Skip to content

cli-reference.md

Latest commit

 

History

History
364 lines (273 loc) · 8.43 KB

cli-reference.md

File metadata and controls

364 lines (273 loc) · 8.43 KB
Raw

CLI Reference Guide

Overview

This guide provides comprehensive documentation for all Trust AI CLI commands, options, and usage patterns.

Global Options

All commands support these global options:

--config, -c <PATH>     Use custom configuration file
--verbose, -v           Enable verbose output
--quiet, -q            Suppress non-essential output
--help, -h             Show help information
--version, -V          Show version information

Commands

trust-ai chat

Start interactive conversations with AI models.

Usage

trust-ai chat [OPTIONS] [MESSAGE]

Options

--continue              Continue previous conversation
--model <MODEL>         Override default model
--temperature <FLOAT>   Set response creativity (0.0-2.0)
--max-tokens <INT>      Maximum response length
--system <TEXT>         Set system prompt
--save <NAME>           Save conversation with name
--load <NAME>           Load saved conversation
--export <FORMAT>       Export conversation (json, markdown, txt)

Examples

# Interactive chat session
trust-ai chat

# Single message
trust-ai chat "Explain machine learning"

# Continue previous conversation
trust-ai chat --continue "Can you elaborate on that?"

# Use specific model with custom temperature
trust-ai chat --model gpt-4 --temperature 0.3 "Write a formal email"

# Set system prompt
trust-ai chat --system "You are a helpful coding assistant" "Help me debug this code"

# Save conversation
trust-ai chat --save "ml-discussion" "Explain neural networks"

# Load and continue saved conversation
trust-ai chat --load "ml-discussion" --continue "What about deep learning?"

trust-ai config

Manage configuration settings.

Subcommands

config show

Display current configuration.

trust-ai config show [--format json|yaml|table]
config set

Set configuration values.

trust-ai config set <KEY> <VALUE>

Available Keys:

  • provider: AI provider (openai, ollama, anthropic)
  • model: Default model name
  • api_key: API key for the provider
  • endpoint: Custom API endpoint
  • temperature: Default temperature (0.0-2.0)
  • max_tokens: Default maximum tokens
  • timeout: Request timeout in seconds
  • cache.enabled: Enable/disable caching
  • cache.max_size: Maximum cache entries
  • cache.ttl: Cache time-to-live in seconds
  • monitoring.enabled: Enable performance monitoring
  • monitoring.metrics_retention: Metrics retention period
config get

Get specific configuration value.

trust-ai config get <KEY>
config reset

Reset configuration to defaults.

trust-ai config reset [--confirm]

Examples

# View all configuration
trust-ai config show

# Set provider and API key
trust-ai config set provider openai
trust-ai config set api_key sk-your-key-here

# Configure Ollama
trust-ai config set provider ollama
trust-ai config set endpoint http://localhost:11434
trust-ai config set model llama2

# Enable caching with custom settings
trust-ai config set cache.enabled true
trust-ai config set cache.max_size 2000
trust-ai config set cache.ttl 7200

# Get specific value
trust-ai config get model

# Reset to defaults
trust-ai config reset --confirm

trust-ai perf

Performance monitoring and optimization commands.

Subcommands

perf status

Display performance dashboard.

trust-ai perf status [--format table|json] [--period 1h|1d|1w]
perf cache

Show cache performance metrics.

trust-ai perf cache [--clear] [--stats]
perf monitor

Real-time performance monitoring.

trust-ai perf monitor [--live] [--interval <SECONDS>] [--duration <MINUTES>]
perf optimize

Get optimization suggestions.

trust-ai perf optimize [--apply] [--dry-run]
perf metrics

Export performance metrics.

trust-ai perf metrics [--format csv|json] [--output <FILE>] [--period <DURATION>]
perf benchmark

Run performance benchmarks.

trust-ai perf benchmark [--iterations <COUNT>] [--model <MODEL>]
perf profile

Profile system performance.

trust-ai perf profile [--duration <SECONDS>] [--output <FILE>]
perf report

Generate performance reports.

trust-ai perf report [--format html|pdf|markdown] [--output <FILE>]

Examples

# View performance dashboard
trust-ai perf status

# Monitor cache performance
trust-ai perf cache --stats

# Real-time monitoring for 5 minutes
trust-ai perf monitor --live --duration 5

# Get optimization suggestions
trust-ai perf optimize

# Apply optimizations automatically
trust-ai perf optimize --apply

# Export metrics to CSV
trust-ai perf metrics --format csv --output metrics.csv --period 7d

# Run benchmark with 10 iterations
trust-ai perf benchmark --iterations 10

# Profile for 30 seconds
trust-ai perf profile --duration 30 --output profile.json

# Generate HTML performance report
trust-ai perf report --format html --output report.html

trust-ai init

Initialize Trust AI configuration and setup.

trust-ai init [--force] [--config-dir <PATH>]

Options

--force              Overwrite existing configuration
--config-dir <PATH>  Custom configuration directory

trust-ai version

Display version and build information.

trust-ai version [--format table|json]

trust-ai completions

Generate shell completions.

trust-ai completions <SHELL>

Supported Shells: bash, zsh, fish, powershell

Examples

# Generate bash completions
trust-ai completions bash > ~/.bash_completion.d/trust-ai

# Generate zsh completions
trust-ai completions zsh > ~/.zsh/completions/_trust-ai

# Generate fish completions
trust-ai completions fish > ~/.config/fish/completions/trust-ai.fish

Configuration File Format

The configuration file (~/.config/trust-ai/config.yaml) supports these options:

# Provider settings
provider: openai              # openai, ollama, anthropic
model: gpt-4                 # Default model
api_key: your-api-key        # API key
endpoint: https://api.openai.com/v1  # Custom endpoint
timeout: 30                  # Request timeout (seconds)

# Response settings
temperature: 0.7             # Creativity (0.0-2.0)
max_tokens: 2048            # Maximum response length
system_prompt: ""           # Default system prompt

# Cache settings
cache:
  enabled: true             # Enable caching
  max_size: 1000           # Maximum cache entries
  ttl: 3600               # Time-to-live (seconds)

# Monitoring settings
monitoring:
  enabled: true            # Enable performance monitoring
  metrics_retention: 7d   # How long to keep metrics
  auto_optimize: false    # Automatic optimization

# UI settings
ui:
  color: auto             # auto, always, never
  progress: true          # Show progress indicators
  timestamps: false       # Show timestamps in output

Environment Variables

Configuration can also be set via environment variables:

# Provider settings
TRUST_AI_PROVIDER=openai
TRUST_AI_MODEL=gpt-4
TRUST_AI_API_KEY=your-key
TRUST_AI_ENDPOINT=https://api.openai.com/v1
TRUST_AI_TIMEOUT=30

# Response settings
TRUST_AI_TEMPERATURE=0.7
TRUST_AI_MAX_TOKENS=2048
TRUST_AI_SYSTEM_PROMPT="You are a helpful assistant"

# Cache settings
TRUST_AI_CACHE_ENABLED=true
TRUST_AI_CACHE_MAX_SIZE=1000
TRUST_AI_CACHE_TTL=3600

# Monitoring settings
TRUST_AI_MONITORING_ENABLED=true
TRUST_AI_METRICS_RETENTION=7d

Exit Codes

Trust AI uses standard exit codes:

  • 0: Success
  • 1: General error
  • 2: Configuration error
  • 3: Network error
  • 4: Authentication error
  • 5: Rate limit exceeded
  • 6: Invalid input
  • 7: File system error

Tips and Best Practices

Performance Optimization

  • Enable caching for frequently used prompts
  • Use appropriate temperature settings for your use case
  • Monitor performance with trust-ai perf status
  • Run periodic optimizations with trust-ai perf optimize

Configuration Management

  • Use environment variables for CI/CD environments
  • Keep sensitive data like API keys in environment variables
  • Use different configuration files for different projects

Troubleshooting

  • Use --verbose flag for detailed output
  • Check configuration with trust-ai config show
  • Monitor performance with trust-ai perf monitor
  • Clear cache if experiencing issues: trust-ai perf cache --clear

For more detailed information, see the Configuration Guide and Best Practices.