descrybe/scripts/seed/README.txt

A1 Slovenija tenant seed (Postgres COPY archive, gzipped)

File: a1-demo-data.sql.gz
Company id: 604f23a8-b66e-4b21-8b45-0d72b68f4790
Legacy MySQL id: 97e1a309-3d23-4aa2-b518-8e8d7afdfec7

Source of truth for a clean local DB (2026-08-08 reseed):
  MySQL dump → C:\Users\Green Eclipse\Downloads\descrybe_new (1).sql
  Loaded into Laragon MySQL DB `descrybe_new`, then A1-only migrator
  (`-company 97e1a309-3d23-4aa2-b518-8e8d7afdfec7`) + seed-demo + user cleanup.
  See docs/migrate-from-descrybe-new.md and docs/demo-user.md.

Expected A1 snapshot (dump-faithful, no local QA/Iso pollution):
  companies=1 (A1 Slovenija only, plus internal __platform_settings__)
  users=3 (demo@descrybe.local, demo@descrybe.test, a1-primary@descrybe.local)
  memberships=3
  categories=119 (108 with AI prompts via overlay)
  attributes=303
  category_attributes=307
  input_feeds=12 / feed_mappings=11 / export_feeds=2
  raw_products=~25355 / processed_products=0 (demo seed clears processed)
  processing_jobs=0 / processing_job_products=0
  plan=A1 wallet total=2500 used=163 remaining=2337
  mapped coverage (typical): description ~98%, feed attrs ~99%, category ~17% (4330 dump-assigned)

Category assignment (dump -> PG):
  MySQL processed_products.category holds category unique_id (e.g. 46, 120).
  Feed XML / feed_mappings do NOT include a category field - assignment was
  historical processing output. Archive restores raw mapped_data (description,
  specs/attrs, category when present). seed-a1 then refreshes
  mapped_data.category from a MySQL dump when available.

  Dump path (first match wins):
    1. -mysql-dump flag / admin body mysql_dump
    2. SEED_A1_MYSQL_DUMP (API process env on Git-Syncer)
    3. Auto-detect: scripts/seed/descrybe_new.sql (deploy root),
       ~/Downloads/descrybe_new (1).sql or descrybe_new.sql

  Git-Syncer / production: copy the dump onto the API host, e.g.
    <deploy-root>/scripts/seed/descrybe_new.sql
  or set SEED_A1_MYSQL_DUMP=/absolute/path/descrybe_new.sql in the API env.
  Then use Admin → Companies → Sync A1 (no CLI DATABASE_URL needed).

  One-off without wipe (CLI; loads monorepo-root .env when unset):
  go run ./cmd/seed-a1 -mode backfill-categories

  Admin Sync A1 = dump category backfill (when dump present) + Fix hygiene
  (prompts, weak hashes, bidirectional category backfill, attr sanitize).
  It does NOT wipe/reimport the ~25k catalog.

  Note: MySQL processed description/attributes are usually NULL; feed-origin
  original description + specs live on raw_products.mapped_data in the archive.
  A1 demo seed keeps processed=0 / jobs=0.

Who sees what:
  demo@descrybe.local � platform admin + membership on Platform Demo only
    (switch/impersonate to a1-primary to act on A1; do not use demo API key for A1)
  a1-primary@descrybe.local � A1 member only (same A1 catalog; no other tenants)

Category prompts overlay (committed WP dump + JSON fallback):
  wp_product_categories.sql   — primary Sync A1 / repair source of truth
  a1-category-prompts.json    — JSON fallback when WP SQL is absent
Applied automatically after every `npm run seed:a1` reimport (match by normalized
category name; rewrites {""OPIS IZDELKA""} -> {{description}},
{""STARO IME IZDELKA""} -> {{name}}). Skip with -skip-category-prompts, or re-apply
without wipe:

  go run ./cmd/seed-a1 -mode apply-category-prompts -file ../../scripts/seed/a1-demo-data.sql.gz
  go run ./cmd/repair-category-prompts -apply
  # uses scripts/seed/wp_product_categories.sql by default (SEED_A1_WP_CATEGORIES overrides)

Normal use (loads DATABASE_URL from repo-root .env if unset):
  npm run seed:a1          # wipe A1-only rows, import archive, prompts + mapped category backfill
  npm run reimport:a1      # alias of seed:a1

Requires migrations already applied. Does not wipe other companies.
Does not run export — it only imports scripts/seed/a1-demo-data.sql.gz.

A1 demo seed skips processed_products / processing_jobs (processed=0, jobs=0).
Use -mode recover-jobs only when restoring legacy job history from a MySQL dump.
feed_sync_jobs stay ephemeral and are not restored; raw_products.sync_job_id is
exported as NULL for that reason. Archive is Postgres COPY text inside gzip, not
MySQL dump SQL.

Full clean rebuild from MySQL dump (when archive is wrong / polluted):
  1. mysql: DROP/CREATE descrybe_new; stream-import the Adminer dump
  2. docker: DROP/CREATE postgres descrybe; goose up
  3. migrator -company <A1 legacy id> -resume -ensure-demo -skip-post-import
  4. migrator -domains jobs -resume
  5. go run ./cmd/seed-demo
  6. keep only demo + a1-primary memberships; delete orphan users / Iso test cos
  7. go run ./cmd/seed-a1 -mode apply-category-prompts
  8. npm run seed:a1:export   # refresh a1-demo-data.sql.gz then commit

Export copies EVERY A1 processing_jobs / processing_job_products row (no LIMIT).
If the archive only has a couple of jobs, live Postgres was missing history —
recover from the MySQL dump first, then re-export:

  set SEED_A1_MYSQL_DUMP=D:\path\descrybe_new.sql
  go run ./cmd/seed-a1 -mode recover-jobs
  npm run seed:a1:export

Maintainer only — regenerate the snapshot when live A1 data must be re-captured:
  npm run seed:a1:export   # snapshot live A1 → a1-demo-data.sql.gz (then commit)
Not part of the day-to-day reset workflow.

Support Knowledge articles (platform FAQ /admin/support/knowledge):
  support-kb-articles.json       � product/user corpus (Agent 6)
  support-kb-articles-tech.json  � technical/ops/admin corpus (Agent 7; tech-* slugs)
Markdown sources for tech bodies (optional edit then regenerate JSON):
  apps/api/cmd/seed-support-kb/content/*.md
Seed (from apps/api, DATABASE_URL set):
  go run ./cmd/seed-support-kb
  go run ./cmd/seed-support-kb -file ../../scripts/seed/support-kb-articles-tech.json
Idempotent ON CONFLICT (slug). Do not reuse product slugs for tech articles.