1. Run and Observe
  2. Run workflows from the CLI

ava is the local command-line front door for Avalanche.

​
Start local services

# Discover flows and serve gRPC.
uv run ava operator --flows ./flows

# Serve the browser UI for a running operator.
uv run ava web --connect localhost:7433

# Start both together.
uv run ava dev --flows ./flows

Use --ray with operator or dev when the selected workflows should use the Ray executor and Ray is installed.

​
Start a workflow run

uv run ava run my_workflow \
  --connect localhost:7433 \
  --input '{"customer_id": "cust_123"}' \
  --context '{"request_id": "req_123"}'

The command prints a run ID. Attach top-level file and workspace fields with --file FIELD=PATH and --workspace FIELD=DIR.

​
Download a terminal result

uv run ava result RUN_ID \
  --connect localhost:7433 \
  --wait \
  --timeout 300 \
  --output-dir ./run-result

--output-dir must name a directory that does not already exist. Avalanche creates it with result metadata and any returned file or workspace content.

​
Inspect local webhook routes

uv run ava webhooks list --connect localhost:7433
uv run ava webhooks get workflow_name --connect localhost:7433

​
Launch the terminal UI

uv run ava tui --connect localhost:7433

See the complete CLI reference for supported arguments, defaults, and the starter-project ava init command.