GitStress

Paste a GitHub URL. Get the first line that dies under load — and the fix.

GitStress analyzes GitHub repositories and local source trees to identify the precise symbol, route handler, or database query that exhausts resources first under concurrent load. It models saturation ceilings using Little's Law queuing theory, assigns an archival concurrency resilience grade, and outputs a ready-to-apply patch that delivers measurable headroom.

┌─────────────────────────────────────────────────────────────────────────────┐
│  GITSTRESS CONCURRENCY FORENSIC DOSSIER                                     │
├─────────────────────────────────────────────────────────────────────────────┤
│  CULPRIT     http.DefaultTransport                  DIES AT 85 concurrent   │
│  FILE        internal/proxy/proxy.go:11                                     │
│  WHY         DefaultTransport MaxIdleConnsPerHost=2 exhausts pool on burst  │
│  FIX         transport := &http.Transport{ MaxIdleConnsPerHost: 20, ... }   │
│  REPRO       hey -n 1000 -c 85 -m POST http://localhost:7477/...             │
│  PROVENANCE  internal/proxy/proxy.go:11 (code default, not measured)         │
│  HEADROOM    [Apply fix in model] -> 85 -> 850 concurrent (+10x)            │
└─────────────────────────────────────────────────────────────────────────────┘

Quickstart

1. Instant Web Diagnostic (URL Swap)

Replace github.com with gitstress.com in your browser address bar:

https://github.com/tkngate/tkngate
           ↓↓↓
https://gitstress.com/tkngate/tkngate

Featured live presets:

• msh-protocol/msh — Go Swarm Telemetry & Fleet Mesh runtime (dies at 2,500 c/s -> 25,000 c/s, +10x headroom)
• tkngate/tkngate — Go HTTP model proxy (dies at 85 concurrent, +10x headroom)
• calcom/cal.com — Next.js + Prisma unindexed attendeeId query (dies at 18 c/s -> 320 c/s, +17.8x headroom)
• posthog/posthog — Django + Postgres event ingestion (dies at 2,400 req/s -> 18,500 req/s, +7.7x headroom)
• supabase/realtime — Elixir Phoenix WebSockets (in-scope refusal, scheduled for v0.12)
• torvalds/linux — Operating system kernel (honest architectural refusal: ring-0 kernel out of scope)

2. Local CLI Audit

Run directly inside any repository without installation:

# Audit current repository
npx gitstress .

# Audit remote repository by owner/repo slug
npx gitstress tkngate/tkngate
npx gitstress msh-protocol/msh

# Audit via full GitHub URL
npx gitstress https://github.com/calcom/cal.com

3. CI Gate Integration

Prevent concurrency regressions in pull requests:

# Fail CI with exit code 1 if a fatal concurrency bottleneck is detected
npx gitstress . --fail-on-fatal

# Output machine-readable JSON for automated reports
npx gitstress . --json

Key Features

• Predictive Concurrency Forensics & Little's Law Saturation: Deterministic mathematical modeling ($L = \lambda W$) predicting the exact concurrent client threshold where queue delay transitions from linear to exponential saturation without requiring expensive load runners or production downtime.
• Instant Culprit Isolation: Highlights the exact line, AST symbol, and file location where connection pooling, thread pool exhaustion, unindexed queries, or event loop starvation occurs.
• Steve Jobs Executive Concurrency X-Ray & 60 FPS Tachometer HUD: 3-second executive comprehension across 3 clean cards (Collapse Ceiling, AST Culprit, 10x Headroom Fix) featuring an interactive 60 FPS analog tachometer with damped spring physics, tactile 10x Headroom Magic Lever, and instant workload quick-switcher.
• Off-CPU Latency Flamegraph & Sequential Waterfall: Sub-millisecond (0.1ms precision) timeline breakdown across Ingress Gateway, Routing, Worker Dispatch, Pool Contention, and Storage IO with comparative baseline vs. tuned (+10x) latency compression.
• Live Real-Time Traffic Surge Simulator & Circuit Breaker Cockpit: Interactive 60 FPS animated oscilloscope canvas in the Simulate tab modeling Spike Hammer, Sustained Soak, and Chaos Jitter traffic profiles with dynamic Little's Law circuit breaker tripping.
• Interactive Live Patch Sandbox & Unified PR Review Bot: In-browser capacity tuner slider recalculating p99 latency in real-time, split code diff viewer, copyable GitHub PR review comment block, and turnkey .github/workflows/gitstress.yml CI gate.
• Archival Concurrency Stack Matrix (/stacks): Curated directory comparing 10 open-source architectures (Go, Rust, TypeScript, Python, C++, Elixir) with runtime filters, text search, and 1-click duel launchers.
• 2D Pixel Concurrency Dynamometer (PixelQueueRadar.tsx): Real-time 60 FPS interactive architecture simulation with bi-directional micro-pixel packet flow (forward requests and return ACK responses), dynamic 2D queue stacking ladders, live node LED indicators, and instant queue flushing upon applying the 10x headroom patch.
• Streamlined Concurrency Engineering Telemetry: Real-time throughput (req/s), queue latency delay (ms), capacity load percentage, and saturation limits with zero academic paper clutter.
• Framework Concurrency Duels (/vs): Compare two repositories or frameworks head-to-head (e.g. gin-gonic/gin vs tokio-rs/axum, tkngate/tkngate vs msh-protocol/msh, FastAPI vs Express) with live interactive load comparators and 1px telemetry circuits.
• Unified Git Apply Patches: Generates verified, syntax-highlighted code patches providing 5x to 20x concurrency headroom increases.
• Turnkey Repro Scripts: Generates executable load-test commands for k6, hey, autocannon, wrk, and curl.
• Intelligent 404 Repository Triage & Mobile First Interface: Graceful handling for non-existent, mistyped, or private repositories with direct token injection, inline corrections, and fully responsive mobile navigation with an interactive drawer.
• Zero Telemetry Leaks: Static code analysis executes locally in memory. No code is transmitted to external servers.

CLI Reference

Usage:
  gitstress [target] [options]

Arguments:
  target                Local folder path, GitHub owner/repo slug, or full URL (default: .)

Options:
  --json                Output results in JSON format
  --fail-on-fatal       Exit with non-zero code if critical concurrency bottleneck is detected
  --port <port>         Start local web UI server on specified port (default: 4199)
  --version, -v         Display version information
  --help, -h            Show usage help

Local Development & Self-Hosting

Prerequisites

• Node.js >= 18.0.0
• npm or pnpm

Getting Started

# Clone the repository
git clone https://github.com/gitstress/gitstress.git
cd gitstress

# Install dependencies
npm install

# Start public web application (gitstress.com on port 4199)
npm run dev

# Run CLI locally
npm run cli -- .

# Run gauntlet test suite
npm test

# Build production bundles
npm run build:com
npm run build:dashboard

# Start production servers
npm run start:com        # http://localhost:4199 (Public Web)
npm run start:dashboard  # http://localhost:4200 (Enterprise Dashboard)

• Public Web Portal: http://localhost:4199
• Enterprise SaaS Dashboard: http://localhost:4200

Project Structure

gitstress/
├── com/                  # gitstress.com production web application (Next.js 15, Vercel ready)
│   ├── app/              # App Router pages (forensics, duels, dynamometer, hall of fame)
│   ├── src/core/         # AST analysis engine & concurrency modeling
│   ├── auth/             # GitHub authentication & rate-limit handling
│   ├── badges/           # Dynamic SVG concurrency badge service
│   ├── server/           # Commercial runner & API integration server
│   └── package.json      # @gitstress/com workspace manifest
├── oss/                  # Open-source core distribution & CLI tool
│   ├── bin/              # CLI executable entry points (bin/gitstress.ts)
│   ├── src/core/         # Static analysis engine & AST parsers
│   ├── gauntlet/         # 20-repo validation suite & benchmark fixtures
│   └── package.json      # gitstress open-source package manifest
├── docs/                 # Architecture specifications & deployment guides
├── package.json          # Root workspace manifest routing default scripts to com/
└── README.md             # Project documentation

Deploying gitstress.com to Vercel

The com/ folder is pre-configured for zero-friction deployment to Vercel:

1. Import the repository in the Vercel dashboard.
2. Set Root Directory to com.
3. Vercel automatically detects Next.js 15, builds the edge bundle, and links your custom domain gitstress.com.

Supported Stacks & Gauntlet

| Stack | Engine | Default Saturation | Tuned Saturation | Headroom | |---|---|---|---|---| | Go HTTP Proxy (tkngate) | net/http pool | 85 concurrent | 850 concurrent | +10x | | Go Swarm Telemetry (msh) | Swarm mesh | 85 concurrent | 850 concurrent | +10x | | Next.js + Prisma (cal.com) | Node event loop / Postgres | 18 c/s | 320 c/s | +17.8x | | Django + Postgres (posthog) | Python WSGI / Gunicorn | 2,400 req/s | 18,500 req/s | +7.7x | | FastAPI + Async (fastapi) | AnyIO worker pool | 450 req/s | 3,200 req/s | +7.1x | | Express.js (express) | Single-threaded V8 loop | 210 req/s | 1,800 req/s | +8.5x |

Contributing

Contributions are welcomed. Please follow these guidelines:

1. Fork the repository and create a descriptive feature branch (feat/feature-name).
2. Adhere to existing code formatting and architectural patterns.
3. Ensure all gauntlet tests pass via npm test.
4. Submit a clear pull request describing the changes and test coverage.

Please review our Contributing Guidelines and Code of Conduct.

License

GitStress is licensed under the MIT License.