SpriteStack CLI

SpriteStack is the complete TestOps orchestration layer for TestSprite MCP. It provides a single CLI (spritestack) to run functional, load, and bulk-prompt testing autonomously, generating rich HTML reports and tracking test runs over time.

Installation

Install globally using npm:

npm install -g spritestack

Quick Start

1. Initialize Project:

   spritestack init

   This creates a spritestack.yml configuration file in your directory.

2. Run Tests:

   spritestack run --round 1

   This kicks off TestSprite MCP. It will run the functional and load tests defined in your YAML file, and generate an HTML report inside the local testsprite_tests/ folder.

3. Check Status:

   spritestack status

   View a historical table of all your runs and test rounds.

4. Iterate and Compare: Once you fix the bugs found in Round 1, run Round 2 to see improvement metrics:

   spritestack run --round 2 --compare

5. Open Visual Dashboard:

   spritestack dashboard

   Note: This works if you have the full repository cloned. It will launch the local TestOps dashboard at localhost:4242.

🤖 Guide for LLMs and AI Agents

SpriteStack is engineered for autonomous AI workflows. If you are an AI/LLM, use the following patterns to integrate SpriteStack into your task loops.

1. Generating Configuration

Create the spritestack.yml file non-interactively using:

spritestack init -y

Or directly write to the file system. Example spritestack.yml:

name: Project Name
projectType: auto
baseUrl: http://localhost:3000
tests:
  functional:
    journeys: [login, user-flow]
  load:
    users: 50
    duration: 2m

2. Autonomous Test Verification

To run tests and get a structured, machine-readable JSON result, always use the --output-json flag and --ci flag. The --ci flag will ensure the process exits with code 1 if tests fail, which makes it easy to integrate into continuous agent loops.

spritestack run all --round 1 --ci --output-json ./test-summary.json

JSON Output Format Example (test-summary.json):

{
  "runId": "abc123d",
  "status": "failed",
  "totalTests": 34,
  "passedTests": 22,
  "failedTests": 12,
  "passRate": "64.7",
  "coveragePercent": "61.0",
  "round": 1,
  "issueCount": 3,
  "reportPath": "/path/to/testsprite_tests/round1/report-abc123d.html"
}

3. Reviewing Automated Output

After a run, read the test-summary.json to extract failedTests and the reportPath. If failures exist, read the corresponding HTML report located at reportPath, explicitly analyze the [CRITICAL] and [HIGH] issues, fix the code, and re-run with --round 2 --compare.

4. Viewing Raw Test History

Use the status command to dump the test history table into your shell context:

spritestack status --limit 5

Testing Modalities

Functional Testing

Defined in tests.functional. Specify the key journeys you want TestSprite MCP to execute. TestSprite will generate assertions automatically for these features.

Load Testing

Defined in tests.load. Simulates concurrent virtual users (VUs) and flags resource exhaustion, latency spikes, and database pool limits.

tests:
  load:
    users: 100        # Concurrent connections
    duration: 5m      # Sustain load for 5 minutes
    slo:              # Service Level Objectives
      p95: 300ms
      error_rate: 1%

Bulk Prompt Testing (For AI apps)

Defined in tests.bulkPrompt. A specific suite designed to test generative AI pipelines under load, tracking LLM latency and cost per 1K requests.

tests:
  bulkPrompt:
    enabled: true
    count: 2000
    endpoint: /api/chat
    maxLatency: 5000ms

CLI Reference

• spritestack init [-y] (Initialize project config)
• spritestack run [all|functional|load|bulk-prompt] [--round <n>] [--compare] [--ci] [--output-json <path>] (Run TestSprite test loop)
• spritestack status [-n <limit>] (View a history of all runs)
• spritestack compare [--r1 <n>] [--r2 <n>] (Compare metrics between two rounds)
• spritestack config (Print current merged configuration)
• spritestack dashboard [-p <port>] (Launch graphical TestOps dashboard)

Powered by TestSprite MCP. Made for the TestSprite S2 Hackathon.