wrundocs
Open the editor

Write it with AI

View as MarkdownOpen the editor

Any AI assistant can write wrun. It is TypeScript syntax, which assistants have read a great deal of, and the editor's compiler checks every draft before anything runs.

The loop

The AI loop: you describe your idea to any AI assistant together with Copy for LLM from the editor; you paste the file it writes into the editor and press Run; if the compiler names errors, Copy for LLM sends the code and every error back; when it compiles, it draws on your chart.compilesyour ideaany AIRunyour chartwith Copy for LLMin the editorerrors? Copy for LLM, paste it back
  1. Open a tab. In the editor, press New indicator, or open the indicator you want to change.
  2. Copy for LLM. Right-click the editor tab and pick Copy for LLM. One paste carries the language rules, the kit's API, your numbered code, every error and your last run.
  3. Ask. Paste it into any assistant and say what you want, naming everything it should draw (an example is below).
  4. Run. Put the complete file it returns into the tab and press Run. The compiler names every mistake with a line and column.
  5. Repeat until it compiles. Copy for LLM again and paste it back. When it compiles, it draws on your chart.

Better results

A request that names everything it wants drawn:

text
An RSI(14) in its own pane with bands at 30 and 70, a dot on every cross
back inside the bands, and an alert on each cross.
  • Name everything you want drawn. Assistants keep the lines and tend to drop labels, per-bar colors, fills, boxes, tables and alerts. List them in your request, then check the legend and the chart against it.
  • Build long scripts in parts. A few hundred lines can outgrow one answer. Ask for the settings and the plotted lines first and press Run; then ask for the drawings, then the alerts, each time with a fresh Copy for LLM.
  • Check the numbers before you publish. Compare the result with a version you trust, on the same market and timeframe.

Porting from Pine

Paste your Pine script with a Copy for LLM and ask for a port; the loop above takes it from there. wrun vs Pine shows one indicator in both languages, line for line.