Operational corpus inventory · 2022-02 — 2026-09

Coverage
through 2026-09

Stored row counts only; no discourse-change findings.

Snapshot of 446 Parquet files across 56 months. Counts describe stored rows, including rows whose text completeness has not been verified.

Monthly CSV · Audit and file inventory · Output hashes

Limits of this audit

• Rows are not deduplicated documents or verified non-empty extracted text.
• Readable footers do not verify data pages, raw blobs, URLs, source completeness or text integrity.
• Checkpoint state is recorded only; candidate-set freshness and fetch success are not revalidated.
• Absent partitions and zero outputs are disclosed; no claim of complete acquisition is made.
• Embedding and passage rows have different units; their count difference is not a processing backlog.
• Evaluation-period metadata is operational inventory only; no model design, fitting or evaluation is performed.
• File inventory fingerprints cover metadata and file stats, not full-file content hashes.

Absent output partitions and read errors

Not yet produced

Evidence-span classification has not been run for 2024-04 – 2026-09 (30 months; media and institutional). No span tables exist for these months.

For 2022-02 – 2024-03 (26 months; media and institutional) the span tables exist and contain 0 rows.

No read errors were recorded.