Skip to main content
The profile command provides performance profiling and benchmarking capabilities for AI agents.

​
Quick Start

praisonai profile benchmark "What is 2+2?"

​
Subcommands

​
Benchmark

Run performance benchmarks on prompts. 
praisonai profile benchmark "Your prompt here"
Options:
OptionDescription
--model, -mModel to use for benchmarking
--iterations, -nNumber of iterations (default: 1)
--output, -oOutput file path for results
--format, -fOutput format: table, json, html
--verbose, -vEnable verbose output
Example: 
praisonai profile benchmark "Write a poem about AI" --model gpt-4o --iterations 3
Output: 
📊 Benchmark Results
┏━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Metric       ┃ Value    ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Iterations   │ 3        │
│ Average Time │ 2500.5ms │
│ Min Time     │ 2100.3ms │
│ Max Time     │ 2900.7ms │
│ Model        │ gpt-4o   │
└──────────────┴──────────┘

​
Compare

Compare performance across multiple models. 
praisonai profile compare "Explain quantum computing" --models gpt-4o,gpt-4o-mini
Options:
OptionDescription
--modelsComma-separated list of models to compare

​
Report

Generate a profiling report from saved results. 
praisonai profile report --input results.json --output report.html

​
Use Cases

​
Model Selection

# Compare models to find the best balance of speed and quality
praisonai profile compare "Complex analysis task" --models gpt-4o,gpt-4o-mini,claude-3-sonnet

​
Performance Optimization

# Benchmark with multiple iterations for reliable metrics
praisonai profile benchmark "Your prompt" --iterations 10 --output benchmark.json

​
CI/CD Integration

# Export results as JSON for automated pipelines
praisonai profile benchmark "Test prompt" --format json --output results.json