-
Notifications
You must be signed in to change notification settings - Fork 28
CLI Commands
All commands accept --config path/to/config.yaml (default: config.yaml in current directory).
| Command | Description |
|---|---|
alfred up |
Start all daemons in background |
alfred up --foreground |
Start attached to terminal (for dev/debug) |
alfred up --live |
Start with real-time TUI dashboard (Textual) |
alfred up --only curator,janitor |
Start only specific tools |
alfred down |
Stop all background daemons |
alfred status |
Show per-tool status overview |
alfred tui |
Launch standalone Ink TUI dashboard (requires Node.js) |
alfred up (without flags) daemonizes by re-executing itself with an internal flag. It writes a PID file to data/alfred.pid and detaches from the terminal.
alfred up --foreground stays attached and prints log output to the terminal. Useful for development and debugging.
alfred up --live starts the Live Dashboard — a Textual TUI showing a 2x2 grid of per-worker interpreted feeds with health indicators, pipeline step tracking, and LLM usage stats.
alfred tui launches the standalone Ink TUI dashboard — a Node.js-based terminal UI with sparklines, inline progress bars, pipeline visualizations, and hex color theming. It reads the same data/ files as --live but runs as a separate process. Requires Node.js 18+.
alfred down creates a sentinel file (data/alfred.stop) and sends SIGTERM. The orchestrator catches the signal and gracefully shuts down all workers.
| Command | Description |
|---|---|
alfred process |
Batch-process all inbox files |
alfred process -j 8 |
Use 8 parallel workers (default: 4) |
alfred process -n 10 |
Process only first 10 files |
alfred process --dry-run |
Preview what would be processed |
Batch processing uses a Rich progress display and runs multiple curator pipelines in parallel. Useful for initial vault population or catching up on a large inbox.
| Command | Description |
|---|---|
alfred ingest <file> |
Split ChatGPT/Anthropic JSON export into inbox files |
alfred ingest <file> --dry-run |
Preview without writing |
Supports conversation export formats from ChatGPT and Anthropic. Each conversation becomes a separate inbox file.
| Command | Description |
|---|---|
alfred curator |
Run curator daemon in foreground |
The curator also runs as part of alfred up. For standalone use, it watches the inbox and processes files as they appear.
| Command | Description |
|---|---|
alfred janitor scan |
One-shot structural scan, print report |
alfred janitor fix |
Scan + apply fixes (with agent for deep issues) |
alfred janitor watch |
Run periodic sweep daemon |
scan is read-only and fast — it just reports issues. fix actually modifies files. watch runs both light and deep sweeps on configurable intervals.
| Command | Description |
|---|---|
alfred distiller scan |
Find extraction candidates and print scores |
alfred distiller run |
Scan + extract knowledge records |
alfred distiller watch |
Run periodic extraction daemon |
scan analyzes vault records for keyword signals indicating latent knowledge. run processes the top candidates through the full extraction pipeline.
| Command | Description |
|---|---|
alfred surveyor |
Run full embed/cluster/label/write pipeline |
The surveyor runs its 4-stage pipeline once and exits. As part of alfred up, it runs as a daemon with configurable intervals.
Requires pip install alfred-vault[temporal] and a running Temporal server.
| Command | Description |
|---|---|
alfred temporal worker |
Start the workflow worker (connects to Temporal, runs workflows) |
alfred temporal run <workflow> |
Trigger a one-off workflow execution |
alfred temporal run <workflow> --params '{"key": "value"}' |
Trigger with JSON parameters |
alfred temporal run <workflow> --id my-run-id |
Trigger with a custom workflow ID |
alfred temporal schedule register <file> |
Register cron schedules from a Python file |
alfred temporal schedule list |
List registered schedules |
alfred temporal list |
List discovered workflow definitions |
The worker discovers workflows from directories configured in temporal.workflow_dirs plus bundled examples. See Kinetic Layer for details on writing workflows and schedules.
Direct CRUD operations on vault records:
| Command | Description |
|---|---|
alfred vault create <type> <name> |
Create a new record |
alfred vault read <path> |
Read a record (JSON output) |
alfred vault edit <path> |
Edit record fields |
alfred vault list [type] |
List records, optionally filtered by type |
alfred vault search --grep <pattern> |
Search vault content |
alfred vault move <old> <new> |
Move/rename a record |
alfred vault delete <path> |
Delete a record |
alfred vault context |
Print vault summary (used by agents) |
alfred vault create project "My Project" \
--set status=active \
--set 'client="[[org/Acme Corp]]"' \
--body-stdin < body.md-
--set key=value— set frontmatter fields -
--body-stdin— read body content from stdin - Without
--body-stdin, the template body (with base-view embeds) is used automatically
alfred vault edit project/My\ Project.md \
--set status=completed \
--append related='[[project/Other]]' \
--body-append "## New Section\nContent here"-
--set key=value— overwrite a field -
--append key=value— append to a list field -
--body-append text— append text to the body
Run external commands with vault environment variables injected:
alfred exec -- <command> # injects ALFRED_VAULT_PATH
alfred exec --scope curator -- <cmd> # also sets ALFRED_VAULT_SCOPEEnvironment variables injected:
-
ALFRED_VAULT_PATH— absolute path to the vault -
ALFRED_VAULT_SCOPE— tool scope (restricts which operations are allowed) -
ALFRED_VAULT_SESSION— path to mutation log session file
alfred quickstartInteractive wizard that scaffolds a new vault and writes configuration. See Installation for details.
Getting Started
Architecture
Workers
Reference