notemd-selection

Overview

This skill enables processing of selected text portions within files, rather than forcing full-file rewrites. This saves tokens, reduces processing time, and allows targeted AI operations on specific sections.

When to Use

• Partial content generation: Generate content for just a section
• Selective linking: Add links to specific paragraphs
• Targeted translation: Translate selected portions
• Efficient processing: Save tokens by processing only what's needed

Function Call Chain

Process Selection
├── Get selected text from editor
├── Get surrounding context (optional)
├── Process selected content (LLM)
├── Replace or append selected content
└── Update file

Selection Processing Patterns

1. Get Selected Text

// From Obsidian editor
const editor = view.editor;
const selection = editor.getSelection();

// From VS Code
const selection = editor.selection;
const text = editor.document.getText(selection);

2. Process with Context

For better results, include surrounding context:

// Get context before and after selection
const lineNum = editor.getCursor().line;
const beforeLines = editor.getRange({line: 0, ch: 0}, {line: lineNum, ch: 0});
const afterLines = editor.getRange({line: lineNum, ch: 0}, {line: maxLine, ch: 0});
const fullContext = beforeLines + selection + afterLines;

3. Replace Selection

// Replace selected text with processed result
editor.replaceSelection(processedContent);

// Or append after selection
editor.replaceRange('\n\n' + processedContent, {line: lineNum + 1, ch: 0});

Use Cases

Wiki-Link from Selection

// Selected: "Machine learning is a subset of AI"
// Process: Add links to key concepts
// Result: "Machine learning is a subset of [[AI]]"
async function addLinksToSelection(app: App, editor: Editor) {
    const selection = editor.getSelection();
    const prompt = getSystemPrompt('addLinks');
    const result = await callLLM(prompt, selection);
    editor.replaceSelection(result);
}

Generate from Selection

// Selected: "Neural Networks"
// Process: Generate expanded content
// Result: Full explanation of neural networks
async function generateFromSelection(app: App, editor: Editor) {
    const selection = editor.getSelection();
    const prompt = getSystemPrompt('generateTitle', {TITLE: selection});
    const result = await callLLM(prompt, '');
    editor.replaceSelection(result);
}

Translate Selection

// Selected: Some text in English
// Process: Translate to target language
// Result: Translated text
async function translateSelection(app: App, editor: Editor, targetLang: string) {
    const selection = editor.getSelection();
    const prompt = getSystemPrompt('translate', {LANGUAGE: targetLang, TEXT: selection});
    const result = await callLLM(prompt, '');
    editor.replaceSelection(result);
}

Settings

Selection Processing

| Setting | Description | |---------|-------------| | includeContextLines | Number of context lines to include | | replaceSelection | Replace or append processed content | | preserveFormatting | Preserve original formatting |

Context Options

| Setting | Description | |---------|-------------| | contextBefore | Lines before selection | | contextAfter | Lines after selection | | includeHeadings | Include parent headings |

Error Handling

Common Errors

1. Empty selection

   • Error: No text selected
   • Solution: Check selection length before processing

2. Selection too large

   • Error: Exceeds token limits
   • Solution: Split into chunks or use full file mode

3. Invalid cursor position

   • Error: Cannot determine insertion point
   • Solution: Check editor state before operations

Platform-Agnostic Approach

Web/Node.js

// Get selection from DOM
const selection = window.getSelection().toString();

// Process with file system
const content = fs.readFileSync(filepath, 'utf-8');
// Find and replace range

CLI (Standard Input/Output)

# Pipe selected content to script
echo "$SELECTED_TEXT" | notemd-process --operation=links

# Output replaces stdin

Best Practices

1. Always validate selection: Check if selection is non-empty
2. Provide context: Include surrounding text for better results
3. Preserve formatting: Maintain markdown structure
4. Preview before replace: Show preview for user confirmation
5. Undo support: Ensure easy rollback if needed