Configuration

Each vault has an optional configuration file at .typemd/config.yaml. It controls CLI defaults, TUI behavior, and AI features. The file is created automatically when you run tmd init or tmd config set.

File location

my-vault/
└── .typemd/
    └── config.yaml

Structure

The config uses YAML with top-level display settings and three namespaces:

date_format: "YYYY-MM-DD"
datetime_format: "YYYY-MM-DD HH:mm:ss"
cli:
  default_type: page
tui:
  debounce_ms: 200
  stats_type_layout: fullscreen
  toast:
    duration_ms: 3000
    dismiss_key: esc
ai:
  enabled: true
  default: claude
  providers:
    claude:
      type: cli
      model: claude-sonnet-4-6-20250627
    ollama:
      type: openai-compatible
      base_url: http://localhost:11434
      model: qwen3-coder:30b
  prompts:
    describe: "Custom prompt for descriptions"
  explore:
    sample_count: 20
    body_truncate: 1000

Display format settings

Key	Type	Default	Description
`date_format`	string	`YYYY-MM-DD`	Display format for `date` properties
`datetime_format`	string	`YYYY-MM-DD HH:mm:ss`	Display format for `datetime` properties and system timestamps (`created_at`, `updated_at`)

These settings control how dates appear in the TUI and CLI. Storage format is unchanged (ISO 8601 / RFC 3339).

Supported tokens: YYYY (year), MM (month), DD (day), HH (24-hour), mm (minute), ss (second). Unrecognized characters pass through as literals.

# European format
date_format: "DD/MM/YYYY"
datetime_format: "DD/MM/YYYY HH:mm:ss"

# Japanese style
date_format: "YYYY年MM月DD日"

CLI settings

Key	Type	Default	Description
`cli.default_type`	string	(empty)	Default type for `tmd object create` when no type argument is given

When set, you can create objects without specifying a type:

# Without default_type — type is required
tmd object create idea "My Idea"

# With cli.default_type: idea
tmd object create "My Idea"

TUI settings

Key	Type	Default	Description
`tui.debounce_ms`	int	`200`	File watcher debounce interval in milliseconds
`tui.stats_type_layout`	string	`fullscreen`	Stats type detail layout: `fullscreen` or `popup`
`tui.toast.position`	string	`bottom-right`	Toast display position: `bottom-right` or `help-bar`
`tui.toast.duration_ms`	int	`3000`	Auto-dismiss duration in milliseconds
`tui.toast.dismiss_key`	string	`esc`	Key to manually dismiss a toast notification
`tui.toast.show_warnings`	bool	`true`	Show warning-level toast notifications
`tui.toast.show_success`	bool	`false`	Show info/success-level toast notifications
`tui.theme.focus_border`	string	`63`	Focused panel border color (ANSI code or hex)
`tui.theme.wiki_link`	string	`33`	Wiki-link display text color
`tui.theme.heading`	string	`3`	Markdown heading color
`tui.theme.bold`	string	(empty)	Bold text color (empty = bold weight only)
`tui.theme.italic`	string	(empty)	Italic text color (empty = italic style only)
`tui.theme.inline_code`	string	`245`	Inline code color
`tui.theme.code_block`	string	`245`	Fenced code block color
`tui.theme.link`	string	`33`	Markdown link color
`tui.theme.blockquote`	string	`8`	Blockquote line color
`tui.theme.hrule`	string	`8`	Horizontal rule color
`tui.keybindings.<action>`	string	(per-action default)	Override a global TUI keybinding by action name. See TUI customization for the full action list and validation rules.

The file watcher monitors objects/ and types/ for changes. The debounce interval controls how quickly the TUI reacts to file modifications — lower values make updates feel more instant, higher values reduce redundant refreshes.

Toast notifications appear in the bottom-right corner of the TUI for transient messages such as sync warnings (unresolved references) and AI operation errors. They auto-dismiss after duration_ms and can be manually dismissed by pressing the dismiss_key. Error-level toasts always show regardless of configuration.

AI settings

AI features support multiple providers. The cli type uses the Claude CLI binary; the openai-compatible type works with any OpenAI-compatible API (Ollama, LM Studio, vLLM, LocalAI, etc.).

Provider configuration

Key	Type	Default	Description
`ai.enabled`	bool	`false`	Enable AI features in the TUI
`ai.default`	string	(empty)	Name of the active provider from `ai.providers`
`ai.providers.<name>.type`	string	—	Provider type: `cli` or `openai-compatible`
`ai.providers.<name>.model`	string	(empty)	Model identifier
`ai.providers.<name>.base_url`	string	(empty)	HTTP endpoint (required for `openai-compatible`)
`ai.providers.<name>.api_key`	string	(empty)	Optional Bearer token for authentication

Example with multiple providers:

ai:
  enabled: true
  default: ollama
  providers:
    claude:
      type: cli
      model: claude-sonnet-4-6-20250627
    ollama:
      type: openai-compatible
      base_url: http://localhost:11434
      model: qwen3-coder:30b
    my-server:
      type: openai-compatible
      base_url: http://192.168.1.100:8080
      model: llama3.2
      api_key: sk-my-key

Switch providers by changing ai.default — no other changes needed.

Prompt and explore settings

Key	Type	Default	Description
`ai.prompts.describe`	string	(built-in)	Custom system prompt for description generation
`ai.prompts.tag`	string	(built-in)	Custom system prompt for tag suggestions
`ai.prompts.explore`	string	(built-in)	Custom system prompt for schema exploration
`ai.explore.sample_count`	int	`10`	Number of objects sampled for schema exploration
`ai.explore.body_truncate`	int	`500`	Max characters of object body included in explore prompt

When ai.enabled is true, the TUI adds two AI keybindings:

g in object detail — open AI action picker (generate description / suggest tags)
ctrl+e from sidebar — enter schema explore mode

Reading and writing config

You can edit .typemd/config.yaml directly, use the CLI, or use the TUI settings page (press , in the TUI to open a browsable config editor):

# Set a value
tmd config set ai.enabled true

# Get a value
tmd config get cli.default_type

# List all set values
tmd config list

# List all known keys with defaults
tmd config list --all

Only known keys are accepted. Setting an unknown key returns an error listing the valid keys.

Defaults

If .typemd/config.yaml does not exist or a key is not set, TypeMD uses sensible defaults:

No default type — tmd object create requires a type argument
File watcher debounce at 200ms
Stats layout in fullscreen mode
Toast notifications: bottom-right, 3s auto-dismiss, Esc to dismiss, warnings shown, success hidden
AI features disabled
Built-in prompts for all AI operations
10 sample objects and 500 character body limit for schema exploration