Configuration - djvolz/coda-code-assistant GitHub Wiki

Configuration

Coda uses a flexible configuration system based on TOML files with support for multiple configuration sources and environment variable overrides.

Quick Setup

# Copy the example configuration
cp examples/config.example.toml ~/.config/coda/config.toml

# Edit with your settings
nano ~/.config/coda/config.toml

# Set required environment variables
export OCI_COMPARTMENT_ID="ocid1.compartment.oc1..example"

Configuration Sources

Coda loads configuration from multiple sources in priority order (highest to lowest):

Environment variables (highest priority)
Project-specific: .coda/config.toml in current directory
User-specific: ~/.config/coda/config.toml
System-wide: /etc/coda/config.toml (Unix-like systems)

Settings from higher priority sources override those from lower priority sources.

Configuration Format

Coda uses TOML format for configuration files. Here's a complete example:

# Default provider to use
default_provider = "oci_genai"

# Provider configurations
[providers.oci_genai]
compartment_id = "ocid1.compartment.oc1..example"
config_profile = "DEFAULT"
config_file_location = "~/.oci/config"

[providers.litellm]
api_base = "http://localhost:8000"
default_model = "gpt-4"

[providers.ollama]
host = "http://localhost:11434"
default_model = "llama3"

# Session settings
[session]
history_file = "~/.local/share/coda/history"
max_history = 1000
autosave = true

# UI settings
[ui]
theme = "default"  # Options: default, dark, light, minimal, vibrant, monokai_dark, monokai_light, dracula_dark, dracula_light, gruvbox_dark, gruvbox_light
show_model_info = true
show_token_usage = false
show_thinking_indicator = true

# Tool settings
[tools]
enabled = true
filesystem_enabled = true
git_enabled = true
web_enabled = true
system_enabled = true
require_confirmation = false  # Ask before executing potentially dangerous operations

# Agent settings
[agent]
enabled = true
max_iterations = 10  # Maximum tool calls per conversation turn
show_tool_calls = true  # Display tool usage in conversation

# Generation parameters
temperature = 0.7
max_tokens = 2000

# Debug mode
debug = false

Core Settings

Basic Configuration

Setting	Type	Default	Description
`default_provider`	string	`"oci_genai"`	Default AI provider to use
`debug`	boolean	`false`	Enable debug mode with verbose logging
`temperature`	float	`0.7`	Generation temperature (0-2)
`max_tokens`	integer	`2000`	Maximum tokens to generate

Provider Configuration

OCI GenAI Provider

[providers.oci_genai]
compartment_id = "ocid1.compartment.oc1..example"  # Required
config_profile = "DEFAULT"                         # Optional
config_file_location = "~/.oci/config"            # Optional

compartment_id: (Required) Your OCI compartment OCID
config_profile: OCI config profile name (default: "DEFAULT")
config_file_location: Path to OCI config file (default: "~/.oci/config")

LiteLLM Provider

[providers.litellm]
api_base = "http://localhost:8000"
default_model = "gpt-4"
api_key = "your-api-key"  # Optional, can use env vars

api_base: Base URL for LiteLLM proxy server
default_model: Default model to use
api_key: API key (prefer environment variables)

Ollama Provider

[providers.ollama]
host = "http://localhost:11434"
default_model = "llama3"
timeout = 30

host: Ollama server host and port
default_model: Default model to use
timeout: Request timeout in seconds

Session Configuration

[session]
history_file = "~/.local/share/coda/history"
max_history = 1000
autosave = true
database_path = "~/.cache/coda/sessions.db"

history_file: Command history file location
max_history: Maximum command history entries
autosave: Enable automatic session saving
database_path: Session database location

UI Configuration

[ui]
theme = "default"
show_model_info = true
show_token_usage = false
show_thinking_indicator = true
prompt_style = "default"

theme: UI color theme. Available themes:
- default - Default balanced theme
- dark - Dark mode optimized for low light
- light - Light theme for bright environments
- minimal - Minimal colors for focused work
- vibrant - High contrast with vibrant colors
- monokai_dark - Monokai color scheme - dark variant
- monokai_light - Monokai color scheme - light variant
- dracula_dark - Dracula theme - dark variant
- dracula_light - Dracula theme - light variant
- gruvbox_dark - Gruvbox retro groove - dark variant
- gruvbox_light - Gruvbox retro groove - light variant
show_model_info: Display current model in prompt
show_token_usage: Show token usage statistics
show_thinking_indicator: Display animated indicator while AI is processing
prompt_style: Prompt appearance style

Environment Variables

Environment variables override configuration file settings:

Variable	Description	Example
`CODA_DEFAULT_PROVIDER`	Default provider	`export CODA_DEFAULT_PROVIDER=ollama`
`CODA_DEBUG`	Enable debug mode	`export CODA_DEBUG=true`
`CODA_TEMPERATURE`	Generation temperature	`export CODA_TEMPERATURE=0.5`
`CODA_MAX_TOKENS`	Maximum tokens	`export CODA_MAX_TOKENS=4000`

Provider-Specific Environment Variables

OCI GenAI:

export OCI_COMPARTMENT_ID="ocid1.compartment.oc1..example"
export OCI_CONFIG_PROFILE="MY_PROFILE"
export OCI_CONFIG_FILE="~/my-oci-config"

LiteLLM:

export LITELLM_API_BASE="http://localhost:8000"
export LITELLM_API_KEY="your-api-key"

Ollama:

export OLLAMA_HOST="http://localhost:11434"

Configuration Examples

Personal Development Setup

default_provider = "ollama"

[providers.ollama]
host = "http://localhost:11434"
default_model = "codellama"

[session]
autosave = true
max_history = 2000

[ui]
show_model_info = true
show_token_usage = true

Team/Corporate Setup

default_provider = "oci_genai"

[providers.oci_genai]
compartment_id = "ocid1.compartment.oc1..corporate"
config_profile = "TEAM_PROFILE"

[providers.litellm]
api_base = "https://internal-llm-proxy.company.com"

[session]
autosave = true
database_path = "/shared/coda/sessions.db"

[ui]
theme = "corporate"
show_token_usage = true

Project-Specific Configuration

Create .coda/config.toml in your project directory:

# Use specific model for this project
default_provider = "oci_genai"

[providers.oci_genai]
compartment_id = "ocid1.compartment.oc1..project_specific"

# Project-specific generation settings
temperature = 0.3  # More deterministic for code
max_tokens = 4000  # Longer responses for complex code

[session]
autosave = true
# Project sessions in project directory
database_path = ".coda/sessions.db"

Setup Guides

First-Time Setup

Create configuration directory:
```
mkdir -p ~/.config/coda
```

Copy example configuration:

cp examples/config.example.toml ~/.config/coda/config.toml

Edit configuration:
```
nano ~/.config/coda/config.toml
```

Set environment variables:

echo 'export OCI_COMPARTMENT_ID="your-compartment-id"' >> ~/.bashrc
source ~/.bashrc

Validating Configuration

Test your configuration:

# Check if configuration loads correctly
coda --version

# Test with debug mode
CODA_DEBUG=true coda "Hello, world!"

# List available models
coda
> /model

Troubleshooting

Common Issues

Configuration not found:

Check file path: ~/.config/coda/config.toml
Verify file permissions are readable
Use absolute paths if relative paths fail

Provider authentication fails:

Verify environment variables are set correctly
Check OCI config file and profile
Test credentials with OCI CLI tools

Permission errors:

Ensure write permissions for session database directory
Check XDG directory permissions
Use chmod to fix permission issues

Debug Configuration

Enable debug mode to troubleshoot:

CODA_DEBUG=true coda

This will show:

Configuration loading process
Provider initialization
API requests and responses
Session operations

Advanced Configuration

Custom Provider Setup

You can configure multiple provider instances:

[providers.oci_genai_dev]
compartment_id = "ocid1.compartment.oc1..dev"
config_profile = "DEV"

[providers.oci_genai_prod]
compartment_id = "ocid1.compartment.oc1..prod"
config_profile = "PROD"

Configuration Validation

Coda validates configuration on startup:

Required fields for enabled providers
Valid TOML syntax
File permissions and accessibility
Environment variable format