PerfWeaver

Weaving real traffic into performance tests

Convert browser HAR (HTTP Archive) traffic into structured load-testing scripts for multiple performance testing tools.

Features

• HAR Parsing: Extract and structure HTTP requests from browser recordings
• Behavior Modeling: Detect sequential blocks, parallel execution, and transaction boundaries
• Correlation Engine: Auto-detect and extract dynamic values (tokens, session IDs, etc.)
• Smart Parameterization: Replace environment-specific values with configurable variables
• Think-Time Modeling: Simulate realistic user pacing based on HAR timestamps
• Multi-Tool Support: Generate scripts for:
  • Apache JMeter
  • k6
  • Gatling
  • Locust
  • Playwright (API load mode)

Architecture

The system uses a layered architecture to ensure extensibility:

HAR Input
    ↓
HAR Parsing Engine
    ↓
Behavior Modeling Engine
    ↓
Correlation Engine
    ↓
Parameterization Engine
    ↓
Tool Adapters
    ↓
Generated Load Test Scripts

Installation

Install from npm:

npm install perfweaver

You can also use it via npx without installation:

npx perfweaver analyze example.har

For development:

git clone https://github.com/ashwin-ak/perfweaver.git
cd perfweaver
npm install
npm run build

Or for development:

git clone <repo>
cd perfweaver
npm install
npm run build

Quick Start

Generate a JMeter Script

perfweaver generate --tool jmeter --har login.har --output login.jmx

Generate a k6 Script

perfweaver generate --tool k6 --har checkout.har --output checkout.js

Analyze HAR File

perfweaver analyze login.har

Visualize HAR Structure

perfweaver visualize login.har --output report.html

Configuration

Create a perfweaver.config.yaml in your project root:

filters:
  ignoreExtensions:
    - png
    - jpg
    - gif
    - css
    - woff
    - woff2
    - ttf
    - svg
  ignoreResourceTypes:
    - image
    - stylesheet
    - font
    - media

parallelDetection:
  overlapThresholdMs: 40

correlation:
  enableAutoCorrelation: true
  extractors:
    - type: json
    - type: regex
    - type: xpath

parameterization:
  envVars:
    - BASE_URL
    - AUTH_TOKEN
    - USER_ID

loadModel:
  threadCount: 10
  rampUpTime: 60
  duration: 300
  iterations: 1

tools:
  enabled:
    - jmeter
    - k6
    - gatling

Module Structure

• core/har-parser: Parse and extract HAR data
• core/behavior-model: Detect transactions and request relationships
• core/correlation-engine: Auto-detect dynamic values
• core/parameterization: Replace values with variables
• core/think-time: Model user pacing
• core/load-model: Configure load profiles
• adapters/: Tool-specific script generators
• cli/: Command-line interface
• tests/: Unit and integration tests

Usage Examples

See the examples directory for sample HAR files and generated scripts.

Documentation

Complete user and developer guides are available in the docs/ folder:

• Quick Start
• API & CLI Reference
• Performance Tuning
• Developer Guide
• Architecture Overview

Contributing

Contributions are welcome! Please follow the coding standards and add tests for new features.

License

MIT

Author

Ashwin Kulkarni