Configuration

hi-shell uses a TOML configuration file for all settings. The config file is automatically created during hi-shell --init.

Config File Location

Full Config Reference

# Unique installation identifier (auto-generated)
installation_id = "550e8400-e29b-41d4-a716-446655440000"

# LLM Provider: "Embedded", "Local", or "Cloud"
llm_provider = "Embedded"

# --- Embedded Model Settings ---
# Hugging Face model repository
embedded_model = "lmstudio-community/Llama-3.2-1B-Instruct-GGUF"
# Specific GGUF file to use
embedded_model_file = "Llama-3.2-1B-Instruct-Q4_K_M.gguf"

# --- Local Provider Settings ---
# Local provider type: "Ollama" or "LmStudio"
local_provider = "Ollama"
# API endpoint URL
local_url = "http://localhost:11434"
# Model name
local_model = "llama3.2"

# --- Cloud Provider Settings ---
# Cloud provider: "OpenRouter", "Anthropic", "OpenAI", "Gemini", or "Custom"
cloud_provider = "OpenRouter"
# Model identifier
cloud_model = "microsoft/phi-3-mini-4k-instruct"
# Custom API URL (for Custom provider only)
cloud_custom_url = ""
# API key
api_key = ""

# --- Telemetry ---
# Anonymous usage statistics (opt-in)
telemetry_enabled = false

# --- Updates ---
# Last update check timestamp (ISO 8601)
last_update_check = "2024-01-15T10:30:00Z"

Provider Configuration

Embedded

llm_provider = "Embedded"
embedded_model = "lmstudio-community/Llama-3.2-1B-Instruct-GGUF"
embedded_model_file = "Llama-3.2-1B-Instruct-Q4_K_M.gguf"

Local (Ollama)

llm_provider = "Local"
local_provider = "Ollama"
local_url = "http://localhost:11434"
local_model = "llama3.2"

Local (LM Studio)

llm_provider = "Local"
local_provider = "LmStudio"
local_url = "http://localhost:1234"
local_model = "TheBloke/Llama-2-7B-Chat-GGUF"

Cloud (OpenRouter)

llm_provider = "Cloud"
cloud_provider = "OpenRouter"
cloud_model = "microsoft/phi-3-mini-4k-instruct"
api_key = "sk-or-..."

Cloud (Anthropic)

llm_provider = "Cloud"
cloud_provider = "Anthropic"
cloud_model = "claude-sonnet-4-20250514"
api_key = "sk-ant-..."

Cloud (OpenAI)

llm_provider = "Cloud"
cloud_provider = "OpenAI"
cloud_model = "gpt-4o-mini"
api_key = "sk-..."

Cloud (Custom)

llm_provider = "Cloud"
cloud_provider = "Custom"
cloud_custom_url = "https://your-llm-api.example.com/v1/chat/completions"
cloud_model = "your-model-name"
api_key = "your-api-key"

CLI Flags

Environment Variables

hi-shell does not use environment variables for configuration. All settings are managed through the TOML config file.

Security Best Practices

1. Protect your config file — It contains API keys

   chmod 600 ~/.config/hi-shell/config.toml

2. Don’t commit your config — Add to .gitignore if sharing dotfiles

3. Rotate compromised keys — If your config is exposed, rotate API keys immediately

4. Use local providers — For maximum privacy, use Embedded or Ollama providers