CLI Commands Overview

Complete overview of Circuit Breaker Labs CLI command structure and usage

Command Structure

The Circuit Breaker Labs CLI follows a hierarchical command structure:

cbl [global-options] <command> [command-options] <provider> [provider-options]

Components

1. Global Options - Apply to all commands (API keys, logging, output)
2. Command - Evaluation type (single-turn or multi-turn)
3. Command Options - Specific to the evaluation type
4. Provider - Model provider (openai, ollama, or custom)
5. Provider Options - Configuration for the selected provider

Available Commands

single-turn

Runs a single-turn evaluation where the model responds to individual prompts without conversation context.

cbl single-turn --threshold 0.5 --variations 2 --maximum-iteration-layers 2 openai --model gpt-4o

View complete single-turn documentation →

multi-turn

Runs a multi-turn conversational evaluation that tests the model across multiple conversation turns.

cbl multi-turn --threshold 0.5 --max-turns 8 --test-types user_persona,semantic_chunks openai --model gpt-4o

View complete multi-turn documentation →

Provider Subcommands

All evaluation commands require a provider subcommand:

openai

Use OpenAI or OpenAI-compatible APIs.

Required:

• --api-key - OpenAI API key (or set OPENAI_API_KEY env var)
• --model - Model name (e.g., gpt-4o, gpt-4-turbo, gpt-3.5-turbo)

Optional:

• --base-url - Custom API endpoint
• --temperature - Sampling temperature (0-2)
• --top-p - Nucleus sampling
• And many more options…

ollama

Use locally-hosted Ollama models.

Required:

• --model - Ollama model name

Optional:

• --base-url - Ollama server URL (default: http://localhost:11434)
• --temperature - Sampling temperature
• --num-ctx - Context window size
• And many more options…

custom

Use custom endpoints with Rhai scripting for request/response translation.

Required:

• --url - Endpoint URL to POST to
• --script - Path to Rhai script file

See examples/providers/ for script examples.

Complete Examples

Single-Turn with OpenAI

cbl --output-file results.json \
    single-turn \
    --threshold 0.3 \
    --variations 3 \
    --maximum-iteration-layers 2 \
    --test-case-groups suicidal_ideation \
    openai \
    --model gpt-4o \
    --temperature 1.0

Multi-Turn with Ollama

cbl --log-level debug \
    multi-turn \
    --threshold 0.5 \
    --max-turns 8 \
    --test-types user_persona,semantic_chunks \
    ollama \
    --model llama2 \
    --temperature 0.8

Custom Provider

cbl single-turn \
    --threshold 0.5 \
    --variations 2 \
    --maximum-iteration-layers 2 \
    custom \
    --url https://api.example.com/v1/chat \
    --script ./my-provider.rhai

Getting Help

Get help for any command:

# General help
cbl --help

# Command-specific help
cbl single-turn --help
cbl multi-turn --help

# Provider-specific help
cbl single-turn openai --help
cbl multi-turn ollama --help

Next Steps

Global Options

Configure API keys, logging, and output settings

Single-Turn

Complete single-turn command reference

Multi-Turn

Complete multi-turn command reference

GitHub Examples

View example scripts and configurations