The idea the rest is built on: every action has matching R you can keep.
Buffer a layer, run a join, draw a feature or import a dataset, and the R that actually ran appears in
the console. Turn on recording and those commands are written to a script
as you go: a new file, the open tab, or an existing .R file on disk. Click
through an analysis and walk away with a script that reproduces it.
Your decisions are in the code: the units a buffer used and how a join treated blanks, not a summary of them.
Run code your way: type in the console, Ctrl+Enter a line, or Shift+Enter a selection.
Documents too: R Markdown and Quarto with runnable chunks, knit to HTML, PDF or Word.
A buffer run from the interface, and the R it produced.
94 point-and-click tools in seven categories, and every one writes clean R.
Fill in a form, pick your layers and columns, and press Run. The toolbox covers the geoprocessing you reach
for every day, and shows the R it will run before it runs it.
Table Operations 11 tools
Filter, select, mutate, sort, distinct, sample and slice.
Joins & Combine 8 tools
Inner, left, right, full, semi and anti joins; bind rows or columns.
Aggregate & Reshape 5 tools
Group-by summaries, counts, pivots and frequency tables.
A GIS should refuse to give you a plausible wrong number. ouRGeo’s does.
Buffer a layer in degrees, join on keys that quietly duplicate rows, or average a column full of missing
values, and most tools hand back a result that looks fine. ouRGeo’s data checks
look at the input first. A clean operation simply runs; one that can’t be trusted is blocked, with
the affected rows named.
Joins that show their work: duplicate and blank keys, unmatched rows on each side, and the predicted row count, before anything runs.
Units you can trust: geographic layers buffer in real metres, and nothing guesses a CRS or a UTM zone.
Missing values on purpose: a summary asks whether to block, propagate or omit them, and an all-missing group is NA, never zero.
Exports that verify themselves: GeoPackage and GeoJSON are read back before they are published.
A left join, blocked before it runs: the right-hand table repeats two county names, and 40 counties have no match.
Build a chart from a form, and keep the ggplot2 code.
Chart Builder turns any table or layer into a ggplot2 chart. Drag columns onto
x, y, colour, size or facets, stack layers such as points and a smoothed trend, set the labels and theme,
and watch the preview update. The code is shown as you build, and lands in the console when you run it.
Ten chart types: scatter, line, bar, column, histogram, box plot, density, violin, area and smooth.
Publication maps: ggplot2 or tmap code from the map on screen, with legend, north arrow and scale bar.
Everything else R does: models, tests and every package on CRAN, in the same window as the data.
Chart Builder with a scatter and a smoothed trend. The generated ggplot2 code is on the right.
Digitizing with CAD-style snapping, and a traverse tool that takes the deed as written.
As you draw, ouRGeo snaps to vertices, edges, midpoints and true intersections on every
visible layer. Type an absolute coordinate, or place the next vertex by distance and bearing. Edits wait in
a Save Edits badge and are written by one batched R statement you can read first.
Distance & bearing
Press B and enter each course the way your notes are written:
150 ⇥ N45°30′E ⏎. Press U for a curve and give any two of
radius, arc length, chord, tangent, central angle and the rest; ouRGeo solves the others and prints them all.
Every notation: azimuths, DMS, quadrant bearings, deflections and angles right; metres, feet, US survey feet and chains.
It shows its working: the numbers on each leg, and the closure as 1:N once a figure could close.
Grid or ground: it always names which one is in force, and a projected layer defaults to its own grid.
Real curves: arcs are stored as CIRCULARSTRING, not chopped into segments.
Real parcels and roads have arcs. Most tools quietly turn them into short straight lines.
Ask sf/GEOS for the length or area of a CIRCULARSTRING and it first
linearizes the curve, which adds error, or refuses it. ouRGeo ships with
sfarc, an R package built for this project that works out the centre,
radius and swept angle of each arc and returns an exact answer.
Checked against GEOS, with linearization only as a warned last resort.
Correct on the globe: geographic curves are measured on the ellipsoid.
Nothing changes for straight lines: they go straight to sf.
The same arc, measured exactly and linearized.
Getting sfarc. It is a separate open-source package still in CRAN review, so for now it installs
from GitHub: remotes::install_github("CCurtis1988/sfarc"). Setup is on the
Download page.
07 The AI Assistant
An assistant that knows your session and can drive the whole app, typed or spoken. Frontier models or fully local.
Ask for the analysis you want in plain language, or say it out loud. The assistant can see your R
environment, write the code, and with Autopilot on, run it and iterate.
Everything it runs lands in your console like any other action. Use a frontier model from Anthropic, OpenAI or
Google, or an open model such as Qwen, Llama or Mistral on your own computer, so nothing leaves it.
OpenAIAnthropicGoogleChatGPT planClaude subscriptionOllamaLM Studiollama.cppvLLM+ more
It can drive the app: more than 170 actions, from maps, layouts and the editor to packages, themes and plugins, plus every toolbox tool. Each one is screen-only, Autopilot-gated or confirm-first. You choose.
Voice: the microphone beside Send dictates your request. Speech is transcribed on your computer by default; a cloud service is optional, and every upload is logged to the console.
Make it yours: describe a theme and it builds one, or describe a tool and it writes the plugin, then asks before switching it on.
Privacy you control: keep any column away from it, switch web access off entirely, and see what was sent under every reply.
A memory you own: kept on your computer, and learned memories are only suggested until you accept them.
The assistant working on a dataset with Autopilot on.
Public data catalogs and your own services, without leaving the app, as reproducible import code.
Browse a live service, build a query, preview the result, and import it. The import is R echoed to your
console, so the data pull can be rerun next month. Arcs from Esri services arrive as true curves.
Close the app mid-thought. Open it tomorrow exactly where you were.
A saved project keeps its whole working state on your computer between saves: documents with their unsaved
edits, the R environment, the map, plots, the console and your chats. R is snapshotted in the background, so
even a crash costs only the last few moments.
Project Files: keep the PDF standard, the client’s spreadsheet and last year’s report with the project, indexed for search and preview.
From any R function: read.csv(ourgeo_project_file("sites.csv")) and ggsave(ourgeo_project_path("figures/map.png"), p).
Shared drives, handled: if a colleague saves over the project, keep theirs, yours, or merge the two.
A script that reads from Project Files and saves into it. The figure joined the list on its own.
New connectors, tools and panels, written by you, by the assistant, or by someone else.
Most plugins are plain JSON with no code at all; R and compiled TypeScript are there when a plugin needs them.
Browse the plugin store on this site or from Settings inside the app. Every listing
was installed and run by a moderator before it appeared.
Installed plugins arrive switched off. Turning one on is your decision.
Publish your own: ouRGeo builds it and puts it in the review queue.