improving ignored content from waczs

2026-06-13 05:38:29 +03:00 · 2023-07-28 12:19:14 +01:00
parent 7a5c9c65bd
commit aa71c85a98
3 changed files with 32 additions and 2 deletions
--- a/src/auto_archiver/enrichers/wacz_enricher.py
+++ b/src/auto_archiver/enrichers/wacz_enricher.py
@@ -180,6 +180,9 @@ class WaczArchiverEnricher(Enricher, Archiver):
                        m.set("src_alternative", record_url)
                    except Exception as e: logger.warning(f"Unable to download best quality URL for {record_url=} got error {e}, using original in WARC.")

+                # remove bad videos
+                if m.is_video() and not m.is_valid_video(): continue
+                
                to_enrich.add_media(m, warc_fn)
                counter += 1
                seen_urls.add(record_url)