Skip to content

CHANGELOG.md

Latest commit

 

History

History
72 lines (53 loc) · 3.19 KB

CHANGELOG.md

File metadata and controls

72 lines (53 loc) · 3.19 KB
Raw

Changelog

v1.2.0 — 2026-05-17

  • Improved report UX with a new culprit-summary/digest section and added summary export to CSV.

v1.1.1 — 2026-03-09

Core Improvements

Per-Core CPU Monitoring

  • System metrics now capture per-core CPU usage (cpu_per_core) at each event via psutil.cpu_percent(percpu=True)
  • cpu_percent is now the system-wide average (sum of all cores / core count) instead of raw per-process value which could exceed 100%
  • Per-core data displayed in both HTML report (color-coded badges) and TUI detail view

Signal Handling Fix

  • Fixed Ctrl+C hang: signal handler now correctly calls stop(), restores default signal behavior, and re-raises the signal
  • Process terminates cleanly even with multiple uvicorn workers

Analyzer Enhancements

  • Added hint field for transition blocks
  • Added async_total_time stat — tracks cumulative duration of slow async tasks separately from blocking time
  • Fixed avg_duration to only count blocking events (previously included async events)

CLI

  • Added --version flag via importlib.metadata

HTML Report Overhaul

  • Grouped View (default) — events grouped by offending user-code location, sorted by total duration. Shows count, total, avg, CPU range per group
  • Timeline View — flat chronological list with per-column filtering and 1000/page pagination
  • Stable Event IDs — each event gets a permanent #ID (0, 1, 2...) that never changes with sorting/filtering
  • Per-core CPU in detail panel — color-coded per-core breakdown (green/orange/red)
  • Stat card tooltips — each summary metric now explains how it's calculated and notes that times are cumulative sums, not wall-clock
  • Column headers in groups — expanded groups show sub-headers so columns are always readable
  • Removed timeline bar chart
  • Full-width layout

Examples

New: app_async.py

  • Fully async version of the demo app using aiosqlite, aiohttp, asyncio.to_thread, asyncio.sleep
  • SharedAiohttpSession — lock-based lazy singleton, auto-recreates if closed
  • SharedAiosqlitePool — connection pool with health checks and SQLite optimizations (WAL, synchronous=NORMAL, cache_size, mmap, temp_store=MEMORY)
  • No comments/docstrings — clean reference implementation

New: mock_server.py

  • Local aiohttp mock server replacing httpbin.org for stress testing
  • Routes: /delay/{s}, /status/{code}, /get, /post, /anything
  • Runs on 127.0.0.1:9999

Upgraded: loadtest.py

  • Concurrent load tester (100-1000 configurable concurrency)
  • Configurable duration, slow request threshold
  • Live progress bar with RPS, success/fail/slow counts
  • Generates standalone HTML + JSON reports with per-endpoint breakdown and P50/P95/P99 latencies

Updated: app.py

  • External endpoints now hit local mock server instead of httpbin.org
  • Switched from httpx to aiohttp for async HTTP
  • Graceful shutdown ensures sentry.stop() is called

Dependencies

  • Added aiohttp, aiosqlite, orjson to [project.optional-dependencies.examples]

Documentation

  • Added logo
  • Added report interpretation guide
  • Updated CLI reference with --version
  • Updated example run instructions for mock server + load tester