instacart-orders

name	instacart-orders
description	Read Instacart order history (canonical item IDs, retailer, dates) from the printing-press CLI's local SQLite DB. Trigger ad-hoc re-imports when the nightly history scrape is stale or missed an order.
version	2.0.0

Instacart Order History

Replaces the old browser-snapshot scraper (retired 2026-05-13) with a deterministic CLI-driven pipeline. Order history now lives in the instacart-pp-cli SQLite DB at ~/Library/Application Support/instacart/instacart.db and is populated nightly by ~/.config/spratt/infrastructure/instacart/history-scrape.py under launchd job com.spratt.instacart-history-scrape.

When to use this skill

• "What was in the last Instacart order?" — query the CLI DB.
• "When did I last buy X on Instacart?" — history search X --json.
• "How often do I buy X?" — ~/.config/spratt/infrastructure/instacart/cadence.py --due-only.
• A nightly scrape missed a specific order and you want to import it now.

For building a cart (add items, place order) use the instacart-api skill instead.

Read-only queries

# Most-recent orders
instacart-pp-cli history list --json --limit 20

# All items at a single retailer, by purchase frequency
instacart-pp-cli history list --store costco --json

# Free-text search across purchased items
instacart-pp-cli history search "honey roasted" --json

# Order + item totals per retailer
instacart-pp-cli history stats --json

The DB exposes canonical Instacart IDs:

• order_id — 17–18 digit numeric (the Apollo cache id, not the URL id).
• item_id — format items_<retailerLocationId>-<legacyProductId> (use this with instacart-pp-cli add --item-id ... to bypass autosuggest). Caveat (2026-05-16): the nightly scraper currently persists RetailerID-prefixed IDs (e.g. items_42-... for QFC RetailerID=42) instead of the LocationID-prefixed form Instacart's add-to-cart API now requires (e.g. items_3092-... for the QFC store branch serving the saved address). Dry-runs still pass; real mutations get rejected as notFoundBasketProduct. The Wed/Sat cart-build.py self-heals by re-searching by item name on rejection and retrying with the fresh LocationID-prefixed ID, but ad-hoc instacart-pp-cli add --item-id calls do NOT have that fallback — search first or expect rejection on stored historical IDs.
• product_id — numeric, useful for cross-referencing.

Ad-hoc re-import (rare)

If you know the nightly scrape missed a specific order — for example because the order is older than the "Load more" boundary at scrape time — trigger a manual import.

# Sweep everything visible on the order-history page (~3 hours wall
# clock for 200+ orders). Idempotent — already-imported orders are
# skipped by `history import`.
python3 ~/.config/spratt/infrastructure/instacart/history-scrape.py --backfill

# Or process only N new (un-tracked) orders. Useful for catch-up after
# the launchd job has been off for a few days.
python3 ~/.config/spratt/infrastructure/instacart/history-scrape.py --max-orders 30

The script writes status to ~/.config/spratt/infrastructure/launchd-status/instacart-history-scrape.json and emits an outbox alert on uncaught failure.

Limits

• Some legacy orders return cache_key_missing during extraction (the URL order_id doesn't line up with the Apollo cache key on a small number of older orders). Those rows are skipped, not retried, and the script records them under errors in the status JSON.
• The CLI's own history sync command is documented as broken upstream ("Instacart has no clean GraphQL op for order history"). Don't use it. Use history-scrape.py above — it drives the CLI's own bundled docs/extract-one.js (per-order Apollo cache reader) and pipes results back through instacart-pp-cli history import -.
• Auth: relies on the Chrome cookies that instacart-pp-cli auth login reads. If auth status --json returns logged_in:false, see the instacart-api skill for the recovery flow.

What lives where now

Old (pre-2026-05-13)	New
orders.sqlite.orders rows with source='instacart'	instacart.db orders + order_items tables
Email-scan fabricated email-XXX order_ids when receipts had no number	Canonical numeric order_id from Apollo
Nightly OpenClaw cron Instacart Order Scraper (agentTurn, Flash)	launchd com.spratt.instacart-history-scrape (deterministic Python)
Purchase cadence: purchase-cadence.py against orders.sqlite for both Amazon and Instacart	purchase-cadence.py for Amazon only; instacart/cadence.py for Instacart
Cart-build via browser autosuggest (reorder-cart-build.py)	cart-build.py via CLI add --item-id (canonical IDs)

The historic orders.sqlite rows stay where they are; nothing actively queries them anymore. The only orders.sqlite table still in use is reorder_notifications, which reorder-nudge.py reads + writes to dedup Wed/Sat nudges across runs.