Jessar Website — Progress

Phase Checklists

Open Questions for Jessar (2 open · 15 resolved)

• 1

  Sub-brand → prefix mapping — ✅ RESOLVED 2026-05-07. Jessar returned the completed mapping (work/sub_brand_mapping_RESOLVED_2026-05-07.csv, 154 groups, 2,560 SKUs, 100% assigned). Headline finding: at the data tier there are four sub-brands, not ten — JS Gourmet (7-, 1,264 SKUs), JS Maison (6-, 236), Jessar Lighting (5-, 358), Xtricity (1-+3-+4-, 702 combined). The other catalogue names (Beauty Cover, Cook & Fresca, IBake, Limpus, Pop Fluff, Veraroma) are marketing identities / product lines that roll up into the four data-tier brands. Action: Sean to review the implication for the homepage (home_sm_v6 currently shows 10-brand portfolio).

  Raised 2026-04-27 · Resolved 2026-05-07 by Jessar
• 8

  Hero product narrowing — Michael picks 20 from 50. Sean delivered the curated list (50 unique product concepts / 88 SKUs); SOW caps full hero PDP treatment at 20. Action: Michael selects 20 at jessar-deliverables.vercel.app/hero_candidates_for_michael; output CSV drops at /work/hero_selections_michael.csv and the build pass flips them to tier=sow_hero. Also: confirm 4 hero SKUs missing from DB (7-99738, 7-98690, 7-99754, 7-99757).

  Raised 2026-04-27 · Intake closed 2026-04-28 · Awaiting Michael's narrowing

Resolved: Q2 eco-fees (orange = ignore) · Q3 discontinued (orange in Items-Jessar) · Q4 source-of-truth (Items-Jessar + Electrical/Beauty Cover Inventory) · Q5 On Hand gap (moot) · Q6 pricing (MSRP from Items-Jessar col AO "Selling Price" → metafield) · Q7 WP export (CSV delivered, REST API not needed) · Q9 tone POC (approved by Jessar) · Q10 FR review (we do the heavy lift; Shopify Translate & Adapt for fallback; cherry-pick human review) · Q11 store locator (replaced with Partners section) · Q12 dealer portal (out of scope, link to existing) · Q13 retailer application (out of scope) · Q14 Costco (live on site) · Q15 catalog versions (newer supersedes if same SKU) · Q16 SSD mount (completed — Seagate drive indexed 2026-05-03) · Q17 brand guidelines + partner logos (we design + procure ourselves).

Recent Decisions

• Page-system native-vs-custom audit (2026-06-29). 6 core templates locked for client review. Contact kept RS (form functional). Dealer (4) + About (6) native Hyper parallels built alongside the RS versions for side-by-side review — nothing removed. Keep-custom: dealer form, About values + brands. Homepage finalized with real retailer logos. 404 branded; /collections hidden.
  2026-06-29 · Sean / RevShift
• Master-carton correction — third source-reading fix (2026-06-03). Sean asked me to double-check that master carton was genuinely missing/unsourced. It was not. The On Hand sheet carries three (L)/(l)/(H) dimension blocks (unit · inner · master); the third (most-filled) block is the master-carton size — proven by the sheet's own per-unit-CBM formula (CBM ÷ Master). I had been counting only the 57 master cartons recovered from box-artwork OCR and missing the On Hand block entirely. build_product_data_supplement.py now picks the most-filled dimension triple and treats On Hand as the primary carton source. Master carton coverage 57 → 906 (54%); remaining genuine gap 1,610 → 761. Re-ran the full downstream chain (inject → aggregate → reviewer → content Excel → data-request) and redeployed both portals. Corrected live data-gap counts: inner 998 · master carton 761 · certs 1,373 (UPC 1,662 + master case 1,661 effectively complete). Third time Sean's instinct to re-check a "missing" claim was right — after the cert/UL-ETL sheet correction and the "~1,145 recoverable certs" overcount. Honest residual gap = inner-case (sparse at source) + certifications (their "To Check" column) + ~761 carton rows for catalog-only SKUs absent from On Hand. Shopify untouched.
  2026-06-03 · Session 21 · Sean / RevShift
• Missing data surfaced + Jessar data-request built (2026-06-02). Client flagged inner case / master case / certifications / UPC as missing. Key correction: I'd been reading the wrong Items-Jessar sheet — the operational On Hand sheet holds UPC/Inner/Master + a UL/ETL cert column. Findings: UPC (1,662) + master case (1,661) we have → now surfaced into the v2 specs, the client reviewer, and a new packaging block in the content Excel. Inner case is genuinely sparse even in their source (669/1,667; many items have no inner pack). Certifications: the UL/ETL column exists but Jessar left it “🔴 To Check” for 812/876 — unfilled on their end; our honest extract is 294 (my earlier “~1,145 recoverable” was a bad estimate that matched the French word “ce”). Genuine gaps packaged into jessar_data_request_2026-06-02.xlsx (998 inner · 1,610 carton · 1,373 cert; amber = please fill, cream = we have, heaviest-gap first), linked from the client hub. New scripts: build_product_data_supplement, inject_supplement_to_v2, build_jessar_data_request_xlsx. Reframes the client conversation: the columns exist in their On Hand sheet — the cert one is still “To Check.”
  2026-06-02 · Session 20 · Sean / RevShift
• Shopify metafield cleanup + content-push plan locked (2026-06-02). Audited the 48 live jessar.* metafield definitions + took a full rollback snapshot of all 1,668 live products. Deleted 10 internal/useless definitions (~6,177 stray values purged store-wide): migration_action, completeness_score, box_features (the noisy OCR field originally flagged — now gone from live), care_source, material_source, color_source, source_catalog, priority_tier, short_description, cord_length. Kept internal_clas_1/2 + future-use empties (certifications/upc/is_hero/catalog_pdf) + all customer-facing → 38 definitions remain. v2 content push plan (locked): wait for client edits → one clean push, EN+FR together, box_features already removed; scope set when edits return. Surfaced for later: store shows 1,668 products (one extra, likely test SKU — left for now) and product_line live 148 vs clean 149 (reconcile at push). Rollback: live snapshot + setup_shopify_phase5.py.
  2026-06-02 · Session 20 · Sean / RevShift
• Client feedback channels for content review (2026-06-02). Two ways for Jessar to return copy changes, both batch-appliable: (1) inline editing in the v2 reviewer — a "✎ Suggest edit" control on every product opens editable fields (title/romance/description/features) + approve-as-is + notes, autosaves in-browser, and Export my changes downloads jessar_content_edits.json (built with safe DOM construction, no XSS). (2) A colour-coded Excel (jessar_content_review_2026-06-02.xlsx) mirroring the reviewer fields EN+FR — muted read-only source columns vs 3 green editable columns (Approve Y/N · Corrections · Revised copy), needs-verification rows amber + sorted to top, product-line tinted, README legend, frozen panes. Both linked from the client hub; client picks their preference, both return to us for batch apply.
  2026-06-02 · Session 20 · Sean / RevShift
• Residual 16 majors cleared — 0 known factual errors (2026-06-02). Resolved the last 16 majors: 9 visible spec errors fixed against cited sources (7-99112 dropped an unsourced "13.7\" blade"; 7-99403 single→double-wall glass; 7-99696/697 wattage 1000→1500/1200W, honestly re-flagged for vendor confirmation; 7-99577/7-99687/1-60412 removed contradictory/unsupported claims) + 7 empty product_line fields populated from box/title (Veraroma, Beauty Cover, iBake), synced to the authoritative inferred_product_lines.csv. The fix→verify discipline paid off again: re-QA caught 2 self-inflicted mismatches (7-99851 left "Celebration Collection" in the description after the romance edit; 7-99913 set product_line=iBake but prose still said VerAroma) — both corrected and re-verified. Final independently-verified state: 0 major · 194 minor (11.6%) · 1,473 clean (88.4%); product_line clean across 6 lines; 0 integrity issues; 39 needs_human_review. Tone/descriptions/substance confirmed client-ready on a cross-tier read. Added a "Product Content — Review & Verify" card to the client hub. Both portals redeployed. Rollback 2026-06-02-pre-residualfix snapshot; Shopify untouched.
  2026-06-02 · Session 20 · Sean / RevShift
• Fix + polish + verify — 694 flagged SKUs cleaned to top quality (2026-06-02). Fix workflow corrected every QA-flagged issue (120 major + 574 minor) with the locked brand voice applied (restrained/factual; banned-words list — no premium/curated/world-class/cutting-edge/discover). Root-cause rule: BOX/PRESENTATION authoritative over xlsx/catalog on physical specs. Wrote to a separate product_content_fixed/ namespace, then merged to v2. Independent re-QA verified the result — major 120→16 (7.2%→1.0%), minor 574→189 (34%→11%), clean 973→1,462 (58%→88%); ~99% publish-ready. Caught + fixed a fix-pass regression: agents had stuffed product_line with junk (model names Valerio/RIPPLE, a description string, VerAroma casing — 18 bad values) → re-asserted from the authoritative source to the clean 149 across 6 lines. Residual 16 majors are mostly product_line ATTRIBUTION edge cases (copy names a line but SKU not in canonical list; 7-99913 Veraroma-vs-AirPlus) + ~6 spec calls (7-99696/697 wattage, 7-99104 dims, 7-99112 blade) — the human/Jessar decision residue; 36 SKUs carry needs_human_review. Refreshed reports live: /jessar-qa-report + /jessar-pdp-v2-review. Rollback 2026-06-01-pre-fix intact; Shopify untouched.
  2026-06-02 · Session 19 · Sean / RevShift
• Category mis-map fix — TAPS is electrical, not faucets (2026-06-02). 17 electrical outlet-tap/adapter SKUs (3-70611…3-70624, 3-70701/703/704) were mis-categorized "Kitchen — Faucets & Taps". Root cause: CATEGORY_MAP["TAPS"] in generate_product_content.py + fill_content_gaps.py mapped the electrical "TAPS" classification code to plumbing faucets, and also routed it to the kitchen writing voice. Fixed at source (4 spots, 2 scripts → "Electrical — Power" + electrical voice) and in derived data (product_content.csv, regen_input, content_batches). Verified: zero genuine faucet products exist in the catalog (nothing wrongly reclassified); v2 content was not voice-contaminated (gold-based regen produced correct electrical tone).
  2026-06-02 · Session 19 · Sean / RevShift
• Adversarial QA of v2 PDP content — all 1,667 SKUs (2026-06-01). Independent QA workflow — 67 auditor agents that did NOT write the content (read-only, Max plan, zero API) — scrutinized every v2 block against its raw sources + old data, hunting source-priority violations, unverifiable claims, dropped features, spec conflicts, brand-line errors, FR quality. Survived a power-failure interruption (stalled at 318 → resumed via missing-only manifest + a durable scheduled task, now disabled). Result: 58.4% clean · 34.4% minor · 7.2% major (120 SKUs with customer-visible factual errors). Per-tier major rate: Jessar Lighting 13.0% (worst), JS Maison 11.1%, Xtricity 6.7%, JS Gourmet kitchen/home 6.0% (best substantive tier). Defect types: dropped_feature 367 · unverifiable_claim 253 · spec_conflict 161 · internal_inconsistency 124 · source_priority 74 · brand_line_error 31 · fr_quality 18. Systemic root cause for most majors: on lighting/bulbs the pipeline trusted xlsx/catalog over the BOX/PRESENTATION on physical specs (e.g. labeling a "DO NOT USE WITH DIMMER" bulb as dimmable; 1-year vs box's 3-year warranty). Self-grading had rated these needs_review=false, so the adversarial pass caught ~120 errors that would otherwise have shipped — vindicating the decision to QA rather than trust self-reports. Report (filterable major/minor, source-cited fixes) live at revshift-internal.vercel.app/jessar-qa-report. No content modified — verdicts in work/qa_findings/ only; rollback point pre-qa-2026-05-27 + data snapshot intact. New scripts: aggregate_qa.py, render_qa_report.py. Next: fix plan pending Sean's go.
  2026-06-01 · Session 18 · Sean / RevShift
• v2 PDP content regeneration — all 1,667 SKUs (2026-05-27). Final assembly pass via background Workflow (56 + gap-fill batches, Max plan, zero API). For every SKU, regenerated the complete front-facing PDP block — title · romance EN/FR · description EN/FR · features EN/FR · specs — FROM the validated gold as atomic source of truth. Old v1 romance/description used as tone inspiration only, never copied; the pre-validation errors gold caught (cast-iron mislabeled "ceramic," carton dims as product size) do not reappear — verified on 7-99512 (enameled cast iron, correct specs). Hard namespace separation: writes only to work/product_content_v2/ → product_content_v2.csv; old product_content.csv is read-only reference (powers the viewer's Compare-with-v1 toggle), never written. Normalized product_line casing (VerAroma→Veraroma, IBake→iBake) + cleared a model-name false-positive (RIPPLE). Final: 1,667/1,667, 0 integrity issues, 149 brand-line SKUs (Beauty Cover 74 · Veraroma 29 · Inspiration Italia 25 · Laguiole 8 · iBake 7 · Limpus 6), 53 needs client verification (each carries its flag note). New scripts: build_regen_input.py, aggregate_pdp_v2.py, render_pdp_v2_review.py. v2 review viewer (per-SKU verify banner + note + Compare-with-v1 toggle + brand-line/needs-verification filters) live at revshift-internal.vercel.app/jessar-pdp-v2-review. Push-time freeze+guard deferred to export build. Shopify untouched — pending Sean review → client verification of the 53 → sync.
  2026-05-27 · Session 17 · Sean / RevShift
• Second pass — brand-line identity + manuals + electrical xlsx (2026-05-27). Pre-review diligence found brand-line identity was barely surfaced — only 9 of 148 line-branded SKUs (~6%); the Laguiole knives worked only because "Laguiole" was incidentally printed on the box. Root cause: product_line was never fed into the feature bundles and sub_brand was the stale generic. Fixed by enhancing the bundle builder with product_line (inferred_product_lines.csv), corrected sub_brand (4-brand prefix map), manual OCR text (work/manual_ocr — 629 manuals extracted 2026-05-13 but never used; 392 primary SKUs), and the Electrical/Beauty Cover Inventory xlsx (newly mined → 863 rows). Re-enriched 196 priority SKUs (148 product-line + 85 JS Gourmet/Maison-with-manuals). Brand-line coverage now 148/148 (100%) — every Veraroma/Beauty Cover/IBake/Inspiration Italia/Limpus/Laguiole SKU surfaces its line as a dedicated brand_line feature. Accuracy guard: dropped false protected-origin claims (Laguiole/Inspiration Italia = style/collection, all made-in-China per box — no false France/Italy claims). Final gold: 9,370 features, 172 brand_line, 4,715 junk dropped, 796 OCR repairs, 59 needs_human_review. Flagged a few product_line mislabels (Veraroma vs Inspiration Italia) for Sean. Shopify untouched.
  2026-05-27 · Session 16 (2nd pass) · Sean / RevShift
• Full LLM feature validation + enrichment pass — all 1,667 SKUs (2026-05-27). Ran entirely on the Max plan via batched general-purpose subagents (zero API spend). Synthesized one authoritative, bilingual (EN + Quebec FR), front-facing feature set per SKU from ALL sources — presentations = primary source of truth, Items-Jessar xlsx = authoritative for spec/material verification, plus catalogs, WooCommerce, and deterministically-cleaned box OCR. New sources mined this pass: 1,957 SSD presentation PDFs (653 primary SKUs), "Presentations for Sean" deck, xlsx desc-an/fr/description columns. Result: 9,304 features (5.6/SKU, 94% high-confidence), 4,727 junk lines removed, 794 OCR misreads repaired, 43 SKUs flagged needs_human_review. Subagents caught substantive errors that would have shipped to live PDPs: cast-iron mislabeled "ceramic," carton dims passed off as product size, "grounded 3-conductor" copy on an ungrounded cord (safety), AI features describing a glass jar on a cutting board, fabricated rated-life/warranty figures, wrong-market voltage specs. Pipeline scripts: extract_presentations.py, build_xlsx_descriptions.py, clean_box_features.py, build_feature_input_bundles.py, aggregate_gold_features.py, render_gold_features_review.py. Output: work/features_gold/*.json + work/features_gold.csv. Shopify NOT touched — Sean reviews the gold set first at revshift-internal.vercel.app/jessar-gold-features-review.
  2026-05-27 · Session 16 · Sean / RevShift
• Product-line cleanup pass (2026-05-15). Three targeted corrections on the live dev store (78 products, 0 errors, all verified live). (1) Cleared jessar.product_line on 12 false positives — 5 Mason/Dublin ("Mason" = generic jar type + a bath pillow; "Dublin" = a Jessar Lighting drum-fixture model name + a glass-cup style — neither a marketing brand) and 7 "Beauty Cover" that were actually electrical junction boxes/covers (Xtricity), mis-tagged by an unreliable 'BC' supplier-code rule. (2) Rerouted 65 genuine Beauty Cover SKUs (prefix 7-) from vendor JS Gourmet → JS Maison — Beauty Cover is home/bath/decor, not kitchen. (3) Promoted 1 USB-touch-lamp outlier (7-99858 "Wooden Tripod Nightstand Lamp") JS Gourmet → Jessar Lighting to match the other 37. Findings: the original "81 Beauty Cover / 5 false positives" framing was off — actual was 74 genuine (65 prefix-7 to reroute + 9 prefix-6 already JS Maison) + 7 electrical false positives; and Shopify Standard Taxonomy has no "Table Lamps" leaf — hg-13-5 "Lighting > Lamps" is the most specific node and the 38 USB lamps were already there + already Jessar Lighting (37/38), so that task was effectively pre-satisfied by the 2026-05-13 vendor fix. Hardened scripts/extract_product_lines.py (dropped Mason/Dublin from KNOWN_LINES; removed the supplier-code source wholesale — 7/7 false positives); inferred lines 160 → 148. New scripts: clear_false_product_lines.py, reroute_vendors_2026-05-15.py. Resume-safe logs: work/product_line_clear_log.csv, work/vendor_reroute_log.csv.
  2026-05-15 · Session 15 · Sean / RevShift
• Product-line discovery + push (2026-05-14). Sean spotted that JS Gourmet has sub-brands / marketing lines (Laguiole knives, Beauty Cover bedding, Veraroma drinkware, IBake bakeware, etc.) that we were flattening to a single sub-brand vendor. Built scripts/extract_product_lines.py to mine these from existing data: box artwork OCR text (90 SKUs detected), WordPress titles in wp_name (63), and supplier code prefix in master xlsx code-pro-frn (7). Discovered 160 SKUs (10% of primary) with identifiable brand lines — Beauty Cover (81), Veraroma (35), Inspiration Italia (18), Laguiole (8), IBake (7), Limpus (6), Mason (3), Dublin (2). Created jessar.product_line metafield definition (48 jessar.* defs total) and pushed all 160 values via batched metafieldsSet (scripts/push_product_lines.py) — 160/160 OK in 7 seconds. Live verified: 7-99096 = vendor JS Gourmet + product_line Laguiole; 7-10001 = vendor JS Gourmet + product_line Beauty Cover. Also rebuilt the Jessar-facing review file as .xlsx instead of CSV (work/jessar_review_2026-05-14.xlsx, scripts/build_jessar_review_xlsx.py) — 3 sheets (README, Products, Certifications), value-inclusive cells (shows the actual extracted value, not just ✓), color-coded green/amber/red for source-of-truth, low-intensity instructions, frozen header. Added 4 new columns Sean requested: warranty_period, certifications, heat_resistance, needs_review_or_correction, general_notes_from_jessar. Heroes sorted first. Product-line column added between sub_brand and category.
  2026-05-14 · Session 14 · RevShift
• Shopify migration push complete + post-push refinement (2026-05-13). Full first-pass migration to Shopify dev store: 1,667 primary SKUs live as ACTIVE products. End-to-end pipeline ran across 4 stages: bulk image upload (1,735 OK + 29 failed at 87.9 min), product create (1,667 OK at ~70 min), Shopify Standard Taxonomy category assignment (1,665 OK at ~22 min), variant SKU/barcode fix (1,667 OK at ~36 min). Found two notable bugs mid-run: Shopify GraphQL 2025-01 moved variant.sku → inventoryItem.sku (the original push silently dropped SKUs — scripts/fix_variant_skus.py repaired all 1,667); and assign_product_categories.py needed resume-safe skip-done logic after parent-shell killed it at 941/1,667. Post-push refinement pass (per Sean): (a) Push errors validated — 0 real misses; the 6 "errors" were the same 3 SKUs from pre-fix API-version-discovery attempts that succeeded on retry. (b) 29 failed images recovered — failures were large Exif-laden JPEGs Shopify's image processor rejected; scripts/retry_failed_images.py re-encodes via sips (max 2048px, format=jpeg, options=85) before re-upload + attaches to products via productCreateMedia. 100% image coverage now. (c) Vendor fix (scripts/fix_vendor_to_subbrand.py) — replaced "Jessar Industries" parent vendor with actual sub-brand per SKU prefix (resolved 4-brand mapping from 2026-05-07): JS Gourmet (883), Xtricity (483), Jessar Lighting (254), JS Maison (18), Jessar Industries fallback for 2-prefix (29). Affects Shopify vendor filter, Search & Discovery facets, Google Merchant feed. (d) 20 metafields pinned at top of product editor (scripts/pin_metafields.py) — Shopify's hard 20-pin limit consumed by: sub_brand, msrp_cad, qty_per_package/container, dimensions (×2 unit systems), weight (×2 unit systems), capacity, set_count, master_carton_dimensions, materials, color_finish, lid_material, care_instructions, safety_claims, heat_resistance, warranty_period, certifications, feature_callouts. Lighting-specific fields (wattage/voltage/lumens/etc.) remain accessible under "View all metafields" but unpinned. (e) Taxonomy validated — distribution looks correct, 8 spot-flagged cases all checked out. Outputs: work/shopify_products_pushed.csv, work/category_assignment_log.csv, work/variant_fix_log.csv, work/vendor_fix_log.csv, work/sku_image_urls.csv, work/reencoded_images/. Commit da64e00 for the push, follow-ups in this session.
  2026-05-13 · Session 13 · RevShift
• Apple Vision OCR pipeline + Jessar meeting brief (2026-05-12, overnight). Built and ran an OCR pipeline over 900 housewares-tier box-artwork PDFs that PyMuPDF couldn't fully read — rasterized/outlined text baked into artwork is invisible to text-layer extraction. Uses a tiny Swift CLI (scripts/ocr_image.swift → scripts/ocr_image) calling Apple Vision's VNRecognizeTextRequest with EN+FR language support — zero pip deps, on-device, no API tokens. Python orchestrator (scripts/ocr_box_artwork.py) renders each PDF page to 250 DPI PNG, batches through the Swift binary, filters print-production noise, saves to work/box_artwork/<sku>/<stem>.ocr.txt. Run completed in 57 min: 852/900 SKUs OCR'd, 940 PDFs, 31,883 lines recovered, zero errors. Pattern discovery on the full corpus surfaced missing patterns; added "Do not put in dishwasher or microwave", single-axis inline dims (w (width): 35 mm), master carton dims, standalone wattage. Re-extracted with --local --fresh. Housewares tier — the priority gap is now mostly closed: 7- JS Gourmet basic spec coverage (≥3 categories) jumped from 31% → 64% (+33pp); rich (≥5 categories) 9% → 24% (+15pp). 6- JS Maison basic 61% → 89%, rich 39% → 56%. Overall primary-tier: 1,667 SKUs, 1,489 (89%) with box-artwork features, 1,477 (89%) with any structured spec, 1,136 (68%) with usable spec table, 544 (33%) rich. made_in 36% → 59%, care (dishwasher/microwave/freezer) 14% → 21%, standalone width/height 6% → 30%+. Built scripts/render_jessar_meeting_brief.py — the Tuesday-call deliverable showing per-sub-brand coverage tables, per-spec-field × prefix heatmap, concrete ranked asks for Jessar (inserts/instructions, UPC list, certifications by family, spot-check on 900-row data request CSV, confirm 4 missing hero SKUs). Live at revshift-internal.vercel.app/jessar-meeting-brief. Companion work/jessar_data_request_2026-05-12.csv (900 rows of 6-/7-prefix housewares with we_have_X and NEED_X columns) attached to meeting prep. Retailer scraper stays parked — Linen Chest is JS-SPA, Amazon returns paid-placement noise; will rebuild as UPC-based once Jessar supplies UPC list.
  2026-05-12 · Session 12 · RevShift (overnight autonomous run)
• Box artwork v2 extraction — structured spec coverage 34% → 96% (2026-05-11). Rebuilt the box-artwork spec extractor after surveying raw text dumps and confirming v1's KEY:VALUE regex was missing the bulk of real callouts. v1 required Material: / Voltage: / Wattage: prefixes but real artwork uses standalone callouts (2W, 120 VOLTS, 5000K, AMBER) plus multi-line dimension blocks (Beauty Cover layout where width: and 163.5 and cm end up on three separate lines after PyMuPDF text extraction). v2 adds a categorised rules table with 30+ patterns binned into 24 categories — wattage, voltage, color_temp, lumens, life_hours, beam_angle, base_type, bulb_shape, dimmable, dimensions, width/height/depth/length/diameter, weight_net/gross, capacity, set_count, made_in, color, care, usage_environment — plus a multi-line dimension recovery pass keyed on label lines (width: / largeur: etc.) that scans 4 lines ahead for numeric + unit. Added --local mode that re-extracts from the existing work/box_artwork/<sku>/*.txt dumps without needing the SSD remounted, and --fresh to truncate before writing (CSV schema changed: new structured_specs JSON column + structured_count). New CSV: 2,697 SKUs, 11,862 total structured values, 96% of SKUs have structured spec data (up from 34% KV-only). Top category coverage: dimensions 74.5%, made_in 65.3%, voltage 26.6%, base_type 26.5%, wattage/lumens/life_hours/color_temp each ~18–19% (lighting tier), capacity 14.5%, color 12.1%. render_product_content_review.py updated to render a structured spec table per card (two-column key/value grid in JetBrains Mono, ordered by spec category, deduped against the raw KV fallback). 87% of primary SKU cards now show structured specs in the dashboard. Live at revshift-internal.vercel.app/jessar-content-review. Practical implication: the PDP spec table on the new site can now be populated from box artwork for ~1,448 of 1,667 primary SKUs without retailer scraping — retailer scraper becomes a cross-validation source, not the primary source.
  2026-05-11 · Session 11 · RevShift
• Sub-brand → prefix mapping RESOLVED by Jessar (2026-05-07). Jessar returned the completed mapping (work/sub_brand_mapping_RESOLVED_2026-05-07.csv): 154 groups across 2,560 SKUs, 100% assigned. Headline finding: at the data tier there are FOUR sub-brands, not ten. JS Gourmet (7-) = 1,264 SKUs; Jessar Lighting (5-) = 358; JS Maison (6-) = 236; Xtricity (combined 1- + 3- + 4-) = 702. The other catalogue names we'd been treating as sub-brands (Beauty Cover, Cook & Fresca, IBake, Limpus, Pop Fluff, Veraroma) do not have unique SKU prefixes — they are marketing identities / product lines that roll up into the four data-tier brands. Specifically: 6-CLEN-* groups (Limpus territory) all mapped to JS Maison, not Limpus. Implication for the homepage: home_sm_v6 currently frames the portfolio as "3 flagship + 7 specialty ribbon = 10 brands." This may need to be reframed as "4 house brands + product lines" to match data-tier reality — or kept as is if Sean decides the catalogue-tier identities (which appear on jessar.ca today) should remain as marketing labels. DECISION NEEDED from Sean. Sub-brand mapping was the single biggest unresolved data-foundation question; Phase 1 is now effectively unblocked.
  2026-05-07 · Session 10 · Jessar
• home_sm_v6 deployed — homepage canonical (sixth iteration, same session). Four refinements vs v5 per Sean's review: (1) Background flip — brand portfolio now anchored on canvas (white-ish) so cards POP like in v3; markline flipped to canvas-warm to maintain alternating rhythm down the page. (2) "In focus this season" (NEW, position 5) — editorial seasonal moment extrapolated from the v4 spotlight idea, but spread across multiple brand domains rather than tied to a single product. 3-up triptych: tabletop (JS Gourmet) / bedroom (Beauty Cover) / living room (Jessar Lighting). Each tile is image placeholder + italic caption. Editorial breath between the two navigational moments (find by brand → editorial → find by category). Single CTA "See the seasonal feature" routes to a curated landing page. (3) Catalogue eyebrow renamed — "Browse by category" (was "Browse the catalogue") to make the by-category nav explicit. Pairs cleanly with the brand portfolio's "Our brands" eyebrow. (4) Wordmark text removed from the nav — real Jessar logo image stands alone. Cleanup propagated across all six versions: v1, v2, v3 had text-only wordmark replaced with logo image (CSS updated); v4, v5 had the wordmark span removed. Live at jessar.revshiftmedia.com/home_sm_v6.
  2026-05-07 · Session 10 (6th iteration) · RevShift
• home_sm_v5 deployed — homepage (superseded by v6 same session). Three substantive changes vs v4 per Sean's review: (1) Flagship brand swap — Xtricity + Jessar Lighting moved to the specialty ribbon; JS Maison + Beauty Cover promoted to flagship. The homepage now leads with the consumer-housewares story (JS Gourmet kitchen / JS Maison home & furniture / Beauty Cover bedroom textiles). All three flagship cards now use real logos (Sean-supplied). (2) Featured product spotlight cut — Jessar isn't e-commerce. The website's job is to tell the brand story and route people into the catalogue, not to push individual products. (3) "Browse the catalogue" (NEW) at position 5 in the spotlight's place — five-tile category navigation moment. Top row: Kitchen + Home & Bath as 50/50 hero tiles (the consumer-everyday categories where the flagship brands live). Bottom row: Lighting + Electrical + Cleaning as third tiles. Each tile maps category → constituent brands so users route from category to brand seamlessly. Together with the brand portfolio above (find by brand), gives two clean entry paths into Jessar's offering. v1–v4 retained as comparison. Live at jessar.revshiftmedia.com/home_sm_v5.
  2026-05-07 · Session 10 (5th iteration) · RevShift
• home_sm_v4 deployed — homepage (superseded by v5 same session). Corrects v3 across five axes: (1) magazine framing cut entirely — no Issue 01, no Spring '26 (this is a homepage, not a publication); (2) "in commission" copy stripped everywhere — clean Shopify-style image placeholder wireframes (line-art icon + dimension label only, no apologetic text); (3) real Jessar logo + Cormorant wordmark in nav; (4) real sub-brand logos for 5 of 10: JS Gourmet, Xtricity (flagship); JS Maison, Beauty Cover, Limpus (specialty). Clean placeholders for the remaining 5 (Jessar Lighting, Cook & Fresca, IBake, Pop Fluff, Veraroma) until logos supplied; (5) hero rebuilt as full-bleed image placeholder + overlay text — drops in any 1920×1080 lifestyle photo and the hero is finished. Structure rebuilt: markline moved from position 6 to position 3 (brand thesis primes the page); featured spotlight restored at position 5 (anchors the abstract "ten brands × ~2,000 SKUs" claim in one tangible product); banners moved from position 4 to position 8 (action surface comes after the page has earned attention); promise strip cut (redundant with credibility numbers); standalone lifestyle moment cut (its purpose absorbed into the image-led hero). Net: 11 sections + footer, down from 13. Narrative arc: Hook → Thesis → Show → Prove → Trust → Mood → Convert → Validate → Decide. Caraway-inspired bones; clean Shopify wireframe placeholders communicate intent without apologizing. Live at jessar.revshiftmedia.com/home_sm_v4.
  2026-05-07 · Session 10 (4th iteration) · RevShift
• home_sm_v3 deployed — homepage (superseded by v4 same session). Corrects v2 across four axes per Sean's review: (1) width — bg full-bleed (canvas to screen edge), inner content max-width 1480px (v2's 1600px page-frame felt cramped); (2) brand count — TEN brands per PROJECT.md §6 architecture, not 3 (3 flagship cards: JS Gourmet sage / Xtricity dusty / Jessar Lighting terra, over a 7-card specialty ribbon: JS Maison · Beauty Cover · Limpus · Cook & Fresca · IBake · Pop Fluff · Veraroma); (3) hero — magazine-cover treatment, NOT PDP-shaped (full-bleed atmospheric bg with subtle texture grid, massive Cormorant headline "A Canadian story, in ten brands.", framing rules above and below carrying eyebrow + Issue tag and lede + CTA pair, source tag bottom-corner — totally different vocabulary from PDP); (4) banners — 3-up promotional row added per Month-2 SOW deliverable and client request (featured collection · accent-gradient dealer banner · dark-ink retailer banner). Plus: promise strip ribbon (mono-caps single-line credibility band below hero), 4-up reels grid (added Beauty Cover), approach narrative updated to ten-brand portfolio story. Bug fix: .footer class collision between page footer and brand-card footer resolved by renaming inner class to .brand-foot. Live at jessar.revshiftmedia.com/home_sm_v3.
  2026-05-07 · Session 10 (3rd iteration) · RevShift
• home_sm_v2 deployed — homepage (superseded by v3). Corrects v1's positioning error: Jessar is a Canadian distributor, not a curator. All curation language stripped. New hero: "A Canadian story, in three brands." New markline: "Three brands. Thirty years. Ten provinces. One Canadian distributor — family-run since 1996, supplying retailers from coast to coast with housewares worth carrying." Net additions vs v1: width-contained 1600px page-frame (no sprawl on wide screens) · lifestyle photography moment (full-bleed, 21:9) · highlighted reels grid (3-up 9:16, in commission) · dedicated brands navigation (3 large interactive cards as primary nav moment, not just descriptive triptych links). PDP design vocabulary applied throughout: 4:5 warm panels with mix-blend-mode multiply, source tags, hairline meta-row beams, oversized Cormorant marks. Live at jessar.revshiftmedia.com/home_sm_v2.
  2026-05-07 · Session 10 cont'd · RevShift
• home_sm_v1 deployed — Soft Modern homepage (superseded by v2). Editorial homepage in Direction B vocabulary, continuous with the PDP. Ten sections: nav · editorial hero ("Housewares, considered.") · 60-word manifesto · sub-brand triptych (JS Gourmet sage / Xtricity dusty / Jessar Lighting terra) · JS Gourmet spotlight · approach narrative ("Family-run. Canada-based. On a first-name basis…") · 4-cell hairline credibility row (3 brands · ~1,200 SKUs · 10 provinces · 30 yrs) · press placeholder (aspirational, dashed-border badge with pulsing dot) · two-paths CTA (Find a Retailer tagged "Decision Pending" per Sean) · footer. Strategy: brand elevation, not catalog port. Sub-brand portfolio treated as curated trio identity moments rather than 5–6 logo strip. Category tile section deliberately omitted (sub-brand triptych implies domain). Government sales section scrubbed (contamination from outdoors-gear reference). Live at jessar.revshiftmedia.com/home_sm_v1.
  2026-05-07 · Session 10 · RevShift
• Migration scope confirmed: 1,667 SKUs (primary tier only). The 3,211 figure in reports is the full DB universe — not the migration scope. Rule: master files (Items-Jessar + Electrical/Beauty Cover) are source of truth. In masters → migrate. On site but not in masters (366 "investigate") → skip, low priority, already validated. Catalog/Master Sku xlsm only (1,018 secondary) → ignore for now. Eco-fees + discontinued (160 excluded) → never. Product content review showing 1,667 is correct.
  2026-05-07 · Sean
• hero_pdp_sm_v2 approved — canonical Soft Modern hero render. Synthesizes Direction B palette/type with structural craft from v0_1_c (Direction C): 4:5 warm image panel (mix-blend-mode multiply), meta-row as hairline-bordered structural beam, 4-col hairline build grid with Cormorant Garamond values, oversized Cormorant sub-brand mark (clamp 54–84px), warm canvas-warm use-tiles, centered story (880px max-width), accent-light dealer CTA. "RECOVERY" annotation tag documents craft recovered from v0_1_c. Sean: "its excellent. Very happy. Ill be buying Hyper and we'll start diving in." Live at jessar.revshiftmedia.com/hero_pdp_sm_v2.
  2026-05-06 · Session 9 · Sean / RevShift
• hero_pdp_sm_v1 deployed — hero PDP in proper Soft Modern. Extends standard_pdp_sm_v1 with 10 sections (vs standard's 7): hero split (romance lede + highlights strip), story (image + narrative, italic headline), pillars (3-col product truths, no italic), made for (full-bleed dark, italic on emotional line), specs (dark panel), full sub-brand block (image + narrative, italic on tagline), dealer CTA (dark panel, left CTA + right B2B data table), curated cross-sell ("Complete the set"). Section annotation drawer maps all metafield sources and hero-vs-standard delta. Live at jessar.revshiftmedia.com/hero_pdp_sm_v1.
  2026-05-06 · RevShift
• standard_pdp_sm_v1 deployed — proper Soft Modern standard PDP. Built from Direction B design logic (not layout-adapted from prior explorations). 55/45 gallery-first hero grid (max-width 1340px), features shown openly (editorial clarity), accent (#C87A5E) limited to 5 places, meta strip understated at column bottom, gallery warm gradient. Italic rule applied throughout. Live at jessar.revshiftmedia.com/standard_pdp_sm_v1.
  2026-05-06 · RevShift
• Italic rule formalized. Cormorant italic reserved for: section-level storytelling labels, brand names in prose paragraphs, and emotional lines within narrative sections (made-for, story, sub-brand tagline). Never on: product titles, navigational headings, pillar attribute names, spec data, or cross-sell headings. Applies across all sm_v1 renders and all future Soft Modern work.
  2026-05-06 · Sean / RevShift
• Exploration renders produced (v3 + v0_1_b). standard_pdp_v3: merged v2 Direction B palette + v0.1a constrained grid, fixing screen-stretch (max-width: 1380px). standard_pdp_v0_1_b: v0.1a layout with Direction B colors replacing Fraunces/Warm Editorial treatment — confirmed doesn't work (palette mismatch; Cormorant Garamond doesn't do the tonal work Fraunces did). Conclusion: build Direction B from its own logic, not by translating from Direction A layouts.
  2026-05-06 · RevShift
• revshift-internal.vercel.app established as RevShift agency-wide cross-client hub. Each client (Jessar, Corbeil, Decorolala, etc.) gets their own section. Future internal tools across all projects are centralised here, keeping client-facing deliverables sites (e.g., jessar-deliverables.vercel.app) clean. Index at /Users/macbookpro/Downloads/RevShift Media/revshift-internal/.
  2026-05-05 · Sean
• Product content review added to deliverables hub. product_content_review.html (all 1,667 primary SKUs, searchable + filterable by category/status) live at jessar-deliverables.vercel.app/product_content_review. Internal RevShift tool — not Jessar-facing.
  2026-05-05 · RevShift
• Direction B renders v2 deployed. Three full Direction B locked renders live at jessar-deliverables.vercel.app: hero_pdp_v2 (JS Gourmet steak knives — 11 sections: story, build, made-for, dark specs, sub-brand block, dealer CTA), standard_pdp_v2 (Xtricity LED — streamlined, 3 accordions, dusty-blue dot), collection_page_v1 (Kitchen & Dining — brand-pill filters, sticky toolbar, 3-col grid with 2-col hero card). RevShift Notes drawers document Hyper section mappings. Nothing touches Shopify until Sean approves.
  2026-05-05 · RevShift
• Brand direction locked — Direction B (Soft Modern), default palette. Coral #C87A5E, off-white canvas #F8F5F0, Cormorant Garamond (display 300–500wt) + Inter (body 400/500) + Geist Mono (SKU/specs). Sub-brand dots: sage #7D9078 / dusty blue #7A8A9C / terracotta #B8755A. Italic-as-emphasis active. Unblocks hero PDP Liquid port, metaobject schema (sub_brand accent_color field), Phase 3 design system build. Walnut & Saffron ruled too close to Caraway; Stone & Plum too narrow for Jessar's product breadth. Decided by Jessar at 1pm meeting.
  2026-05-05 · Jessar
• Certifications strategy pivoted. Vision extraction pass deprioritised. 234 SKUs already have cert data from text extraction. Remaining 1,433 active SKUs compiled into work/cert_request.csv (High/Med/Low priority by category) for Jessar to fill in directly. Faster and more reliable than image logo extraction. If Jessar cannot provide promptly, vision pass will run overnight.
  2026-05-05 · RevShift
• Content generation complete. All 93 batches done. work/product_content.csv generated — 1,667 rows, 10 columns including certifications. 1,388 SKUs with AI-generated romance + description copy; 279 with extracted data only. Unblocks Phase 4 standard PDP population.
  2026-05-05 · RevShift
• Content extraction pipeline complete. All 1,557 box artwork PDFs processed; 110 SKUs had no PDF (stubs written). 17 fields extracted per SKU from PDF text, WooCommerce spec tables, and name-embedded specs. Weight merged from Items-Jessar for Sean.xlsx (poid-stoc column): 2.2% → 60.3% coverage (969 fills). Dimensions: 75.9% → 94.2% (305 fills). Key output: work/extracted_content.json (1,667 records) + work/content_gaps.csv. Data gap report and phase1 one-pager updated. Certifications: logos are image-embedded in PDF artwork — 14% from text extraction; 1,343 SKUs queued for vision pass (scripts/extract_certifications_vision.py). Care instructions: not a universal gap — correctly null for lighting/electrical; real gap is KITCH category (32.9% coverage within 608 kitchen SKUs). Romance copy: 845 EN SKUs to generate using title + category + specs + POC voice rules.
  2026-05-03 · RevShift
• SSD asset index complete. Seagate drive fully walked: 9,430 real files (5,740 box artwork PDFs, 3,679 studio photos from Quasimodo), covering 2,865 of 3,211 DB SKUs (89%). Primary-tier coverage: 99.8% (1,663 of 1,667 SKUs matched). Key gap confirmed: zero lifestyle photography exists for any hero candidate — needs commission or AI-generation decision before hero PDP build begins. 4 hero SKUs (7-99738, 7-98690, 7-99754, 7-99757) not in unified DB — Michael to confirm. 4 malformed primary SKUs (extension cords with product names in SKU field) flagged for manual cleanup. See ssd_asset_coverage.
  2026-05-03 · RevShift
• Jessar-facing project hub shipped. Parallel landing page at jessar-deliverables.vercel.app/jessar_hub — only client-facing items, grouped by purpose: (1) Read first (Phase 1 one-pager), (2) Decision needed (customer-edition brand direction), (3) Tools for your team (Product DB Viewer + Data Gap Report), (4) Tasks for specific people (hero candidates for Michael + intern brief), (5) Coming with launch (draft management guide). Internal artifacts deliberately omitted: open items dashboard, project state snapshot, internal PDP renders with annotation drawers, site architecture, brand guide v0.x with internal pills, investigation reports, WP audit. Sean's existing internal index now carries a top-card link to the Jessar hub so he can find it. Audience-split is more reliable than relying on the client to ignore "For Sean" tags on a shared hub.
  2026-04-30 · Sean / RevShift
• Standard PDP rendered alongside the 3 hero directions. Same product (7-99096) so the visual delta between hero and standard is unambiguous. Specs-forward, fast to scan, B2B-friendly — specs panel is the centerpiece (full B2B detail, dark inverse). No story / build / made-for sections (storytelling metafields skipped at render when jessar.is_hero is false). Sub-brand reduced to a single-line inline credit. Cross-sell uses Shopify auto-recs (vs hero's curated related_products). "Behind the page" drawer maps every section + flags hero-only metafields as "skipped" with explicit comparison notes vs the hero PDP. Direction-agnostic — B/C palette/typography swaps the same way the hero renders do. View →
  2026-04-30 · RevShift
• Site architecture locked. PROJECT.md §6 placeholders (sitemap, URL/taxonomy, filter/facet strategy) all filled in. Working reference shipped at jessar-deliverables.vercel.app/site_architecture — 27 page specs, four fields each (Purpose · Sections · Data sources · Done criteria). Sitemap: home → 10 sub-brand pages → 5 cross-cutting category collections → product PDPs (hero + standard templates) → static (partners, about, become-a-dealer, contact) → policies → utility (search, 404). URL pattern: /brands/{handle}, /collections/{handle}, /products/{handle}, FR mirror via /fr/. Filter strategy: native Hyper UI, metafield-driven, no third-party app. Out of scope (locked): store locator, dealer portal, retailer application form, customer accounts, cart, blog, refund/shipping policy pages.
  2026-04-29 · RevShift
• Hero PDP rendered in all 3 directions on real product 7-99096 (JS Gourmet steak knife set). Same EN+FR copy, same structure, same "Behind the page" metafield-map drawer; differs only in typography (Fraunces / Cormorant Garamond / Newsreader), palette (warm walnut / soft coral / acacia-sage), italic-emphasis treatment, and sub-brand expression (filled pill / 6px dot / bordered material card). Direction badge top-left for instant identification. A · B · C.
  2026-04-29 · RevShift
• Shopify code commenting + change-log convention locked in. Every theme edit gets an inline RevShift - YYYY-MM-DD - What it is comment + a per-file change log at /docs/shopify_code_changes/<flattened-path>.md showing original + final only (no intermediates). Saved to memory so it persists across sessions. Required for client handoff: Jessar can grep "RevShift" to find every customization, and read a single change-log file per theme file to understand what we did.
  2026-04-29 · Sean
• Client management how-to skeleton shipped at jessar-deliverables.vercel.app/management_guide_v0_1. Section-by-section reference mirroring the hero PDP exactly: what drives each section, where to edit in admin, bilingual flow, step-by-step. Skeleton level for now (screenshots get added once Hyper is live + metafields populated).
  2026-04-29 · RevShift
• Asset indexer scaffolding ready (scripts/index_ssd_assets.py). Walks $JESSAR_ASSETS, extracts SKUs from filenames + parent folders, joins against unified DB + hero list, outputs per-SKU coverage report (HTML + CSV + JSON). Handles "volume not mounted" gracefully. The moment SSD finishes copying: one-command run produces a coverage report flagging which heroes need lifestyle imagery commissioned.
  2026-04-29 · RevShift
• Hero candidates selection tool shipped to Vercel. jessar-deliverables.vercel.app/hero_candidates_for_michael — single self-contained HTML with all 50 unique product concepts as cards, grouped by category, live counter targeting 20. Output CSV drops at /work/hero_selections_michael.csv; the build pass flips Michael's picks to tier=sow_hero.
  2026-04-28 · RevShift
• SOW hero cap = 20. 50 candidates from Sean's curated list are now scope-tiered: Michael picks 20 → tier=sow_hero (full Caraway-style PDP). Remaining 30 stay tier=candidate and either bump to featured_extra if effort allows once brand direction locks, or defer to retainer. Curating 50 ≠ committing 50; protecting the 20-cap protects margin and timeline. Selection tool also hands Michael the call on which 20 — he knows which sell, we don't.
  2026-04-28 · Sean
• Master corrections pipeline live. work/master_corrections.csv is the canonical override list, applied last in build_unified_database.py so corrections always win regardless of source-master errors. First entry: 7-98211 flipped to primary/preserve (Sean confirmed misclassification in Electrical/Beauty Cover Inventory). Migration buckets shifted: preserve 1099→1100, skip 1098→1097, secondary 1385→1384. Pattern: Sean adds rows, re-run build, deliverables refresh.
  2026-04-28 · Sean / RevShift
• Cross-reference flag — 15 missing hero SKUs. 7-99738 family (S&P), 7-9975x families (Tea), 7-98690. Cross-check against new data when SSD/updated Items-Jessar arrives. If they're still absent from the canonical master after that pull, they shouldn't be hero candidates — flag back to Sean.
  2026-04-28 · Sean / RevShift
• Privacy Policy + ToS drafted on Shopify (template). Sean emailed Jessar for legal confirmation. Refund + Shipping policies skipped per Sean — no transactions, so they don't apply. Phase 5 net change: 17 → 19 items (added Privacy/ToS [done], FR translation pass, footer wiring).
  2026-04-28 · Sean
• Hero list received from Sean — 50 products / 88 SKUs. Parsed into work/hero_products_assigned.csv across 21 categories. 73 of 88 SKUs validated against unified DB; 15 missing (clustered around newest variant families — Salt & Pepper, Tea, Over-the-Door Hanger — likely from recent catalogs). Scope concern: 2.5-3.3x the original 15-20 hero PDP scope. Recommend tiering: top 15-20 get TRUE hero PDP treatment (story, lifestyle imagery, custom layout), other ~30 get "featured" treatment (premium copy + home placement, standard PDP). Awaits Sean's call.
  2026-04-28 · Sean / RevShift
• Metafield revision per Sean — 22 → 17 defs. Removed 8 (4 bilingual _fr dupes since Translate & Adapt handles them, 2 _en suffixes collapsed, why_stock as forward-leaning, original_wp_url as post-launch dead-weight). Added jessar.catalog_pdf (file_reference) for downloadable B2B sell sheet pattern.
  2026-04-28 · Sean / RevShift
• Phase 5 setup pass — 3 of 4 punch-list warnings cleared. Location updated to Jessar HQ (341 Allée du Golf, Saint-Eustache, QC J7R 0L5) · FR locale enabled + published (hreflang auto-emitted; dev plan only accepts base fr not fr-CA, sufficient for Quebec content) · 17 jessar.* metafield definitions on PRODUCT owner (revised set). Only remaining warning: Hyper demo→main flip (awaits Sean's purchase). Settings → General "About this store" address is read-only via API — values printed for manual UI paste.
  2026-04-28 · RevShift
• Afternoon batch (7 items shipped): test product live at admin/products/10306878865730 (10 metafields applied) · Hyper validation removed from public hub · customer-facing brand guide derivative built · public hub redesigned (left-bar accent removed per Sean, palette retained) · redirect map draft (2,172 rows) · brand voice skeleton · Phase 5 Shopify smoke test ran clean (3 warnings for punch list).
  2026-04-28 · RevShift
• Project State snapshot shipped — comprehensive where-we-stand / where-we’re-going deliverable. Phase grid, action lists for Sean + Jessar, brand-decision impact map, deliverables map by audience. View snapshot. Also: Shopify Access Setup section corrected to 4/4 Complete (was stale since OAuth exchange).
  2026-04-28 · RevShift
• Brand guide v0.3 shipped — italic-as-emphasis pattern (per Our Place reference) wired into every mock + spec, plus 5 alternative Caraway-spirited palettes (Walnut & Saffron, Bone & Brass, Linen & Forest, Stone & Plum, Almond & Saddle). Pick a direction and a palette — mix-and-match. View v0.3.
  2026-04-28 · RevShift
• Brand guide v0.2 shipped in response to Sean's feedback that v0.1 mocks looked alike. v0.2 loads real Google Fonts (Fraunces, Cormorant Garamond, Newsreader, Inter), uses all 9 palette colors per direction with semantic roles, and gives each direction its own distinct mock layout. View v0.2.
  2026-04-28 · RevShift
• Overnight: 3 new deliverables live on jessar-deliverables.vercel.app. (1) brand_guide_options_v0_1 — 3 direction-setting candidates honoring Sean's serif+sans soft-Caraway constraint. (2) investigate_skus_report — 366 WP-only SKUs pre-triaged. (3) wp_category_audit — 147 categories scored (34 abandoned, 11 weak, 102 healthy) for IA input.
  2026-04-28 · RevShift (overnight)
• Deliverables live on Vercel: jessar-deliverables.vercel.app. Built as replicable cross-client pattern (templates in polished-deliverables skill). Sean can spin up <client>-deliverables.vercel.app for any project in ~5 min.
  2026-04-27 · RevShift
• Hyper theme VALIDATED. All 3 criteria pass — 5 collection templates, 9 product templates, native filtering/swatches/sorting, 93 sections incl. brand-logos, collapsible-tabs, comparison-table. FR locale built in. See /work/hyper_theme_validation.html. Pebble ruled out.
  2026-04-27 · RevShift
• WP migration scope is just products + 301 redirects. No plugins, no blog, no menus (existing menu is messed up — we rebuild), no static pages exporting cleanly (we pull what's needed from products). Significantly tighter than first assumed.
  2026-04-27 · Sean
• AI image + video generation pipeline for hero PDPs and social — feed base product photos, generate lifestyle imagery / short videos. Logged in Phase 6 / retainer scope; design hero templates with AI-imagery placeholder slots so they drop in cleanly.
  2026-04-27 · Sean
• Phase 5 expanded with 3 critical migration items: 301 redirect map (WP slugs → Shopify URLs from Permalink column), Google Search Console setup, Google Analytics 4 setup. Confirm existing accounts first. Phase 5: 15 → 17 items.
  2026-04-27 · Sean
• Built product_database_viewer.html — single self-contained file, 3,211 SKUs interactive (sort/filter/search/expand). For Jessar's team to triage at speed. Color-coded tier + action badges, completeness bars, missing-fields chips. Light + dark mode.
  2026-04-27 · RevShift
• Bug fixes: dark-mode card on 1-pager (hardcoded cream → palette var); responsive breakpoint bumped to 1100px so vertical screens collapse to single-column before right column gets cramped. Skill CSS updated too.
  2026-04-27 · RevShift
• Additional WP exports requested: Pages, plugin list, menu screenshot. Products export already has the URL slugs needed for 301 redirects.
  2026-04-27 · Sean
• Tone POC approved by Jessar. Proceed with content gen in the existing voice. Q9 closed.
  2026-04-27 · Jessar
• FR copy: we do it ourselves. Items-Jessar desc-fr already has 2,659 SKUs covered — we enhance rather than translate from scratch. Cherry-pick human review only. Shopify Translate & Adapt for fallback. Q10 closed.
  2026-04-27 · Jessar
• MSRP populated from Items-Jessar col AO ("Selling Price"). Stays gitignored; pushed to Shopify as msrp metafield behind dealer login. Q6 fully resolved.
  2026-04-27 · Sean / Jessar
• Partner logos: Sean procures from public sources (Costco, Linen Chest, Staples). Q17 fully resolved.
  2026-04-27 · Sean
• Jessar intern available for data tasks. First brief: sub-brand mapping template.
  2026-04-27 · Jessar
• Secondary product tier added to unified DB. Long-tail products from Master Sku / catalogs get priority_tier=secondary + parallel CSV. Ready to add later without pipeline rebuild.
  2026-04-27 · Sean
• Data foundation summary refresh after SSD pull — current 1-pager is accurate to what's local; expect updates after asset bulk-copy.
  2026-04-27 · Sean
• Dark mode + theme toggle on all Jessar HTML deliverables (progress.html, data_gap_report.html, phase1_data_state_one_pager.html). Respects OS preference; ☾/☀ toggle top-right persists to localStorage.
  2026-04-27 · RevShift
• Unified product database is the Phase 1 deliverable. 3,211 unique SKUs, 5 migration buckets (preserve 1,099 · add 567 · cleanup 81 · investigate 366 · skip 1,098). Completeness scored. Pricing kept out of repo.
  2026-04-27 · RevShift
• Shopify OAuth complete. shpca_* token stored in .env. Smoke-tested: shop = Jessar Industries Inc., CAD, EN primary, dev plan, 0 products. Granted scope = all admin (~70 scopes — Partner custom apps grant full access by default).
  2026-04-27 · RevShift
• Hyper + Pebble both pre-installed as demo themes on the dev store (current main = Horizon). Hyper validation can proceed immediately via Shopify CLI pull of theme code without changing the live role.
  2026-04-27 · RevShift
• WP vs masters reconciliation: 1,226 live + active (preserve), 110 live but discontinued (must clean up), 805 WP-only (status unclear), 440 active but missing from current site (must add).
  2026-04-27 · RevShift
• Catalog PDFs are searchable text — no OCR needed. PyMuPDF extracted 1,767 SKU rows across 11 PDFs, 451 pages. Full text saved per page for content gen reuse.
  2026-04-27 · RevShift
• Orange exclusion lives in Items-Jessar's On Hand sheet (not the SKU master). 119 SKUs flagged discontinued.
  2026-04-27 · RevShift
• WP SKUs are in the Name field, not the SKU column. Built regex extractor.
  2026-04-27 · RevShift
• WC CSV export is sufficient — REST API not needed for now. Sean uncertain about full API access; 2,196-product CSV covers SKUs, names, descriptions, categories, images, stock, published.
  2026-04-27 · RevShift
• Brands are sub-brands of Jessar. Top-level identity is Jessar; 10 sub-brands (JS Gourmet, JS Maison, Beauty Cover, Cook & Fresca, IBake, Limpus, Pop Fluff, Veraroma, Jessar Lighting, Xtricity) live inside.
  2026-04-27 · Jessar
• Orange highlight in Items-Jessar = ignore (discontinued + eco-fees). Read cell fill colors programmatically.
  2026-04-27 · Jessar
• Source-of-truth confirmed: Items-Jessar + Electrical/Beauty Cover Inventory = primary. Master Sku xlsm = late-stage gap-filling reference.
  2026-04-27 · Jessar
• No store locator. Replaced with "Partners" section showing major retailers (Costco, etc.) as logos.
  2026-04-27 · Jessar
• Dealer portal out of scope — link to existing system. Retailer application out of scope (or simple contact form).
  2026-04-27 · Jessar
• Costco items live on the site (not Costco-exclusive). Catalog supersession: newer wins only if same SKU; otherwise add regardless of year.
  2026-04-27 · Jessar
• No MSRP yet — add empty msrp metafield as placeholder.
  2026-04-27 · Jessar
• Brand guide: we design 2–3 options. Jessar has no existing guide. New explicit Phase 3 deliverable.
  2026-04-27 · Jessar
• Pricing is NOT a gap — Items-Jessar already has prix-list, cout-std, prix-cout, prix. Lives behind dealer login. Local only, never in repo.
  2026-04-27 · RevShift
• 108 eco-fee / non-product SKUs excluded from migration (accounting line items, not products).
  2026-04-27 · RevShift (pending Jessar confirm)
• Items-Jessar confirmed as canonical SKU master (provisional). ~2,560 real SKUs across 6 prefixes, fully bilingual, pricing present.
  2026-04-27 · RevShift
• External SSD primary; Tailscale as fallback. Windows App (RDP) on Mac is not programmatically accessible — SSD is the right call for bulk processing.
  2026-04-27 · Sean
• Lockstep enforcement via hook + CLAUDE.md. PostToolUse hook injects reminder when PROJECT.md is edited; CLAUDE.md auto-loads at session start. Activate with /hooks once.
  2026-04-27 · Sean
• Veraroma identified as JS Gourmet sub-line (logo: Veraroma by JS Gourmet). Same pattern likely for other sub-brands — confirm with Jessar.
  2026-04-27 · Sean
• Custom Shopify Partner app created against jessar-industries.myshopify.com dev store. OAuth flow to capture access token; secret stored in .env (gitignored).
  2026-04-27 · Sean
• Repo at folder root with smart .gitignore for raw client data.
  2026-04-27 · Sean
• 6-phase engagement with sign-off gates and renders before commits.
  2026-04-27 · Sean
• Tooling stack (provisional): Hyper theme + Shopify CLI + Python scripts. Migration tool (Matrixify vs. API) TBD.
  2026-04-27 · Sean
• Items-Jessar for Sean.xlsx as primary data source (provisional).
  2026-04-27 · Sean
• PROJECT.md + progress.html in lockstep as canonical state.
  2026-04-27 · Sean
• 1 TB external SSD for bulk box-artwork + image access.
  2026-04-27 · Sean
• Confidential files NEVER in repo: SOW, raw xlsm, future pricing, .env.
  2026-04-27 · Sean

Quality Bar (Always-On)

• 1. Brand elevation is the goal. Migration is the vehicle.
• 2. Render before commit. No design or IA is "done" until visualized + approved.
• 3. Phase gates are real. Don't start N+1 until N is signed off.
• 4. Data first, structure second.
• 5. Dual-audience always — consumer + B2B on every page.
• 6. Bilingual EN/FR from day one.
• 7. Truth over speed. Mark gaps, don't invent data.
• 8. Reference Caraway/Our Place + Linen Chest/Sur La Table — synthesize, don't copy.
• 9. PROJECT.md is the single source of truth.
• 10. Append-only history.