recover: enrich Lily swatches from local copy (Henry USB unmounts mid-run)

324313fda2c1498bf4f863e907617b9053586732 · 2026-07-01 14:17:38 -0700 · Steve Abrams

Files touched

A onboarding/sangetsu-lilycolor/scripts/local-recover.sh

Diff

commit 324313fda2c1498bf4f863e907617b9053586732
Author: Steve Abrams <steve@designerwallcoverings.com>
Date:   Wed Jul 1 14:17:38 2026 -0700

    recover: enrich Lily swatches from local copy (Henry USB unmounts mid-run)
---
 .../sangetsu-lilycolor/scripts/local-recover.sh    | 48 ++++++++++++++++++++++
 1 file changed, 48 insertions(+)

diff --git a/onboarding/sangetsu-lilycolor/scripts/local-recover.sh b/onboarding/sangetsu-lilycolor/scripts/local-recover.sh
new file mode 100755
index 00000000..0c667366
--- /dev/null
+++ b/onboarding/sangetsu-lilycolor/scripts/local-recover.sh
@@ -0,0 +1,48 @@
+#!/bin/bash
+# Re-enrich the 1828 error SKUs from the LOCAL image copy (~/dw-lily-images-local),
+# not the flaky Henry USB mount — so a mid-run USB unmount can't fail us again.
+# Seeded off the 806 good rows, remerges (last-wins) into final. Local, $0, reversible.
+set -u
+PROJ="$HOME/Projects/Designer-Wallcoverings/onboarding/sangetsu-lilycolor"; cd "$PROJ" || exit 1
+ST="$PROJ/staging"
+LOCAL_IMG="$HOME/dw-lily-images-local"
+GOOD="$ST/enrichment-full-063026A.good-seed.jsonl"
+REC="$ST/enrichment-full-063026A.local-recover.jsonl"
+FINAL="$ST/enrichment-full-063026A.final.jsonl"
+log(){ echo "[localrec $(date '+%H:%M:%S')] $*" >> "$ST/enrich-localrec.log"; }
+
+[ -d "$LOCAL_IMG" ] || { log "FATAL local image dir missing: $LOCAL_IMG"; exit 1; }
+[ -f "$GOOD" ] || { log "FATAL good-seed missing: $GOOD"; exit 1; }
+log "start. local images=$(ls -1 "$LOCAL_IMG" | wc -l | tr -d ' ') good-seed=$(wc -l < "$GOOD" | tr -d ' ')"
+
+# Enrich from LOCAL dir only (point Sangetsu dir at a nonexistent path so only lily is done).
+ENRICH_HENRY_DIR="$LOCAL_IMG" \
+ENRICH_SANGETSU_DIR="/nonexistent-sangetsu" \
+ENRICH_SHARD_COUNT=1 \
+ENRICH_OLLAMA_URL=http://127.0.0.1:11434 \
+ENRICH_DONE_SEEDS="$GOOD" \
+python3 -u scripts/enrich-full.py "$REC" >> "$ST/enrich-localrec.log" 2>&1
+log "enrich pass done. recover rows=$(wc -l < "$REC" 2>/dev/null | tr -d ' ')"
+
+# Remerge good + recover (last wins) -> final
+python3 - "$GOOD" "$REC" "$FINAL" <<'PY' 2>> "$ST/enrich-localrec.log"
+import json,sys
+srcs,final=sys.argv[1:-1],sys.argv[-1]
+by={}
+for p in srcs:
+    try:
+        for line in open(p):
+            line=line.strip()
+            if not line: continue
+            try: o=json.loads(line)
+            except: continue
+            s=o.get("sku")
+            if s: by[s]=o
+    except FileNotFoundError: pass
+ok=sum(1 for o in by.values() if not o.get("error") and o.get("usable") is not False)
+with open(final,"w") as f:
+    for s in sorted(by): f.write(json.dumps(by[s])+"\n")
+print(f"[remerge] final unique={len(by)} ok={ok} err={len(by)-ok}")
+PY
+log "REMERGE done. final=$(wc -l < "$FINAL" | tr -d ' ') rows"
+touch "$ST/LOCAL-RECOVER-DONE"; log "ALL DONE"

← dcd19e57 enrich-full.py: treat mid-run Henry unmount as transient (Vo · back to Designer Wallcoverings · audit: Sangetsu spec/desc gap is source-data (not merge) — L d7dea0f1 →