---

name: django-index-design description: Design safe Django and PostgreSQL indexes from measured query plans, including B-tree, composite, covering, partial, expression, GIN/JSONB, BRIN, and concurrent migration patterns. Use when a Django query plan suggests missing or mismatched indexes, a production table needs an index without write downtime, or existing indexes need review. license: MIT compatibility: Codex, Claude Code, and other Agent Skills-compatible clients. metadata: version: "0.1.0" displayName: Django Index Design category: Django tags: django,database,indexes,postgres,performance

Django Index Design

Use this skill to add or change indexes only after profiling shows a specific query can benefit. Indexes speed selected reads and constraints, but they add write cost, migration risk, and maintenance overhead.

Workflow

1. Start from a query plan.

   • Capture the SQL, filters, joins, ordering, selected columns, table sizes, and current indexes.
   • Confirm that the query is important enough to optimize.

2. Choose the index shape.

   • Equality filters first, then range/order fields when useful.
   • Match ORDER BY for LIMIT queries when possible.
   • Use partial indexes for highly queried subsets.
   • Use covering indexes only when selected columns are stable and index-only scans are plausible.
   • Use GIN for JSONB containment, arrays, and full-text patterns; do not expect a B-tree to solve those.

3. Express the index in Django models or migrations.

   • Prefer Meta.indexes for normal model-owned indexes.
   • Prefer AddIndexConcurrently and RemoveIndexConcurrently for live PostgreSQL tables.
   • Use raw SQL only when Django cannot express the index.

4. Validate on production-like data.

   • Re-run QuerySet.explain() or EXPLAIN ANALYZE.
   • Verify the planner uses the index for the target query.
   • Check write-path impact when the indexed table is hot.

See index-patterns.md for index examples and migration templates.

Production Rules

• For PostgreSQL live tables, use concurrent index operations and set the migration atomic = False.
• Keep names short, stable, and explicit.
• Do not add duplicate indexes that are already covered by a unique constraint or a left-prefix composite index.
• Remove unused indexes only after checking production usage and deployment rollback needs.

Verification

Finish with before/after plan evidence, the exact migration operation, and the operational risk assessment for locks, writes, and rollback.