Input manifest schema

The existing discovery engine writes workflow_metadata/slides.tsv and workflow_metadata/slides.json. The TSV is the human-reviewable roster; the JSON has a top-level slides array with the same rows.

Field	Meaning
sample_id	Collision-safe output/matrix identifier
slide_path	Resolved primary L0 path
relative_path	Primary path relative to input root
size_bytes	L0 file size
mtime_ns, ctime_ns, device, inode	Filesystem identity metadata
fingerprint	Stream-safe primary identity used in workflow caching
l2_path, l2_exists, l2_size_bytes, l2_mtime_ns	Companion location/state
l2_content_sha256	L2 content checksum when the companion is used
l2_fingerprint	Companion identity or explicit missing/not-used state

The inspect command adds review-oriented format, pyramid, MPP, duplicate, and storage information in its own manifest/report while retaining discovery semantics.

Sample sheet input

The user sample sheet is UTF-8 CSV or TSV containing at least:

sample_id,slide_path
sample_001,case_001/1_L0_rgb.tif

Paths may be relative to the input root or absolute within it. Empty or duplicate IDs, duplicate primary paths, reserved IDs, files outside the input root, and missing required L2 companions are rejected.

Sample IDs and paths can be sensitive. Use neutral aliases and do not commit patient identifiers or study manifests.