Skip to content

Archived: integrated graph ROI writer and superseded reopen receiving - #404

Closed
trissim wants to merge 11 commits into
mainfrom
fix/result-directory-publication-134
Closed

trissim wants to merge 11 commits into
mainfrom
fix/result-directory-publication-134

Conversation

@trissim

@trissim trissim commented Oct 1, 2026 •

Copy link
Copy Markdown
Collaborator

Superseded receiving/archive disposition

Closed without merge. Published head dd3242d and all original source proposals, archives, failed graph reopen evidence and comments remain preserved. Do not apply this old branch wholesale to main; production writer integration was delivered by Root394.

The later merged541 receipt docs/validation/neurite-public-pixel-acceptance-20261005.rst is the determining actual saved graph inventory -> native reopen acceptance: three feature-bearing Shapes paths/69coordinates, original channel2/site1/Z1/time1/A01, personally opened matched raw/graph/combined, exact closed custody. This replaces the older assertion that no installed graph receipt exists. Issue134 is already CLOSED.

This is pixel-space engineering graph persistence/reopen, not biological tracing accuracy, physical-unit calibration or exact historical calibrated Site001 acceptance. The receipt explicitly preserves empty acquisition spacing and native legacy micrometers default readback. Merged893 adds its calibrated synthetic observability acceptance but is not a graph-specific physical-unit proof.

That remaining legacy source-default-calibration boundary is separately named and Singer-owned in issue907. It does not keep this superseded archive PR dangling or block ready science/package work. No implementation/source/test/build/runtime/replay was performed for this disposition; original refs/history are retained.

@trissim

trissim commented Oct 1, 2026

Copy link
Copy Markdown
Collaborator Author

Actual #134 repair proposal is now visible at 4510610: docs/validation/persisted-roi-directory-owner-134.patch (three exact shared hunks). It applies cleanly to original main-derived source and pinned latest #394 fdf5024. It declares the original compiled artifact results destination on RuntimeArtifactMetadataTarget, carries the destination through for_directory, and makes the existing OutputTarget writer serialize its complete plate-relative path instead of dropping nested ancestry with basename. No consumer/registry/renderer edits or second writer.

The explicit source projection (unapplied in production; existing WT only) executes the original shared owner and public inventory. Four focused cases PASS, 1 retained-live admission case explicitly deselected because already qualified: original two mixed-output reds plus a new discovered declaration with simple/nested destinations. New declaration composes original ProducedImageMetadataCapability and an independent audit capability; cooperative super executes before/after audit around the original writer, then public kind=result inventory admits the exact ROI. Generic consumers remain unchanged. Original unmodified-owner control preserves 1 PASS / 1 RED: nested new declaration loses its directory due to basename projection. Projected check 3.90s / 230468 KiB; original control 5.67s / 288748 KiB, both kernel 512MiB/one CPU/no swap/60s/plugin/provider-free.

Production and foreign WTs remain untouched pending named #394 contact/release. Shared source proposal is a minimal first checkpoint for the actual mixed-image/ROI witness; ROI-only publication needs the coordinated saved-output participation/empty-image transaction extension, and graph archive provenance remains separate. This is continuing implementation, not diagnostic-only completion or installed readiness. No parent viewer-qa/manifest files or frozen Dalton runtime touched. Original R0 of the unapplied exact proposal tree and byte-exact archival receipt follow.

@trissim trissim changed the title Diagnose persisted ROI directory admission at native publication (#134) Propose declared result-directory publication through existing owners (#134) Oct 1, 2026
@trissim trissim changed the title Propose declared result-directory publication through existing owners (#134) Track persisted ROI acceptance after PR394 publication integration Oct 2, 2026
@trissim trissim changed the title Track persisted ROI acceptance after PR394 publication integration Qualify graph ROI source roundtrip for Root134 integration Oct 2, 2026
@trissim

trissim commented Oct 2, 2026

Copy link
Copy Markdown
Collaborator Author

Docs-only recorder correction pushed at f3e5664. Read-only verified actual /home/ts/wt/openhcs-issue-batch-20260929/neurite-development-skill383-20261001/output/assisted-2-post426-1.stdout:7605 names the exact candidate6 graph ZIP and requires native ROI source metadata; line7602 records plate_file_stream_failed, line7603 ValueError. Corrected receiving receipt. Prior ddbc receipt/history and original archives retained unchanged. No production/test/patch change, repeated test, install/native operation or competing integration. Root retains394 writer ownership; this graph proposal is not included in parent430/PolyStore0.3.1 acceptance.

trissim added a commit that referenced this pull request Oct 4, 2026
Preserve exact public label receipt reopen, full projected pixels, materialized graph ROI source/features, matched render-complete captures and terminal ownership evidence. Retain failed controller/fixture checks and state the synthetic source/bounds/negative-control limits.

Refs #134, #404, #576
@trissim

trissim commented Oct 5, 2026

Copy link
Copy Markdown
Collaborator Author

Cross-machine production-freeze handoff (5 Oct)
This is the older DRAFT graph-writer/receiving follow-up, not an independent current writer implementation to apply over main. Root394 and later native source/geometry #748 are already merged. Planck owns the remaining reconciliation of exact persisted reopen acceptance and superseded branch closure.

Please compare the current determining main diff and retain the historical failed reopen evidence rather than rerunning or merging the old branch blindly. Do not close #134 solely because an older source-only graph writer proposal or wheel existed; use the later actual native reopen receipt where it satisfies that issue. This draft is not a blanket hold on ready PR merges or production freezing.

@trissim

trissim commented Oct 5, 2026

Copy link
Copy Markdown
Collaborator Author

Current determining status for the production freeze: the writer implementation is already integrated by Root #394 (original integration45f3a721); this draft has no remaining production diff to apply over main. There is no current Root seam-release hold.

Singer confirms no later public inventory -> owned native source-bearing graph ZIP reopen/Site001 receipt in the retained receiving context. Therefore #748's accepted Points/scalar/Shapes path is NOT being substituted for graph topology/features/calibration acceptance, and this receiving follow-up stays open at that exact boundary. #134 being closed alone is not behavioral proof. Preserve original graph failure and source archives; do not merge or replay the historical proposal. #541's graph-unit work is separate and in progress.

#472's old wheel staging is superseded and closed independently. This receiving-only draft is not a pending writer implementation or a blanket production-freeze hold.

@trissim

trissim commented Oct 5, 2026

Copy link
Copy Markdown
Collaborator Author

Additional retained receiving witness, NOT a new graph defect: frozen H002_FRESH15_89 HANDOFF.rst64 records diagnostic-image refusal. Exact original mcp.stdin264 requests kind=result with explicit result_directory=attempt02/plate_v1t2/checkpoints, path_contains=h002_support.tif, no source_receipt; mcp.stdout78308 returns Explicit result images require an exact source receipt. Current original PlateStreamingService195-202 refuses that explicit-directory image route before _inventory_source_projection235-243. #399 ordinary inventory/native metadata projection is merged. This does not prove missing persisted provenance, a pixel swap, or a graph-reader defect. Singer retains the receiving route/source-metadata comparison in #404: use existing typed native inventory/projection under a later assigned lane, not invented receipts, filename inference or guard relaxation. Closed original client/native/viewer and all failed inputs remain immutable; no replay/new client or production patch for this witness. #541 unit implementation stays independent.

@trissim

trissim commented Oct 5, 2026

Copy link
Copy Markdown
Collaborator Author

Read-only current-main reconciliation for Planck, existing #404 disposition owner (no source changes or receiving replay). Main674ddefe5 contains the later accepted #541 receipt: docs/validation/neurite-public-pixel-acceptance-20261005.rst lines31–38 establish ordinary public inventory of the freshly materialized graph ZIP, three feature-bearing native Shapes paths/69 coordinates, channel2/site1/Z1/time1/A01, matched original raw/result/combined and exact closed custody. That is a determining later graph writer-to-reopen path, not the Points-only #748 substitute excluded by the older comment. Lines43–62 explicitly retain absent acquisition calibration and legacy native micrometers default readback; do not convert this into physical calibration or exact historical Site001 acceptance. Graph/provenance tests are already integrated and migrated on current main; this old branch must not merge wholesale. Please reconcile the fulfilled persisted graph inventory/reopen subset against #404 closure and keep any materially distinct calibrated-source/subject-feature obligation narrowly named rather than leaving the pre-#541 no-receipt status unchanged. Source-unit default follow-up remains Singer-owned; no new implementation, build, runtime or broader biological claim.

@trissim trissim changed the title Accept integrated graph ROI writer; track native reopen134 Archived: integrated graph ROI writer and superseded reopen receiving Oct 5, 2026
@trissim

trissim commented Oct 5, 2026

Copy link
Copy Markdown
Collaborator Author

Superseded by integrated Root394 writer plus later actual merged541 graph inventory/reopen receiving. Not merged wholesale. Historical dd3242d/source/failure archives preserved. Pixel graph feature/placement success is distinct from unverified historical calibrated Site001 and physical units; remaining Singer-owned legacy source-default calibration is issue907. No rerun/build or source changes.

@trissim trissim closed this Oct 5, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant