mcpField
D
django-query-profiling

django-query-profiling

github.com/lvtd-llcDevelopment162 installsUpdated 7w ago
Repository

About django-query-profiling

A workflow skill for Django developers tracing reported slowness to a specific queryset, SQL statement, and reproducible scenario. It combines production evidence, code mapping, and local reproduction to guide profiling toward the right fix.

Key features

  1. 01Capture the slow user-visible path
  2. 02Collect traces and query evidence
  3. 03Map SQL to Django code
  4. 04Reproduce with representative data
  5. 05Route findings to a fix

Use cases

  1. 01Investigate a slow API action
  2. 02Find SQL behind a slow page
  3. 03Reproduce a background task slowdown

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/lvtd-llc/skills --skill django-query-profiling

After installing, type/django-query-profilingin Claude Code to confirm it loaded.

Bundled filesreferences/tooling-and-reproduction.md

Details

From the project README

Django Query Profiling

Use this skill to turn a vague slow-Django report into a specific queryset, SQL statement, and reproducible scenario.

Workflow

  1. Start from the user-visible slow path.

    • Record the URL, API action, background task, management command, or report.
    • Capture request parameters, user or tenant shape, pagination state, and data volume.

  2. Collect production or staging evidence.

    • Prefer APM transaction traces when available.
    • Use database slow-query logs for SQL that is slow independent of Python.
    • Use Django Debug Toolbar for server-rendered pages in local development.
    • For APIs or jobs, add targeted query logging around the suspicious block.

  3. Map SQL back to Django code.

    • Search model table names and column names.
    • Inspect view get_queryset(), serializer fields, template loops, managers, model properties, and signal handlers.
    • Check whether related-object access happens after the initial queryset was evaluated.

  4. Reproduce locally or in a safe shell.

    • Use production-like row counts when possible.
    • Disable unrelated instrumentation and debug-only middleware when measuring.
    • Keep a repeatable script, test, or shell snippet that exercises the slow path.

  5. Route the fix.

    • Query count issue: use django-orm-query-optimization.
    • One slow SQL statement: use django-query-plan-reading.
    • Large loops or writes: use django-queryset-batch-processing.

See tooling-and-reproduction.md for concrete profiling snippets and pitfalls.

Good Evidence

  • Query count before/after for the specific path.
  • The slow SQL or normalized SQL fingerprint.
  • Stack or code pointer that explains where the SQL originates.
  • Timing from the same environment and representative data shape.

Verification

Do not finish profiling with only a hunch. Finish with a named queryset, code path, and command or request that another agent can re-run.

Comments

Markdown supported