silk-debug
Skill119 starsChanged 3 months agoMITView on GitHub
Analyze Django Silk profiling data to debug slow requests, detect N+1 queries, and optimize database performance. Use when analyzing request IDs, investigating slow endpoints, or optimizing query performance.
In this repository5 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- loki-logsSkill
- update-changelogSkill
- updating-depsSkill
- whats-newSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 21
When given a Silk request ID (UUID), use --full for comprehensive analysis:
- Distinctive: line 68
When you see duplicate queries like:
- Very specific: line 122
When the same for_user() query appears multiple times:
- Distinctive: line 123
Create a method that accepts already-checked objects
- Distinctive: line 128
When creating multiple objects:
- Very specific: line 129
Use bulk_create() instead of individual .save() calls
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Silk Debug Tool
A CLI tool for analyzing Django Silk profiling data to debug slow requests, detect N+1 queries, and optimize database performance.
Tool Location
.venv/bin/python scripts/silk_debug.pyQuick Reference
Analyze a Specific Request
When given a Silk request ID (UUID), use --full for comprehensive analysis:
.venv/bin/python scripts/silk_debug.py <request_id> --fullThis shows:
- Request info (path, method, status, time, query count)
- Duplicate/similar queries (N+1 detection)
- Slow queries (>5ms by default)
- Queries grouped by table
- Query execution timeline
- Python cProfile data (if enabled)
List and Filter Requests
# List recent requests
.venv/bin/python scripts/silk_debug.py --list
# Sort by different criteria
.venv/bin/python scripts/silk_debug.py --list --sort queries # Most queries
.venv/bin/python scripts/silk_debug.py --list --sort duration # Slowest total time
.venv/bin/python scripts/silk_debug.py --list --sort db_time # Most DB time
# Filter requests
.venv/bin/python scripts/silk_debug.py --list --path /api/events --min-queries 20
.venv/bin/python scripts/silk_debug.py --list --method POST --min-time 100Aggregate Analysis
# Overall statistics
.venv/bin/python scripts/silk_debug.py --stats
# Endpoint summary (grouped by path pattern, shows P95)
.venv/bin/python scripts/silk_debug.py --endpoints
# Find slow endpoints
.venv/bin/python scripts/silk_debug.py --slow-endpoints --slow-endpoint-threshold 100Interpreting Results
N+1 Query Detection
When you see duplicate queries like:
🔴 15x similar queries:
SELECT "events_ticket"."id" FROM "events_ticket" WHERE "events_ticket"."event_id" = '<UUID>'This indicates an N+1 problem. Fix with:
select_related()for ForeignKey fieldsprefetch_related()for reverse relations or M2M fields
Slow Queries
Common causes of slow queries:
- COUNT on complex DISTINCT: Pagination wrapping complex visibility subqueries
- Fix: Materialize IDs in Python first, then filter with simple
INclause
- Fix: Materialize IDs in Python first, then filter with simple
- Missing indexes: Full table scans
- Fix: Add database indexes on filtered/joined columns
- Complex JOINs: Multiple related tables
- Fix: Optimize query structure or denormalize if appropriate
Timeline Analysis
The timeline shows query execution order with visual bars:
1. + 0.0ms [ 2.5ms] █ "accounts_reveluser"
7. + 60.2ms [ 2.8ms] █ "__count"Look for:
- Large gaps between queries (indicates Python processing time)
- Queries that could run in parallel but are sequential
- Expensive queries that block subsequent operations
Common Optimization Patterns
Expensive COUNT with DISTINCT
When you see:
SELECT COUNT(*) FROM (SELECT DISTINCT ... complex subquery ...)Fix by materializing IDs:
# Before (slow COUNT)
qs = Event.objects.for_user(user).filter(...).distinct()
# After (fast COUNT)
event_ids = list(Event.objects.for_user(user).values_list("id", flat=True).distinct())
qs = Event.objects.full().filter(id__in=event_ids)Redundant Visibility Checks
When the same for_user() query appears multiple times:
- Create a method that accepts already-checked objects
- Cache visibility results within the request
Batch Operations
When creating multiple objects:
- Use
bulk_create()instead of individual.save()calls - Fetch shared data (settings, related objects) once before the loop
- Send notifications in batches, not per-item
CLI Options Reference
Single Request Analysis
--full, -f: Run all analyses--duplicates, -d: Show N+1 candidates--slow, -s: Show slow queries--slow-threshold N: Slow query threshold in ms (default: 5)--tables, -t: Group queries by table--timeline: Show execution timeline--traceback, -tb: Show code locations for duplicates--profile, -prof: Show Python cProfile data
Listing and Filtering
--list, -l: List requests--limit N: Number of results (default: 20)--sort {recent,queries,duration,db_time}: Sort order--path, -p: Filter by path (contains)--method, -m: Filter by HTTP method--status: Filter by status code--min-queries N: Minimum query count--min-time N: Minimum response time (ms)--min-db-time N: Minimum DB time (ms)
Aggregate Views
--stats: Show aggregate statistics--endpoints: Show endpoint summary with P95--slow-endpoints: Group slow requests by endpoint--slow-endpoint-threshold N: Threshold in ms (default: 200)--min-count N: Minimum requests for endpoint summary
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

