Bulk metadata
Monthly metadata exports as zstd-compressed NDJSON, one publication per line, with identifiers, timestamps, categories and every content hash. Publication bodies are not in these files: fetch those from the permanent URLs, which are in every line.
| publications-2026-10.ndjson.zst | 13 bytes |
|---|---|
| publications-2026-09.ndjson.zst | 18,282,671 bytes |
Line format
{"publication_id": "OT-2026-000000001", "uuid": "0199...", "published_at": "2026-09-15T08:43:17.483921Z", "title": "...", "categories": ["material-handling/conveying-systems"], "canonical_url": "https://opentechnical.org/p/OT-2026-000000001", "canonical_text_sha256": "...", "pdf_sha256": "...", "manifest_sha256": "..."}
Every line carries the permanent URL of the publication it describes, so the export is enough to fetch anything in the library without asking us for an index first.
Batch releases
Publications that arrived as a batch release are stored as archive segments, published whole with their release manifests at /data/releases/. Every text byte in them is recoverable with zstd alone.