- Drop the empty batch_alter_table block and the meaningless
set_config call from migration 0015. Temp tables still persist
for the migration's session (Alembic's transactional_ddl).
- New backend/scripts/persist_aisle_backup.sql creates
public.ingredient_aisle_backup_0015 and
public.grocery_item_aisle_backup_0015 permanent tables for
operators who want a recoverable record beyond the migration.
- Update Review/sprint2-verification.md, Review/ui-nielsen-audit.md
and fix-ui-audit.md with the correct container-based deploy
steps: docker compose exec db psql -U mealplanner -d mealplanner
-f /dev/stdin < ...sql. Host psql is not available on the
deployment host; the db runs inside the container.
- Dashboard MealCard: title truncate -> line-clamp-2, image shrinks to
40x40 on <md to give the title room (B6).
- MealDetail: hero reworked to normal flow with stronger gradient;
description runs through new cleanDescription() helper that strips
14 spoonacular SEO patterns and trims to the last full sentence.
Raw description moved to a 'Notes from source' disclosure (B7).
- Pantry: free-text aisle/unit replaced with <Select> populated from
the new PANTRY_AISLES canonical enum; ingredient name field marked
required. New PANTRY_AISLES export + PantryAisle type in types (B8).
- backend: alembic 0015_normalize_pantry_aisles maps free-text
ingredient.aisle and grocery_item.aisle to canonical labels in a
single transaction; downgrade raises (restore from snapshot).
backend/scripts/dry_run_aisle_migration.sql is the read-only
preview helper.
- ShoppingList: human-readable AISLE_LABEL map replaces raw snake_case
aisle keys; 3-col stat grid with compact mobile sizing (B9 + S3.3).
- Pantry table: role/aria-label region and a right-edge white
gradient hint at mobile horizontal overflow (B10).
- Recipes: pending/applied filter split, Apply and Reset buttons,
active-count chip on the Filters button, role=region + aria-label
on the panel (B11).
- Review/sprint2-verification.md and fix-ui-audit.md updated.
Build: npm run build (tsc + vite) green. tsc emits 0 errors.
Co-located audit + plan docs kept in sync: Review/ui-nielsen-audit.md
gains a Sprint 2 status block; fix-ui-audit.md has implementation
notes for each Sprint 2 task.
- Migration 0012 adds score (float) and components (jsonb) to meal_plan_item
- generate.py: populates score and components at create time
- schemas/MealPlanItemResponse: include score + components fields
- GET /api/meal-plans/{id}: returns persisted values instead of zeros
Matcher improvements (matcher.py):
- Plural normalization: 'tortillas'→'tortilla', 'thighs'→'thigh' so
subset recall check works without stemmer
- Precision floor lowered 0.45→0.30: allows 'Bacon'→'Wright Brand Bacon'
(1/3=0.33) while exclusion words still block category contaminants
- _EXCLUSION_WORDS now normalized through same singularizer for consistency
LLM second-pass (llm_matcher.py):
- run_llm_match_job(): for each still-unmatched ingredient, collects top-12
candidates from grocery catalog ranked by fuzzy×precision (same metric as
AUTO matcher), then asks Ollama to pick the best match
- Candidate scoring: combined = (partial_token_sort_ratio/100) × precision
ensures "McCormick Black Pepper" outranks "Dr Pepper" for 'Black Pepper'
- Stores picks as source='auto_llm' (confidence=0.750)
- Ollama Cloud endpoint: https://ollama.com/v1, model: kimi-k2.6:cloud
Migration 0010: adds 'auto_llm' to ingredient_match_source_enum
Config: OLLAMA_BASE_URL / OLLAMA_API_KEY / OLLAMA_MODEL settings
Docker-compose: wires all three Ollama + Spoonacular env vars to backend/scheduler
Scraper service: calls run_llm_match_job after run_match_job on every scrape
Results: AUTO matcher went from 36→25 unmatched (plural normalization fix),
LLM added 3 more (Black Pepper, Zucchini, Chicken Thighs).
Remaining 22 are genuine Lucky CA catalog gaps (standalone olive oil,
dried spices, etc. not in Swiftly weekly ad).
Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
admin
