Configuration Reference

This document covers all configuration options available in caro.

Configuration Location

caro stores configuration in platform-specific locations:

Platform	Config Path
macOS	~/Library/Application Support/caro/config.toml
Linux	~/.config/caro/config.toml
Windows	%APPDATA%\caro\config.toml

Configuration File Format

Configuration uses TOML format:

# caro configuration file

[general]
# Default backend to use
backend = "mlx"

# Enable colored output
color = true

# Show safety warnings
safety_warnings = true

[model]
# Model to use for inference
name = "qwen2.5-coder-1.5b-instruct"

# Model format
format = "gguf"

# Quantization level
quantization = "q4_k_m"

[backends.mlx]
# MLX-specific settings
enabled = true
threads = 4

[backends.ollama]
# Ollama backend settings
enabled = false
host = "http://localhost:11434"
model = "qwen2.5-coder:latest"

[backends.vllm]
# vLLM backend settings
enabled = false
url = "http://localhost:8000"

Configuration Options

General Settings

Option	Type	Default	Description
backend	string	"mlx"	Default inference backend
color	bool	true	Enable colored terminal output
safety_warnings	bool	true	Show safety level warnings
confirm_execution	bool	true	Require confirmation before execution

Model Settings

Option	Type	Default	Description
name	string	"qwen2.5-coder-1.5b-instruct"	Model name
format	string	"gguf"	Model file format
quantization	string	"q4_k_m"	Quantization level
cache_dir	string	(auto)	Custom model cache directory

Backend: MLX

Option	Type	Default	Description
enabled	bool	true	Enable MLX backend
threads	int	4	Number of CPU threads
gpu	bool	true	Use GPU acceleration

Backend: Ollama

Option	Type	Default	Description
enabled	bool	false	Enable Ollama backend
host	string	"http://localhost:11434"	Ollama server URL
model	string	"qwen2.5-coder:latest"	Ollama model name
timeout	int	30	Request timeout in seconds

Backend: vLLM

Option	Type	Default	Description
enabled	bool	false	Enable vLLM backend
url	string	"http://localhost:8000"	vLLM server URL
timeout	int	30	Request timeout in seconds

Environment Variables

Configuration can also be set via environment variables:

# Override default backend
export CARO_BACKEND=ollama

# Custom cache directory
export CARO_CACHE_DIR=~/custom/cache

# Enable debug logging
export RUST_LOG=debug

# Disable color output
export NO_COLOR=1

# Ollama host
export OLLAMA_HOST=http://localhost:11434

Command-Line Overrides

Command-line flags override configuration file settings:

# Use specific backend
caro --backend ollama "list files"

# Disable color output
caro --no-color "list files"

# Skip confirmation
caro --yes "list files"

# Verbose output
caro --verbose "list files"

Cache Directory

Model cache location:

Platform	Cache Path
macOS	~/Library/Caches/caro/models/
Linux	~/.cache/caro/models/
Windows	%LOCALAPPDATA%\caro\cache\

Managing Cache

# Show cache location and size
caro cache info

# Clear model cache
caro cache clear

# Download specific model
caro cache download qwen2.5-coder-1.5b-instruct

Example Configurations

Development Setup

[general]
backend = "mlx"
color = true
safety_warnings = true

[backends.mlx]
enabled = true

Server Deployment

[general]
backend = "vllm"
color = false
confirm_execution = false

[backends.vllm]
enabled = true
url = "http://inference-server:8000"
timeout = 60

Multi-Backend Setup

[general]
backend = "ollama"  # Primary

[backends.mlx]
enabled = true

[backends.ollama]
enabled = true
host = "http://localhost:11434"
model = "qwen2.5-coder:latest"

[backends.vllm]
enabled = true
url = "http://backup-server:8000"