mrslimslim/awesome-prompt

This is a small Node.js project that packages up an archive of MeiGen prompt examples, both still images and video thumbnails, together with their text prompts. The author appears to be working from a local scrape of MeiGen content sitting in a Windows folder, and the script reads from that source, compresses the media, and writes the results back into the repo so they can be served straight from GitHub. To use it, you run npm install to pull the dependencies, then npm run build:meigen to do the actual work. The default source directory is hard-wired to a path on the author's E drive, but you can pass extra flags. The README mentions --quality=76 to set the image quality target, --max=1280 to cap the image edge length, and --concurrency=8 to control how many files are processed in parallel. A --raw-base flag lets you override the base URL the script writes into the metadata, and --clean=false skips wiping the output before a rebuild, which is handy when you only want to refresh the JSON URLs after adding a GitHub remote. The output is organized into a few folders and files. There is an images directory for compressed prompt images and a video-thumbnails directory for video thumbnails. The data folder holds JSON metadata in three forms: prompts-all.json for everything with CDN image URLs, prompts-images.json for only image prompts, and prompts-videos.json for only video prompts. There is also a CSV export at prompts-all.csv and a summary.json with build and compression stats. For previewing the result locally, the README suggests npm run serve, which starts a small static server on port 4173. Once the repo has a GitHub remote attached, the script can also pick up the right raw.githubusercontent.com base URL automatically, so the metadata points at the GitHub raw URL pattern shown in the README. The README is short and is essentially a build and preview reference, not a user-facing description of MeiGen itself.