[riskpy_standalone]
Choose an illustrative example, Load example, review uncertain inputs and fixed assumptions, then Run. Start with 1,000 iterations. Inspect the output selector, histogram, cumulative probability curve and sensitivity ranks. Download draws CSV or save the complete JSON project.
Uncertain inputs have unique Python identifier names and a distribution with labelled parameters. Fixed assumptions are named numeric constants. The Python model must define model(inputs, constants) and return a dictionary of output names to arrays, one value per trial, or a numeric scalar expanded to all trials. Input samples are NumPy arrays. Example: return {"Profit": inputs["volume"] * constants["margin"]}.
Supports 13 distribution families, Monte Carlo or independent Latin hypercube sampling, reproducible seeds, and Gaussian-copula correlation in Monte Carlo mode. Correlation is a latent-normal matrix, not a guarantee that final nonnormal variables have identical Pearson correlation. Maximum 200,000 iterations with memory-based input/output limits. Correlated LHS is unsupported.
Statistics use all simulated trials. Draws CSV exports at most the first 20,000 trials, and the interface states this limit. Spearman sensitivity is association, not causation or variance decomposition. The examples are illustrative; investment examples omit tax, depreciation and working capital. Excel workbook simulation and worksheet functions are intentionally unavailable in this standalone edition.
Download the individual plugin ZIP. In WordPress choose Plugins > Add New > Upload Plugin, select the ZIP, Install Now, then Activate. Create a page, insert a Shortcode block, and paste the relevant shortcode. Publish and open the page. You do not need Excel, Python, a desktop launcher, PowerShell or an Office manifest. These are separate plugin directories and can be active together.
Plugin PHP code targets PHP 7.4 or newer and WordPress 5.5 or newer. Prefer a currently maintained PHP/WordPress release where possible. Use HTTPS and a current desktop browser with WebAssembly and workers. First use downloads Pyodide and packages from jsDelivr; large packages can take a minute or longer on a slow connection. Calculations then run locally. Offline first use and mobile performance are not guaranteed.
CSV uploads and numerical calculations remain in the browser. The tool does not upload source rows to WordPress. Loading the runtime contacts the jsDelivr CDN, which receives normal network metadata. Optional Spyglass AI sends question text, column names/types and prior plans to an external service. User-written Python can access network APIs: review code before running it. The shortcode embeds the tool in an isolated sandboxed frame; do not remove its sandbox or add allow-same-origin.
Use Download CSV for completed data. CSV files can be opened in Excel, LibreOffice and other spreadsheet tools. This release exports CSV rather than native XLSX workbooks. Models and scripts have separate JSON/Python download controls where provided. Nothing is automatically saved to a user account or server database. Save your work before closing or reloading the page.
Shortcodes accept height, for example height="1400". Exclude the plugin app assets from aggressive JavaScript combining/defer rewrites. A Content Security Policy must allow the local app scripts, blob: workers, WebAssembly execution and https://cdn.jsdelivr.net runtime/package access. Server Cross-Origin-Resource-Policy: same-origin can prevent app scripts loading in an opaque sandbox; use a compatible policy for these public assets. Avoid placing all four heavy tools on one page.
Actual Pyodide 0.29.3 WebAssembly tests in Node passed for script execution/PNG plotting, Spyglass regression, RiskPy simulation/export and ExcelBlocks first-order kinetics. RiskPy native engine: 15 tests passed; all ten ExcelBlocks examples returned finite results; PHP syntax parsed and worker/CSV contracts passed. A full WordPress site, live paid AI service and browser visual/sandbox behaviour were not verified in this environment. Test each plugin on a staging page before making it public.
If Python stays loading, check CDN access, browser console, content-security restrictions and cached old assets. Try one tool per page in a fresh tab. Stop/reset terminates the calculation worker. Large simulations may exceed browser memory; reduce trials or shorten the model. Only Pyodide-compatible packages work: no Windows DLLs, COM, xlwings or direct desktop file paths. This release does not provide paywalls, licences, membership permissions or commercial support automation.