Data maintenance pipeline

data-maintenance runs CookDex’s cleanup and audit stages in a fixed order:

dedup -> junk -> names -> parse -> foods -> units -> categorize -> yield -> quality -> audit

Each stage and option is listed under data-maintenance in the task reference. Run the full pipeline for broad maintenance, or select individual stages when you only need a targeted operation.

1. Open CookDex at /cookdex.
2. Go to Tasks and select Data Maintenance Pipeline.
3. Leave Preview Run enabled.
4. Choose specific stages only if you want a targeted run.
5. Run the task and review the log.
6. For live cleanup stages, switch to Run Live, enable Apply Cleanup Writes, and confirm the owner policy unlock.

Stage	Purpose
dedup	Finds recipes imported from the same source URL.
junk	Detects non-recipe pages such as listicles, how-to posts, digests, placeholders, and bad scrapes.
names	Normalizes names derived from URL slugs.
parse	Parses raw ingredient lines into structured fields.
foods, units	Merges duplicate foods and units. Unit merges use each unit’s abbreviation, plural and aliases in Mealie.
categorize	Applies rule-based organization and optional AI categorization.
yield	Normalizes yield and servings fields.
quality	Scores recipe completeness.
audit	Audits taxonomy usage and duplicates.

The labels, tools, taxonomy and cookbooks stages were retired with the Taxonomy Editor. Saved schedules that still list them skip them and say so in the log. Edit those in Organize, or import a taxonomy file there.

Create an interval schedule in the Web UI for recurring maintenance. Use a once schedule for a one-time future run.

For routine monitoring, schedule data-maintenance with Preview Run enabled or schedule health-check for a read-only report.