@canvas-js/cli

Canvas provides a command line interface for running applications inside Node.js.

This package lets you run individual application servers that join the peer-to-peer network.

Installation

To install the CLI, run:

npm install -g @canvas-js/cli

Usage

canvas <command>

Commands:
  canvas init <path>    Initialize a new application
  canvas info <path>    Show the model schema and action names in a contract
  canvas run <path>     Run a Canvas application
  canvas export <path>  Export the action log as dag-json to stdout
  canvas import <path>   Import an action log from stdin

Options:
  --version  Show version number                                       [boolean]
  --help     Show help                                                 [boolean]

Running an application

The main command is canvas run <path>. This will start a libp2p node, SQLite database, QuickJS VM for processing actions, and an HTTP API server. Use the --help flag to learn more.

Joining the libp2p mesh

In the likely case that your machine is behind a NAT layer, you have to provide the CLI with both an internal port bind a WebSocket server using the --listen option, and a public external address using the --announce option. Both of these must be formatted as multiaddrs, the generic composable network address format used by libp2p.

For example, this tells the CLI to listen on port 4444 and advertise wss://foobar.com:8000 as the public address:

$ canvas run ./myapp --listen /ip4/0.0.0.0/tcp/4444/ws --announce /dns4/foobar.com/tcp/443/wss

This assumes that you've configured your server to handle incoming secure websocket connections over TLS on port 443, do TLS termination, and proxy the connection to your internal port 4444.

A few things to note:

• You can pass as many --announce addresses as you want, but only one --listen address.
• You don't have to use TLS, but if you don't, browsers running embedded Canvas apps won't be able to talk to your peer.
• If you'd prefer to announce on a public static IP instead of a DNS name, announce on /ip4/{publicIP}/tcp/{port}/wss.
• If your DNS name has IPv6 AAAA records, you can use /dns6 in addition to /dns4.
• You can pass a --listen address without an --announce address to delegate to libp2p's autonat and identify services, which are works in progress. It's best to provide a public address if you have one.

Almost always, --listen will be of the form /ip4/0.0.0.0/tcp/${port}/ws, and --announce will be of the form /dns4/${hostname}/tcp/{port}/wss.

HTTP API

Running a Canvas app with canvas run will serve an HTTP API at http://127.0.0.1:8000/api/ by default. You can change the port with the --port option.

The basic routes are:

• GET /api - get application metadata
• GET /api/models/:model - get model records
• GET /api/models/:model/:key - get a model record by primary key
• GET /api/clock - get the next logical clock value from the log
• GET /api/messages - query ranges of log messages
• GET /api/messages/:id - get a message from the log
• POST /api/messages - apply a signed message
• GET /api/connections - current libp2p connections
• GET /api/peers - current libp2p pubsub peers
• POST /api/ping/:peerId - ping a peer via libp2p

Exposing optional endpoints

Some optional API endpoints are disabled by default, since they should't be exposed publicly.

• --metrics - serve Prometheus metrics at /metrics

The metrics reported to Prometheus include default NodeJS metric, internal libp2p metrics, and some additional metrics specific to Canvas Core.

canvas_sync_time

A histogram of MST sync times.

label name	type
topic	string
duration	number
peer	string

canvas_messages

A counter of messages applied

label name	type	description
topic	string
type	string	"action" or "session"

Serving static content alonside the API

--static [directory] can be used to serve a static directory alongside the application API. This is the easiest way to bundle a frontend that uses Canvas as a backend. If the --static flag is provided, the root path / serves the files in [directory].