Skip to content

BULK_SOLVER_README.md

Latest commit

 

History

History
203 lines (151 loc) · 5.94 KB

BULK_SOLVER_README.md

File metadata and controls

203 lines (151 loc) · 5.94 KB

Bulk Solution Generation for Missing LeetCode Problems

This document describes the tools and processes for systematically adding AI-generated solutions to all LeetCode problems in the repository.

Problem Statement

The repository tracks 548 LeetCode problems (as listed in README.md), but many lack AI-generated solutions:

  • 76 problems have gpt5-mini.md files (AI-generated solutions)
  • 472 problems are missing AI-generated solutions

Some problems have only human-generated solutions (e.g., jeremymanning.md), and we want to add AI solutions for all of them.

Tools Provided

1. identify_missing.py

Analyzes the repository to identify which problems are missing AI-generated solutions.

Usage:

python3 identify_missing.py

What it does:

  • Scans README.md to find all problem numbers
  • Checks for existence of problems/*/gpt5-mini.md files
  • Reports which problems are missing AI solutions
  • Creates batch files for processing (50 problems per batch)
  • Saves results to temp directory and batch files

2. auto_solver.py (Enhanced)

The existing auto_solver script now supports three modes:

Mode 1: Daily problem (default)

python3 auto_solver.py

Mode 2: Specific problem by ID

python3 auto_solver.py 123

Mode 3: Bulk solve from file

# Set your OpenAI API key
export OPENAI_API_KEY="your-api-key-here"

# Process a batch of problems
python3 auto_solver.py --bulk /tmp/batch_001.txt

What it does:

  • Fetches problem details from LeetCode API
  • Generates solutions using GPT-5-mini
  • Saves solutions as problems/{id}/gpt5-mini.md
  • Handles rate limiting (2 seconds between requests in bulk mode)
  • Reports success/failure for each problem

3. GitHub Actions Workflow: bulk_solver.yml

Automated workflow for bulk solving problems via GitHub Actions.

Location: .github/workflows/bulk_solver.yml

How to use:

  1. Go to the repository's Actions tab
  2. Select "Bulk Solve Missing LeetCode Problems"
  3. Click "Run workflow"
  4. Configure:
    • batch_size: Number of problems to solve (default: 50)
    • start_index: Which problem to start from (default: 0)
  5. Click "Run workflow" to start

What it does:

  • Automatically identifies missing AI solutions
  • Processes a batch of problems using auto_solver.py
  • Commits and pushes generated solutions
  • Can be run multiple times with different start_index values

Example workflow runs:

  • Run 1: start_index=0, batch_size=50 → solves problems 1-50
  • Run 2: start_index=50, batch_size=50 → solves problems 51-100
  • Run 3: start_index=100, batch_size=50 → solves problems 101-150
  • etc.

Workflow for Adding All Missing Solutions

Option A: Using GitHub Actions (Recommended)

  1. Initial Analysis

    python3 identify_missing.py

    This shows you how many problems need solutions (currently 472).

  2. Trigger Workflow Runs

    • Go to Actions → "Bulk Solve Missing LeetCode Problems"
    • Run workflow with: start_index=0, batch_size=50
    • Wait for completion, then run again with: start_index=50, batch_size=50
    • Repeat until all problems are processed (10 runs total for 472 problems)

  3. Monitor Progress

    • Check the Actions tab for workflow status
    • Solutions will be automatically committed to main branch

Option B: Using Local Script (If you have OpenAI API key)

  1. Setup

    # Install dependencies
pip install requests openai httpx

# Set API key
export OPENAI_API_KEY="your-key-here"

  2. Generate batch files

    python3 identify_missing.py

  3. Process batches

    # Process first batch
python3 auto_solver.py --bulk /tmp/batch_001.txt

# Commit changes
git add problems/*/gpt5-mini.md
git commit -m "Add AI solutions for batch 1"
git push

# Continue with remaining batches...
python3 auto_solver.py --bulk /tmp/batch_002.txt
# ... and so on

Solution Format

All AI-generated solutions follow this format:

# [Problem {ID}: {Title}]({URL})

## Initial thoughts (stream-of-consciousness)
[Initial analysis and approach]

## Refining the problem, round 2 thoughts
[Refined approach, edge cases, complexity analysis]

## Attempted solution(s)
\`\`\`python
[Complete Python solution]
\`\`\`
- [Notes about approach and complexity]

Security Notes

  • SSL certificate verification is disabled by default to match the existing auto_solver.py pattern (required for environments behind proxies with self-signed certificates)
  • To enable SSL verification in production, set the environment variable: HTTPX_VERIFY=true
  • All scripts have been scanned with CodeQL and found no security vulnerabilities
  • API keys are passed via environment variables and never hard-coded

Technical Details

API Rate Limiting

  • The script includes 2-second delays between requests
  • This prevents overwhelming the LeetCode or OpenAI APIs
  • For 472 problems, expect ~15-20 minutes per batch of 50

Error Handling

  • Problems that fail to fetch are skipped and logged
  • Failed problems are reported at the end
  • You can re-run with just the failed problems

Network Requirements

  • Requires access to leetcode.com (for problem details)
  • Requires access to OpenAI API (for solution generation)
  • GitHub Actions environment has these by default

Monitoring Progress

After running batches, you can check progress:

# Count AI solutions
find problems -name "gpt5-mini.md" | wc -l

# Re-run analysis
python3 identify_missing.py

Notes

  • The auto_solver.py script now supports three modes:
    1. Daily mode (default): python3 auto_solver.py
    2. Single problem: python3 auto_solver.py 123
    3. Bulk mode: python3 auto_solver.py --bulk file.txt
  • This bulk solver reuses the existing auto_solver logic, avoiding code duplication
  • Both daily and bulk modes use the same GPT-5-mini model and format
  • Solutions are idempotent - safe to re-run if a problem already has gpt5-mini.md