Reference

Configuration Reference

Complete reference for the omni.toml configuration file.

File Location

Omni reads its configuration from a TOML file. The file is created automatically on first launch with sensible defaults. You can edit it manually or through the Settings panel in the UI.

Linux

~/.config/omni/omni.toml

macOS

~/Library/Application Support/Omni/omni.toml

Windows

%APPDATA%\Omni\omni.toml

All sections are optional. Omni uses defaults for any missing values.

[general]

Top-level application settings like logging and data storage.

Key

Type

Default

Description

data_dir

string?

OS default

Override the data directory path. Uses OS-specific default if not set.

telemetry

bool

false

Enable anonymous usage telemetry.

log_level

string

"info"

Log verbosity. Values: trace, debug, info, warn, error.

max_history

integer

1000

Maximum number of messages to keep in session history.

omni.toml

[general]

telemetry = false

log_level = "info"

max_history = 1000

[providers.<name>]

Configure one or more LLM providers. Each provider has a unique key (e.g., providers.openai). You can configure multiple providers and Omni will automatically fall back between them.

Key

Type

Default

Description

provider_type

string

required

Provider type: openai, anthropic, ollama, gemini, bedrock, custom.

default_model

string?

—

Default model to use (e.g., "gpt-4o", "claude-opus-4-6").

endpoint

string?

—

Custom API endpoint URL. Required for ollama and custom providers.

max_tokens

integer?

—

Maximum tokens per response.

temperature

float?

—

Sampling temperature (0.0 – 2.0). Lower = more deterministic.

enabled

bool

true

Whether this provider is active.

API keys are stored securely in the OS keychain (or via environment variables), not in the config file.

omni.toml

[providers.openai]

provider_type = "openai"

default_model = "gpt-4o"

max_tokens = 4096

temperature = 0.7

[providers.anthropic]

provider_type = "anthropic"

default_model = "claude-opus-4-6"

temperature = 0.8

[providers.ollama]

provider_type = "ollama"

default_model = "llama3.1"

endpoint = "http://localhost:11434"

[agent]

Controls the AI agent's behavior — the system prompt, how many tool-use iterations it can perform, and the overall timeout.

Key

Type

Default

Description

system_prompt

string?

—

Custom system prompt prepended to every conversation.

max_iterations

integer

Max tool-use iterations per turn before the agent stops.

timeout_secs

integer

120

Maximum seconds for a single agent turn.

omni.toml

[agent]

system_prompt = "You are a helpful assistant."

max_iterations = 25

timeout_secs = 120

[guardian]

Configure the Guardian anti-injection pipeline that scans all inputs and outputs for prompt injection attacks.

Key

Type

Default

Description

enabled

bool

true

Enable or disable the Guardian pipeline.

sensitivity

string

"balanced"

Detection sensitivity: strict, balanced, or permissive.

custom_signatures

string?

—

Path to a custom regex signatures JSON file.

allow_override

bool

true

Allow users to override blocked content from the UI.

omni.toml

[guardian]

enabled = true

sensitivity = "balanced"

allow_override = true

[permissions]

Default behavior for the permission system. Individual extensions can request specific permissions through their manifests.

Key

Type

Default

Description

default_policy

string

"deny"

What happens when no rule matches: deny (block silently) or prompt (ask user).

trust_verified

bool

false

Auto-approve permissions for marketplace-verified extensions.

audit_enabled

bool

true

Log all permission decisions to the audit trail.

omni.toml

[permissions]

default_policy = "deny"

trust_verified = false

audit_enabled = true

[ui]

Customize the appearance of the desktop application. All settings here can also be changed from Settings → Appearance in the UI.

Key

Type

Default

Description

theme

string

"system"

Color theme: light, dark, or system.

accent_color

string

"#3b82f6"

Primary accent color as a hex value.

font_family

string

"system"

Font family: system, Inter, JetBrains Mono, Fira Code, Source Sans 3.

font_size

integer

Base font size in pixels.

line_height

string

"normal"

Line spacing: normal, relaxed, or loose.

ui_density

string

"comfortable"

UI spacing: compact, comfortable, or spacious.

sidebar_width

integer

250

Sidebar width in pixels.

message_style

string

"bubbles"

Chat message layout: bubbles, flat, or compact.

max_message_width

integer

Maximum message width as a percentage of the chat area.

code_theme

string

"dark"

Code block theme: light, dark, or auto.

show_timestamps

bool

false

Show timestamps on chat messages.

border_radius

integer

Corner radius in pixels for UI elements.

reduce_animations

bool

false

Disable animations for accessibility.

high_contrast

bool

false

Increase text contrast for readability.

auto_update

bool

true

Automatically check for and install updates.

[channels]

Pre-configure channel instances and bindings. Instances define which messaging platforms to connect to; bindings route incoming messages to specific extensions.

Channel Instances

Each instance is keyed by a compound ID in the format {type}:{instance_id}.

Key

Type

Default

Description

channel_type

string

required

Channel type: discord, telegram, slack, whatsapp_web, etc.

display_name

string?

—

Human-readable label shown in the UI.

auto_connect

bool

false

Connect automatically on startup.

Channel Bindings

Bindings route incoming messages from a channel instance to an extension. Use glob patterns in peer_filter and group_filter to match specific senders or groups.

omni.toml

[channels.instances."discord:production"]

channel_type = "discord"

display_name = "Main Server"

auto_connect = true

[channels.instances."telegram:alerts"]

channel_type = "telegram"

display_name = "Alert Bot"

[[channels.bindings]]

channel_instance = "discord:production"

extension_id = "com.example.support-bot"

peer_filter = "*"

group_filter = "support-*"

priority = 100

[marketplace]

Marketplace connection settings.

Key

Type

Default

Description

api_url

string

"https://omniapp.org/api/v1/marketplace"

Marketplace API endpoint. Override for self-hosted instances.

Full Example

A complete configuration file showing all sections with typical values.

omni.toml

[general]

telemetry = false

log_level = "info"

max_history = 1000

[providers.openai]

provider_type = "openai"

default_model = "gpt-4o"

max_tokens = 4096

temperature = 0.7

[providers.anthropic]

provider_type = "anthropic"

default_model = "claude-opus-4-6"

[providers.ollama]

provider_type = "ollama"

default_model = "llama3.1"

endpoint = "http://localhost:11434"

[agent]

system_prompt = "You are a helpful assistant."

max_iterations = 25

timeout_secs = 120

[guardian]

enabled = true

sensitivity = "balanced"

allow_override = true

[permissions]

default_policy = "deny"

trust_verified = false

audit_enabled = true

[ui]

theme = "system"

accent_color = "#3b82f6"

message_style = "bubbles"

[channels.instances."discord:production"]

channel_type = "discord"

display_name = "Main Server"

auto_connect = true

[[channels.bindings]]

channel_instance = "discord:production"

extension_id = "com.example.bot"

priority = 100