How the index is built.

PastTrace ingests source records through versioned adapters, validates rights deterministically, and publishes only records that pass provenance and quality checks.

Ingestion and search

The current host runs application code but has no production database binding. This release uses a reproducible build-time JSON catalog and serialized MiniSearch index. Official Wikimedia Commons, NASA, Internet Archive, and Sound & Vision APIs are cached under the pipeline; uncertain records are quarantined. Search supports fuzzy, prefix, and curated multilingual query expansion in the browser. It is not PostgreSQL.

Shot segmentation

Eligible source video is processed temporarily. The V0.1 experiment measures scene scores with FFmpeg, emits timestamped boundaries, and creates a distinct preview frame. AI-assisted descriptions remain in a separate field and are labelled in the interface.

Location

Country filters are derived only from discoverable source metadata. Unknown locations remain unknown rather than being guessed.