from a bundled GenX.jl example (the new case defaults to the example's name), or rename / duplicate / delete your own. Each card shows size and whether results are stale relative to inputs. |
| Inputs | Browse resources/ · system/ · policies/ · settings/ in a file tree. Edit CSV tables and .yml settings inline with save, each with a Column / Settings reference pulled from the GenX docs. An All Resources Graph view draws the fleet as a zone → type → resource treemap and a bus/tie-line topology diagram. |
| Results | Reads the GenX output CSVs and renders: zone-aware Key Metrics and per-asset Energy / LCOE & cost tables, Capacity Built, Supply to Load Mix (per-zone donuts that reconcile to each zone's demand, with an Imports slice), Interzonal Transfers (net import/export per zone, line flows vs. rating), Unserved Energy timing, Cost Breakdown, hourly curtailment / power, and an inferred Storage Charging Source. Exports a standalone HTML report. |
| Archives | Browse saved runs with headline metrics (system cost, NSE, capacity) and the GenX commit they ran against; open any archive in Results. |
| Help | Searchable GenX.jl reference (settings, inputs, outputs, time-domain reduction, solvers). Same content that backs the inline tooltips on the Inputs page. |
Runner — run preview + optimize-vs-fixed resource summary
Cases — create from an example, rename / duplicate / delete
Inputs — inline CSV editing
Inputs — All Resources Graph (fleet composition + zone topology)
Results — capacity, per-zone supply mix
Install Julia ≥ 1.9 from julialang.org and make sure it is on
your PATH (the Runner shells out to julia):
julia --versionInstall GenX into your default Julia environment:
julia -e 'using Pkg; Pkg.add("GenX")'
julia -e 'using Pkg; Pkg.status()' # should list GenXThe Runner executes each case with julia --project=. Run.jl from inside the case folder, so
using GenX has to resolve from that environment. (Alternatively, give a case its own
Project.toml that depends on GenX.)
Clone GenX next to this repo:
parent/
├── GenX.jl/ # git clone https://github.com/GenXProject/GenX.jl.git
└── GenXUI/ # this repo
GenXUI uses the sibling ../GenX.jl/ for two things: the bundled example systems the Cases
page imports from, and a live copy of the GenX docs for the Help page. Neither is required to
run your own cases — a docs snapshot ships in reference/genx/ — but the example importer is
empty without it.
Python ≥ 3.10. Install dependencies:
pip install -r requirements.txtFrom the GenXUI/ directory:
python -m streamlit run app.pyThen open http://localhost:8501.
On first launch GenXUI asks for a workspace folder. It creates two directories inside it:
<workspace>/
├── data/ active cases (inputs + their latest results/)
└── archive/ saved run snapshots
The choice is remembered in ~/.genxui/config.json and can be changed any time from the
sidebar. GenXUI never scans arbitrary folders for cases — only data/.
-
From an example — Cases page → New case from a GenX.jl example → pick one → Create. It's copied into
data/under the name you choose (defaults to the example name). -
Your own — drop a case folder into
<workspace>/data/. A folder counts as a case if it contains aRun.jl. Standard GenX layout:MyCase/ ├── Run.jl ├── resources/ Thermal.csv, Vre.csv, Storage.csv, Vre_stor.csv, … ├── system/ Demand_data.csv, Generators_variability.csv, Fuels_data.csv, Network.csv ├── policies/ CO2_cap.csv, … └── settings/ genx_settings.yml
See the GenX input-file docs for the full spec.
GenXUI-launched runs set OverwriteResults: 1 so each run replaces the case's results/ in
place (no results_1/, results_2/, … fan-out). To keep a run, use Archive this run on the
Runner or Results page — it snapshots inputs + results into archive/ with headline metrics
and the GenX git commit.
- One case at a time — no side-by-side case comparison.
- Multi-zone / transmission runs are supported for viewing and analysis; the network-flow attribution in Supply to Load Mix is a documented approximation (local generation assumed to serve local load first).
- No multi-stage (
MultiStage) investment support. - Julia startup latency (~30–90 s) before the first solver output appears.
pip install pytest ruff
python -m pytest tests/ # pure-logic tests (no Julia, no Streamlit)
python -m ruff check src/ pages/Repo layout: app.py + pages/ are the Streamlit UI; src/ holds the pure logic
(workspace, metrics, run_preview, run_diagnosis, help_docs, fleet_view, …);
reference/genx/ is a bundled GenX docs snapshot for the Help page.
Developed by Alex Panchula with Claude Code (Anthropic).
GenX is developed and maintained by the GenX Project team.