OpenRouter MCP Server

# Benchmark Results Directory This directory contains benchmark test results and reports generated by the OpenRouter MCP Server benchmarking system. ## Directory Structure ``` benchmarks/ ├── *.json # Raw benchmark results in JSON format ├── *.csv # Benchmark reports in CSV format ├── *.md # Benchmark reports in Markdown format └── README.md # This file ``` ## File Naming Convention - **Benchmark Results**: `benchmark_YYYYMMDD_HHMMSS.json` - **Test Benchmarks**: `test_benchmark_YYYYMMDD_HHMMSS.json` - **CSV Reports**: `test_report_YYYYMMDD_HHMMSS.csv` - **Markdown Reports**: `test_report_YYYYMMDD_HHMMSS.md` ## Usage ### Generating New Benchmarks Use the benchmark tools to create new benchmark results: ```python from src.openrouter_mcp.handlers.mcp_benchmark import benchmark_models results = await benchmark_models( models=["openai/gpt-4", "anthropic/claude-3"], prompt="Your test prompt", runs=3 ) ``` ### Viewing Results - **JSON files**: Raw benchmark data for programmatic analysis - **CSV files**: Import into spreadsheet applications for analysis - **MD files**: Human-readable reports with formatted tables ### Cleaning Up Old Results To remove old benchmark files, you can use the cleanup script: ```python from src.openrouter_mcp.handlers.benchmark_cleanup import cleanup_old_benchmarks # Remove files older than 30 days cleanup_old_benchmarks(days_to_keep=30) ``` ## Data Privacy ⚠️ **Note**: Benchmark files may contain API responses and prompts. Do not commit sensitive data to version control. ## See Also - [Benchmarking Guide](../docs/BENCHMARK_GUIDE.md) - Complete benchmarking documentation - [API Documentation](../docs/API.md) - API reference for benchmark tools --- **Last Updated**: 2025-01-12