1.3 KiB
1.3 KiB
WP-03 — Ingestion, provenance & publishing
Purpose: Turn raw inputs into trustworthy reproducible snapshots.
Dependencies: see INDEX.md. Status: planned. Owner: to assign. Estimate: to refine after WP-00.
Work packages
- Implement build-time JSON/NDJSON and Markdown frontmatter connector APIs.
- Make import idempotent with source digests and connector version.
- Resolve links in second pass; report dangling/conflicts; distinguish tombstones.
- Model source licenses, citations and attribution; quarantine unsupported content.
- Generate snapshot, manifest and machine/human import reports; atomic publication.
Acceptance gates
- Two connector types produce identical normalized snapshots for equivalent input.
- Repeated ingest yields same content hash.
- Every published sourced claim has a retrievable evidence reference.
- Rights violations prevent publication.
Key risk / mitigation focus
R5: rights metadata missing; quarantine instead of guessing. Log evidence and deviations in the workplan review before marking complete.
Evidence artifacts
Pull requests, test logs, pinned artifacts, screenshots or comparison report where applicable. Add explicit ImplementationNotes.md and Review.md on completion; do not mark a gate passed without a verifiable link.