Getting Started

Get up and running with PawMate in under 5 minutes.

Prerequisites

Before you begin, make sure you have:

• Node.js version 18.0 or higher
• npm (comes with Node.js)
• An AI coding assistant to test (Cursor, GitHub Copilot, etc.)
• 10-15 GB free disk space (for Node modules and generated code)

Check Your Node Version

node --version
# Should show v18.0.0 or higher

Need to Install Node.js?

Download from nodejs.org or use a version manager:

macOS/Linux (nvm):

curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.0/install.sh | bash
nvm install 20
nvm use 20

Windows (nvm-windows): Download from github.com/coreybutler/nvm-windows

Step 1: Install the CLI

Install PawMate globally so the pawmate command is available everywhere:

npm install -g pawmate-ai-challenge

Verify Installation

pawmate --version
# Should show version number like 1.0.0

pawmate --help
# Shows available commands

Step 2: Create a Project Directory

Create a dedicated directory for your benchmark run:

mkdir my-pawmate-benchmark
cd my-pawmate-benchmark

TIP

Use a descriptive name like cursor-model-a-run1 to track multiple runs easily.

Step 3: Initialize a Benchmark Run

Choose a profile and initialize:

pawmate init --profile model-a-rest --tool "Cursor" --tool-ver "v0.43"

Profile Options

Profile	Model	API Style	Description
model-a-rest	A (Minimum)	REST	Best for first-time runs
model-a-graphql	A (Minimum)	GraphQL	Model A with GraphQL
model-b-rest	B (Full)	REST	Advanced with auth
model-b-graphql	B (Full)	GraphQL	Advanced with GraphQL

Start with Model A

If this is your first run, start with model-a-rest. It's simpler and faster (~2-4 hours vs 4-6 hours for Model B).

What Gets Created?

The init command creates a run directory:

pawmate-run-20260103T141320/
├── start_build_api_prompt.txt    # API build prompt
├── start_build_ui_prompt.txt     # UI build prompt (optional)
├── run.config                    # Run metadata
├── PawMate/                      # Workspace for generated code
└── benchmark/                    # Benchmark artifacts
    └── result_submission_instructions.md

Step 4: Copy the Prompt

Find your generated prompt file:

ls pawmate-run-*/

cat pawmate-run-*/start_build_api_prompt.txt

The prompt file contains:

• Your tool name and version
• Selected model (A or B)
• Selected API style (REST or GraphQL)
• Complete specification reference
• Detailed build instructions
• Automated testing requirements

Step 5: Submit to Your AI Agent

1. Open a new chat/session in your AI coding assistant
2. Copy the entire contents of start_build_api_prompt.txt
3. Paste it as the first message (don't modify it!)
4. Let the AI work - Do not interrupt or provide additional context

Do Not Modify the Prompt

The prompt is carefully crafted to ensure fair, comparable benchmarks. Modifying it will invalidate your results.

What Happens Next?

Your AI agent will:

1. ✅ Record the generation_started timestamp
2. ✅ Generate all code files (API, database, tests)
3. ✅ Run npm install to build the project
4. ✅ Load seed data into the database
5. ✅ Start the API server
6. ✅ Run automated tests (iterate until all pass)
7. ✅ Generate benchmark artifacts

Expected Duration:

• Model A: 2-4 hours
• Model B: 4-6 hours

Autonomous Completion

The AI should work autonomously from start to finish. If it stops before completing all steps, send continue to prompt it to resume.

Monitoring Progress

Watch for these completion indicators:

✓ All code written
✓ Build successful (npm install completed)
✓ Seed data loaded and verified
✓ API server started and responsive
✓ All tests passing (100% pass rate)
✓ Benchmark artifacts generated

Common Issues

"Cannot find module" errors

The AI needs to request network permissions for npm install:

required_permissions: ['network']

Most modern AI agents handle this automatically.

AI stops before completion

Send a simple message:

continue

Keep sending continue until all completion criteria are met.

Tests failing

The AI should iterate and fix test failures automatically. If it stops with failing tests, send continue to prompt it to keep working.

Hidden Directory Option

By default, pawmate init creates a visible directory (pawmate-run-<timestamp>/).

Power users can use the --hidden flag for a cleaner directory listing:

pawmate init --profile model-a-rest --tool "Cursor" --hidden
# Creates: .pawmate-run-<timestamp>/

Next Steps

Once your AI agent completes the build:

• Learn how to verify and test your implementation →
• Submit your results →
• Understand the workflow in detail →

Quick Reference

# Install CLI
npm install -g pawmate-ai-challenge

# Initialize a run
pawmate init --profile model-a-rest --tool "YourAI"

# Submit results
pawmate submit pawmate-run-*/benchmark/result.json

Need Help?

• FAQ - Common questions and answers
• CLI Reference - All commands and options
• GitHub Issues - Report problems or ask questions