Skip to content

QUICKSTART.md

Latest commit

 

History

History
209 lines (157 loc) · 4.61 KB

QUICKSTART.md

File metadata and controls

209 lines (157 loc) · 4.61 KB

Quick Start Guide

🔑 Get FREE API Key (Google Gemini)

  1. Visit: https://aistudio.google.com/app/apikey
  2. Click "Create API Key"
  3. Copy your key
  4. Paste in config.yaml under api_keys.google

No credit card required! Gemini Flash has a generous free tier.

📝 Set Your Problem

Edit PROBLEM.txt with your problem statement:

Given an array of integers, find the maximum sum...

Input Format:
...

Output Format:
...

Constraints:
...

🚀 Run the Solver

python main.py

This will:

  1. Load problem from PROBLEM.txt
  2. Generate test cases
  3. Create brute force solution
  4. Iteratively generate optimal solutions (up to 5 attempts)
  5. Save all attempts with verdicts
  6. Generate workspace/results.json

🎨 View Results

Option 1: Direct Open (Recommended for local files)

# Windows
start viewer.html

# macOS
open viewer.html

# Linux
xdg-open viewer.html

Option 2: HTTP Server (Required for CORS if loading from file://)

python -m http.server 8000

Then visit: http://localhost:8000/viewer.html

📊 What You'll See

Dashboard Stats

  • ✓/✗ Success status
  • Total attempts made
  • Success rate percentage
  • Accepted solutions count

Problem Statement

  • Formatted with LaTeX support
  • Use $formula$ or $$formula$$ for math

Test Data

  • Input test cases
  • Expected output (from brute force)

Brute Force Solution

  • Reference correct solution
  • Syntax highlighted Python code

Optimal Attempts (Reverse Order)

Each card shows:

  • Attempt Number
  • Verdict Badge:
    • 🟢 Accepted (Green)
    • 🔴 Wrong Answer (Red)
    • 🟠 Runtime Error (Orange)
    • ⚫ Generation Failed (Gray)
  • Code (syntax highlighted)
  • Error Message (if any)
  • "Show Output Diff" button (for wrong answers)

Diff Modal

Click "Show Output Diff" to see:

  • Column 1: Input
  • Column 2: Expected Output
  • Column 3: Actual Output

🔧 Customize Your Problem

Edit main.py and change the problem_statement variable:

problem_statement = """
Your problem here...

Input Format:
...

Output Format:
...

Constraints:
- Use $n \leq 10^5$ for LaTeX math
...
"""

Then run python main.py again and refresh the viewer!

📁 Generated Files

All in workspace/:

  • small_inputs.txt - Test cases
  • brute.py - Brute force solution
  • small_outputs.txt - Expected outputs
  • optimal_attempt_1.py - First attempt
  • optimal_attempt_1_output.txt - First attempt output
  • optimal_attempt_2.py - Second attempt
  • ... (up to 10 attempts)
  • results.json - Viewer data

⚙️ Configuration

Edit config.yaml:

# Google Gemini (FREE!)
models:
  tester_agent: "google:gemini-2.5-flash"
  brute_agent: "google:gemini-2.5-flash"
  optimal_agent: "google:gemini-2.5-flash"

execution:
  max_optimal_attempts: 5                # Max retry count
  timeout_seconds: 30                    # Execution timeout

Available Google Gemini Models (FREE Tier):

  • google:gemini-2.5-flash - Fast, efficient (250 free requests/day) ⭐
  • google:gemini-2.5-flash-lite - Faster, cheaper (1000 free requests/day)
  • google:gemini-2.5-pro - Most capable (100 free requests/day)

🐛 Troubleshooting

"Loading results..." stuck

  • Run python main.py first
  • Check workspace/results.json exists
  • Use HTTP server instead of file://

Unicode errors on Windows

  • Already fixed with UTF-8 encoding
  • If still issues, use: chcp 65001 before running

Viewer not loading files

  • Use HTTP server: python -m http.server 8000
  • Files must be in workspace/ directory

API key errors

For Google Gemini (FREE):

🎯 Example Workflow

# 1. Get FREE Gemini API key
# Visit: https://aistudio.google.com/app/apikey
# Add to config.yaml under api_keys.google

# 2. Run solver
python main.py

# 3. Check console output
# See attempts, verdicts, and success/failure

# 4. Open viewer
start viewer.html  # Windows
# or
python -m http.server 8000  # Then visit localhost:8000/viewer.html

# 5. Explore results
# - View all attempts in reverse order
# - Click "Show Output Diff" to debug
# - Check syntax-highlighted code

🌟 Tips

  1. LaTeX Math: Use $...$ for inline math, $$...$$ for display math
  2. Small Test Cases: TesterAgent generates 3-5 small cases for speed
  3. Feedback Loop: Each failed attempt gets detailed feedback for next try
  4. All Saved: Every attempt preserved - nothing is lost!
  5. Reverse Order: Latest attempts shown first for easy access

Enjoy! 🎉