Repeated knowledge-ingest of identical content (same sha256, new
asset_id -> new canonical object key) reuses the Document row but
appends a second ORIGINAL_PDF artifact, because ensure_artifact identity
is (document_id, storage_key). process_document_id then crashed with
MultipleResultsFound on scalar_one_or_none and Celery retried forever.
New latest_artifact helper picks the newest row deterministically
(created_at DESC, id DESC) — every duplicate references byte-identical
objects (sha256 verified at ingest), so the latest reference is always
safe.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The artifact-upsert helper was duplicated four times (scanner.py,
table_processor.py, figure_processor.py, pipeline.py) with slightly
different signatures. Consolidates into a single keyword-only function
keyed on (document_id, storage_key) - the identity the schema already
enforces - so re-running the pipeline never creates duplicate rows.
scanner / table_processor / figure_processor now import the shared
helper directly. pipeline.py keeps a thin local wrapper to preserve
the positional call sites at three artifact upsert points (OCR_PDF,
MARKDOWN, DOCLING_JSON).
Tests: 24 passed (5 health + 19 original).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
