| name | marp-lens |
| description | Semantic search for Marp presentations using vector embeddings. Use when finding relevant slides by topic, retrieving slide content, or exploring presentation materials. Triggers on "find slides about...", "search presentations for...", "get slide content", "what slides cover...", or any Marp/presentation search query. |
marp-lens
Vector-based semantic search CLI for Marp presentations.
Quick Reference
npx marp-lens search "機械学習の基礎" --limit 5
npx marp-lens get "presentation.md #3"
npx marp-lens stats
npx marp-lens index -d ./slides
npx marp-lens index -d ./slides --with-images
Commands
Search
npx marp-lens search "<query>" [options]
| Option | Description |
|---|
-l, --limit <n> | Max results (default: 10) |
-t, --threshold <0-1> | Min similarity (default: 0) |
-o, --format <type> | Output: table or json |
Get Slide
npx marp-lens get "<file> #<number>"
Retrieves full content of a specific slide. Supports partial path matching.
Index
npx marp-lens index [options]
| Option | Description |
|---|
-d, --dir <path> | Directory to index |
-f, --file <path> | Single file to index |
-r, --rebuild | Clear and rebuild index |
-i, --with-images | Include image descriptions |
Stats
npx marp-lens stats
Shows file count, slide count, embedding count, and database size.
Workflow
- Find relevant slides:
npx marp-lens search "topic"
- Get full content:
npx marp-lens get "file.md #N" for slides of interest
- Re-index if needed:
npx marp-lens index -d ./slides after adding files
Output Format
Search results show similarity score, file path, slide number, heading, and content preview. Use --format json for structured output.
