PlantAI-dev/docs/full_chain_baseline_job_1779966938704_aa384a_2026-05-29.md

Full-chain baseline record: job_1779966938704_aa384a

Date: 2026-05-29

This document records the completed bbav5 gene-family analysis job used as a regression baseline after the data preparation to subfamily-classification chain was run end to end.

Recorded paths

Input job directory:

/home/xia-lab/wutx/00test/bbav5_0502_Yang_Tihuan.7880/input/job_1779966938704_aa384a

Workspace job directory:

/home/xia-lab/wutx/00test/bbav5_0502_Yang_Tihuan.7880/workspace/job_1779966938704_aa384a

Main output directory:

/home/xia-lab/wutx/00test/bbav5_0502_Yang_Tihuan.7880/workspace/job_1779966938704_aa384a/final

Why this job is recorded

This job should be treated as a real-data baseline for later changes around:

• config.xlsx / upload auto-fill behavior.
• Per-species versus multi-species planning.
• Scope-aware artifact lookup.
• Subfamily review popup evidence display.
• Naming cleanup from identify through phylogeny, motif, structure, domain, combo, and reference classification.

It is not a final naming-standard example. The output currently contains both newer scope-aware artifacts and older compatibility names, so it is best used to verify behavior and artifact discoverability before and after revisions.

Inventory snapshot

Commands used to capture the snapshot:

find input/job_1779966938704_aa384a -type f | wc -l
find workspace/job_1779966938704_aa384a/final -maxdepth 1 -type f | wc -l
find workspace/job_1779966938704_aa384a/final -maxdepth 1 -type f -name '*.manifest.json' | wc -l

Observed counts:

• Input files: 131
• final/ files: 770
• Manifest files in final/: 104

Manifest stage/status distribution:

Stage	Status	Count
config_autofill	success	1
genome_prepare	success	1
identify	success	11
identify_refine	success	5
phylogeny	success	17
motif	success	12
structure	success	12
domain	success	12
msa_region_plot	success	10
reference_classification	success	20
subfamily_confirm	success	3

All inspected manifest JSON files were parseable.

Input-side baseline

The input job contains:

• config.xlsx
• At.SPL.pep.fa
• At.SPL.subfamily
• Raw target/reference files for Arabidopsis, Cp, Dl, Lc, Os, PPY, and Sl.
• Normalized species folders including At/, Cp/, Dl/, Lc/, Os/, PPY/, and Sl/.
• Ref_combined.SPL.pep.fasta
• genome_prepare/ outputs.
• final/ copies of identify-related outputs.

This path is important for validating that upload/config auto-fill uses the current job input directory and does not accidentally reuse stale paths from previous jobs or reference-data folders.

Output-side baseline

The workspace/job_1779966938704_aa384a/final directory contains completed outputs across the main chain:

• Genome preparation:
  • genome_prepare.genome_prepare.manifest.json
• Config auto-fill:
  • config_autofill.manifest.json
• Identify and identify refinement:
  • Per-species *.SPL.identify.manifest.json
  • Per-species *.SPL.identify_refine.manifest.json
  • *.final_members.*
• Target plus reference analyses:
  • *.target_plus_reference.phylogeny.manifest.json
  • *.target_plus_reference.motif.manifest.json
  • *.target_plus_reference.structure.manifest.json
  • *.target_plus_reference.domain.manifest.json
• Combo outputs:
  • *.target_plus_reference.combo.png
  • *.target_plus_reference.combo.pdf
• Reference subfamily classification:
  • *.target_plus_reference.reference_classification.manifest.json
  • *.target_plus_reference.reference_classification.assignment.tsv
  • *.target_plus_reference.reference_classification.confidence.tsv
  • *.target_plus_reference.reference_classification.subfamily.tsv
• Scope-aware subfamily prediction and evidence outputs:
  • Target.SPL.target-*.ref-At.subfamily_prediction.reference_classification.*
  • Target.SPL.target-*.ref-At.subfamily_evidence.phylogeny.*
  • Target.SPL.target-*.ref-At.subfamily_evidence.motif.*
  • Target.SPL.target-*.ref-At.subfamily_evidence.structure.*
  • Target.SPL.target-*.ref-At.subfamily_evidence.domain.*

Species-level observation matrix

This matrix records what was observed from files and manifests. It separates classification/prediction outputs from user confirmation outputs.

Species	Identify	Phylogeny	Motif	Structure	Domain	Reference classification	Combo image	Subfamily evidence	Confirm manifest observed
Lc	yes	yes	yes	yes	yes	yes	yes	yes	yes
Dl	yes	yes	yes	yes	yes	yes	yes	yes	yes
Cp	yes	yes	yes	yes	yes	yes	yes	yes	yes
Os	yes	yes	yes	yes	yes	yes	yes	yes	no
PPY	yes	yes	yes	yes	yes	yes	yes	yes	no
Sl	yes	yes	yes	yes	yes	yes	yes	yes	no

Observed subfamily_confirm manifests:

Cp.SPL.subfamily_confirm.manifest.json
Dl.SPL.subfamily_confirm.manifest.json
Lc.SPL.subfamily_confirm.manifest.json

Therefore, this job supports the statement that the classification/prediction/evidence chain has run for the listed target species. It should not be used to claim that every species has completed user-confirmed subfamily review unless the relevant subfamily_confirm artifacts are present.

Known API validation

The following representative subfamily review API call was verified against the running web service on port 7880:

curl -s 'http://127.0.0.1:7880/api/subfamily/review?job_id=job_1779966938704_aa384a&prefix=Target.SPL.target-Lc.ref-At.subfamily_prediction'

Observed behavior:

• success was true.
• missing was empty.
• Rectangular combo evidence resolved to Lc.SPL.target_plus_reference.combo.png.
• Circular tree evidence resolved to Target.SPL.target-Lc.ref-At.subfamily_prediction.reference_classification.tree.circular_tree.png.
• Tree evidence resolved to Target.SPL.target-Lc.ref-At.subfamily_prediction.reference_classification.tree.png.
• Confidence and report files were present.

This API case is the key regression check for the previous false-negative popup message: 未找到 combo PNG.

How to use this baseline during future revisions

Before and after changing scope, naming, planning, upload auto-fill, or subfamily review display, run:

/home/xia-lab/miniconda3/envs/test4py311/bin/python scripts/check_autofill_regression.py
/home/xia-lab/miniconda3/envs/test4py311/bin/python scripts/check_subfamily_evidence_scope.py
/home/xia-lab/miniconda3/envs/test4py311/bin/python -m pytest -q tests/unit/test_autofill_entrypoints.py tests/unit/test_subfamily_evidence_scope.py

Then verify a real-data review endpoint:

curl -s 'http://127.0.0.1:7880/api/subfamily/review?job_id=job_1779966938704_aa384a&prefix=Target.SPL.target-Lc.ref-At.subfamily_prediction'

Expected result:

• Review popup data should load successfully.
• The response should not report missing combo PNG when the same-scope combo artifact exists.
• Artifact resolution should prefer manifest/scope-aware paths before legacy glob fallbacks.

For file-level comparison, use:

find input/job_1779966938704_aa384a -maxdepth 2 -type f -printf '%P\n' | sort
find workspace/job_1779966938704_aa384a/final -maxdepth 1 -type f -name '*.manifest.json' -printf '%f\n' | sort
find workspace/job_1779966938704_aa384a/final -maxdepth 1 -type f -printf '%f\n' | sort

Caveats

• The job contains legacy and newer names at the same time, including species-main files and Target.SPL.target-*.ref-At.* scoped files. This is useful for compatibility testing but should not define the final cleanup target.
• subfamily_confirm manifests were observed for Cp, Dl, and Lc only. The remaining species have classification and evidence outputs, but user-confirmation artifacts should be verified separately before claiming confirmed status.
• The baseline should be interpreted with the 2026-05-26 and 2026-05-27 naming/scope documents: keep manifest identity authoritative, avoid private glob-only artifact discovery, and keep user-facing filenames readable while preserving scope_id in manifests.