This project demonstrates how to call Python scripts from a .NET 9 Razor Pages web application, using the recently published CSnakes library to enable Python execution within .NET.

• The application is built on .NET 9 Razor Pages and integrates with CSnakes, allowing seamless Python script execution within .NET.
• A minimal API endpoint is used, together with HTMX, to call Python scripts and return dynamically generated stock graph images (PNG format) using the matplotlib and pandas libraries in Python.

• .NET 9 SDK (the solution targets net9.0)
• Python 3.12 — optional. CSnakes is configured to fall back to a NuGet-supplied CPython 3.12.4, so a local install is only needed if you want to run the Python tests directly, or if you set Python3_ROOT_DIR.

cd PythonNet.App.Web
dotnet run

Then open the URL printed in the console (see Properties/launchSettings.json).

On first start CSnakes creates PythonNet.Python/.venv and pip-installs PythonNet.Python/requirements.txt into it. This takes a minute or two; progress is visible in the console output. Subsequent starts reuse the venv.

Run from PythonNet.App.Web, not the solution root — Program.cs defaults the Python home to <cwd>/../PythonNet.Python, and util.read_csv resolves data files as Data/<file> relative to the working directory.

To point at a Python home elsewhere (a published layout, a shared venv), override either setting via appsettings.json or the environment:

Python__Home=/srv/pythonnet Python__VirtualEnvironment=/srv/venv dotnet run

dotnet test PythonNetSolution.sln            # xUnit: endpoint validation + routing
cd PythonNet.Python && pip install -r requirements.txt pytest && python -m pytest

Both are run by .github/workflows/build.yml on every push and pull request.

1. PythonNet.Python holds the Python sources. Each one is listed as an <AdditionalFiles> entry in PythonNet.Python.csproj, which is what makes the CSnakes source generator pick it up.
2. At build time the generator reads the type-annotated function signatures in util.py and plot.py and emits matching C# interfaces — IUtil and IPlot — plus IPythonEnvironment extension methods (.Util(), .Plot()) that return them.
3. Program.cs configures the runtime (WithHome, WithVirtualEnvironment, FromNuGet("3.12.4"), WithPipInstaller) and registers IUtil/IPlot as singletons, so they can be injected anywhere.
4. MinimalApiExtension.cs injects them into GET /api/dashboard/stock, which validates its query parameters, then chains read_csv → date_filter → sort_data → plot → save_plot and returns an <img> fragment for HTMX to swap into the page. Each request renders to its own wwwroot/plots/plot_{ticker}_{guid}.png — served from /dynamic-plots with caching disabled — so concurrent callers cannot overwrite each other's chart. Files older than five minutes are swept on the next request. The naming/ error-handling/sweep logic is shared (via RenderPlot) with the endpoint below.
5. GET /api/dashboard/test-plot is the same pipeline, minus the CSV: it calls util.generate_test_data(ticker, days, start_date) — a deterministic sine-wave price series, no file or network I/O — then sort_data → plot → save_plot as above. Useful for exercising the chart/theme/date-range UI without needing a real dataset. Try it via the "Test Data" page, or directly: GET /api/dashboard/test-plot?tickerDd=1&days=30&openCb=on&closeCb=on&theme=light.
6. POST /api/dashboard/custom-plot takes a JSON body (ticker, dates, open, close, plus optional title/showOpen/showClose/theme) and calls util.build_price_data(dates, open, close) to turn those arrays straight into the same sort_data → plot → save_plot pipeline — for plotting caller-supplied data without a CSV or the synthetic generator, e.g.:

   curl -X POST http://localhost:5000/api/dashboard/custom-plot \
     -H "Content-Type: application/json" \
     -d '{"ticker":"ACME","dates":["2024-10-24","2024-10-25"],"open":[10,12],"close":[10.5,12.5]}'

Adding a Python function is therefore: annotate it, rebuild, call the generated C# method. Adding a new .py file also needs an <AdditionalFiles> entry.

PythonNet.App.Web/Data/ ships three static CSVs (Date,Close,Open,High,Low) covering 24–27 Oct 2024 for AAPL, MSFT and IBM, so the dashboard works offline with no market-data API key. The dropdown values map to them in MinimalApiExtension.Tickers; add a ticker by dropping a CSV in Data/, adding a <None Update> copy entry in PythonNet.App.Web.csproj, and adding one line to that dictionary and to IndexModel.Tickers.