nx_revimg_crawl.nx

nx_revimg_crawl.nx -- the DISCOVERY DRIVER: turns real image files (a crawl's fetched media) into the fingerprint corpus that nx_revimg_discover reasons over. This is the glue that makes the reverse-image discovery loop run on REAL media: each candidate image is decoded through the shared nx_img_to_gray (PNG today, JPEG since the FF-D8-FF wiring -- so real web media works), perceptually hashed (nx_phash dHash), and appended to the (site, fingerprint, path-id) corpus. Undecodable/unsupported files are SKIPPED (graceful degradation, never fatal -- one bad image can't abort a crawl). The network I/O (fetch HTML, extract <img> via nx_html_extract_imgs, fetch each image to a local file) is the operator-driven half wired to nx_https_get; this driver is the pure decode->fingerprint->corpus half, so it is fully gateable against local fixtures and identical whether the bytes came from disk or the live web. Composes nx_img_to_gray + nx_phash + nx_revimg_discover -- ZERO new logic. ORIGINAL