Rapid Power-Up Development

Developing Rapid Power-Ups allows Synapse power users to extend the capabilities of the Storm query language, provides ways to implement use-case specific commands, embed documentation, and even implement customized visual workflows in Optic, the commercial Synapse UI.

A Rapid Power-Up consists of a Storm Package which is a JSON object which defines everything used to extend the Storm language and provide additional documentation. Storm Packages can be loaded directly into your Cortex.

In this guide we will discuss the basics of Storm Package development and discuss a few best practices you can use to ensure they are secure, powerful, and easy to use.

The example acme-hello power-up discussed in this guide is included in the Synapse repository within the examples/power-ups/rapid/acme-hello folder. You can find that at Acme-Hello Example.

Anatomy of a Storm Package

A Storm Package consists of a YAML file which defines the various commands, modules, documentation, and workflows embedded within the package.

Minimal Example

As you can see in the minimal example below, the Storm Package is defined by a YAML file that gets processed and loaded into your Cortex.

acme-hello.yaml:

name: acme-hello
version: 0.0.1

synapse_version: '>=2.144.0,<3.0.0'

genopts:
    dotstorm: true # Specify that storm command/module files end with ".storm"

author:
  url: https://acme.newp
  name: ACME Explosives and Anvils

desc: Acme-Hello is a minimal example of a Rapid Power-Up.

modules:
  - name: acme.hello
  - name: acme.hello.privsep
    asroot:perms:
        - [ acme, hello, user ]

commands:
  - name: acme.hello.sayhi
    descr: Print the hello message.

Note

First, a note on namespacing. To ensure your Storm Package is going to play well with other packages, it is important to choose an appropriate namespace for your power-up. In this case, the acme part of the name is meant to be replaced with your company name or an abbreviated version of it. The hello part is meant to be replaced with an indicator of the type of functionality the Storm Package contains.

Namespace now, thank yourself later.

When you define commands and modules, they will be loaded from files using the location of the Storm Package YAML file to locate their contents:

acme-hello.yaml
storm/

    modules/
        acme.hello.storm
        acme.hello.privsep.storm

    commands/
        acme.hello.sayhi.storm

storm/modules/acme.hello.storm:

function woot(text) {
    $lib.print($text)
    return($lib.null)
}

storm/commands/acme.hello.sayhi.storm:

$hello = $lib.import(acme.hello)
$hello.woot("hello storm!")

Building / Loading

To build and load Storm Packages, use the genpkg tool included within Synapse. For this example, we will assume you have deployed your Synapse environment according to the Deployment Guide:

python -m synapse.tools.genpkg acme-hello.yaml --push aha://cortex...

Note

If you added an alternate admin user or used a non-standard naming convention you may need to adjust the aha://cortex... telepath URL to connect to your Cortex.

Once your Storm Package has loaded successfully, you can use the Storm CLI to see it in action:

invisigoth@visi01:~$ python -m synapse.tools.storm aha://cortex...

Welcome to the Storm interpreter!

Local interpreter (non-storm) commands may be executed with a ! prefix:
    Use !quit to exit.
    Use !help to see local interpreter commands.

storm> acme.hello.sayhi
hello storm!
complete. 0 nodes in 1 ms (0/sec).
storm>

Storm Modules

Deploying Storm Modules allows you to author powerful library functions that you can use in automation or Storm Commands to facilitate code re-use and enforce privilege separation boundaries.

A Storm Module is specified within the modules: section of the Storm Package YAML file.

modules:

  - name: acme.hello
    modconf:
        varname: varvalu
        othervar: [1, 2, 3]

The modconf: key can be used to specify variables which will be mapped into the module’s Storm runtime and accessible using the implicit variable $modconf:

function foo() {
    $lib.print($modconf.varname)
    return((10))
}

function bar() {
    for $i in $modconf.othervar {
        // Do something using $i...
    }
}

Privileged Modules

In order to facilitate delegating permission for privileged operations, Storm modules may specify permissions which allow the module to be imported with admin privileges. It is a best-practice to declare these permissions within the Storm package using the perms: key before using them:

perms:
  - perm: [ acme, hello, user ]
    gate: cortex
    desc: Allows a user to call privileged APIs from Acme-Hello.

modules:

  - name: acme.hello.privsep
    asroot:perms:
        - [ acme, hello, user ]

To minimize risk, you must very carefully consider what functions to implement within a privileged Storm module! Privileged modules should contain the absolute minimum required functionality.

An excellent example use case for a privileged Storm module exists when you have an API key or password which you would like to use on a user’s behalf without disclosing the actual API key. The Storm library $lib.globals.set(<name>, <valu>) and $lib.globals.get(<name>) can be used to access protected global variables which regular users may not access without special permissions. By implementing a privileged Storm module which retrieves the API key and uses it on the user’s behalf without disclosing it, you may protect the API key from disclosure while also allowing users to use it. For example, acme.hello.privsep.storm:

function getFooByBar(bar) {

    // Retrieve an API key from protected storage
    $apikey = $lib.globals.get(acme:hello:apikey)

    $headers = ({
        "apikey": $apikey
    })

    $url = $lib.str.format("https://acme.newp/api/v1/foo/{bar}", bar=$bar)

    // Use the API key on the callers behalf
    $resp = $lib.inet.http.get($url, headers=$headers)
    if ($resp.code != 200) {
        $lib.warn("/api/v1/foo returned HTTP code: {code}", code=$resp.code)
        return($lib.null)
    }

    // Return the JSON response (but not the API key)
    return($resp.json())
}

Notice that the $apikey is being retrieved and used to call the HTTP API but is not returned to the caller.

Storm Commands

Adding Storm Commands to your Cortex via a Storm Package is a great way to extend the functionality of your Cortex in a CLI user-friendly way.

Command Line Options

Every Storm command has the --help option added automatically. This means that it is always safe to execute any command with --help to get a usage statement and enumerate command line arguments. The desc field specified in the command is included in the output:

storm> acme.hello.sayhi --help

Print the hello message.

Usage: acme.hello.sayhi [options]

Options:

  --help                      : Display the command usage.
complete. 0 nodes in 4 ms (0/sec).
storm>

Storm Commands may specify command line arguments using a convention which is similar (although not identical to) Python’s argparse library.

A more complex command declaration:

commands:

  - name: acme.hello.omgopts
    descr: |
        This is a multi-line description containing usage examples.

        // Run the command with some nodes
        inet:fqdn=acme.newp | acme.hello.omgopts vertex.link

        // Run the command with some command line switches
        acme.hello.omgopts --debug --hehe haha vertex.link

    cmdargs:

      - - --hehe
        - type: str
          help: The value of the hehe optional input.

      - - --debug
        - type: bool
          default: false
          action: store_true
          help: Enable debug output.

      - - fqdn
        - type: str
          help: A mandatory / positional command line argument.

A more complete example of help output:

storm> acme.hello.omgopts --help

This is a multi-line description containing usage examples.

// Run the command with some nodes
inet:fqdn=acme.newp | acme.hello.omgopts vertex.link

// Run the command with some command line switches
acme.hello.omgopts --debug --hehe haha vertex.link


Usage: acme.hello.omgopts [options] <fqdn>

Options:

  --help                      : Display the command usage.
  --hehe <hehe>               : The value of the hehe optional input.
  --debug                     : Enable debug output.

Arguments:

  <fqdn>                      : A mandatory / positional command line argument.
complete. 0 nodes in 6 ms (0/sec).

Command line options are available within the Storm command by accessing the implicit $cmdopts variable.

storm/commands/acme.hello.omgopts.storm:

// An init {} block only runs once even if there are multiple nodes in the pipeline.

init {

    // Set global debug (once) if the user specified --debug
    if $cmdopts.debug { $lib.debug = $lib.true }

    if ($cmdopts.hehe) { $lib.print("User Specified hehe: {hehe}", hehe=$cmdopts.hehe) }

    // Normalize the FQDN in case we want to send it to an external system
    ($ok, $fqdn) = $lib.trycast(inet:fqdn, $cmdopts.fqdn)
    if (not $ok) {
        $lib.exit("Invalid FQDN Specified: {fqdn}", fqdn=$cmdopts.fqdn)
    }

    // Maybe call an API here or something...
    $lib.print("FQDN: {fqdn}", fqdn=$fqdn)
}


// You may also act on nodes in the pipeline
$lib.print("GOT NODE: {repr}", repr=$node.repr())

if $lib.debug { $lib.print("debug mode detected!") }

// Any nodes still in the pipeline are sent as output

Command Option Conventions

--help

This option is reserved and handled automatically to print a command usage statement which also enumerates any positional or optional arguments.

--debug

This option is typically used to enable debug output in the Storm interpreter by setting the $lib.debug variable if it is specified. The $lib.debug variable has a recursive effect and will subsequently enable debug output in any command or functions called from the command.

--yield

By default, a command is generally expected to yield the nodes that it received as input from the pipeline. In some instances it is useful to instruct the command to yield the nodes it creates. For example, if you specify inet:fqdn nodes as input to a DNS resolver command, it may be useful to tell the command to yield the newly created inet:dns:a records rather than the input inet:fqdn nodes. Commands frequently use the divert Storm command to implement --yield functionality.

--asof <time>

To minimize duplicate API calls, many Storm packages cache results using the $lib.jsonstor API. When caching is in use, the --asof <time> option is used to control cache aging. Users may specify --asof now to disable caching.

Specifying Documentation

Documentation may be specified in the Storm Package file that will embed markdown documentation into the package. While there are not currently any CLI tools to view/use this documentation, it is presented in the Power-Ups tab in the Help Tool within the commercial Synapse User Interface.

Markdown documents may be specified for inclusion by adding a docs: section to the Storm Package YAML file:

docs:
    - title: User Guide
      path: docs/userguide.md
    - title: Admin Guide
      path: docs/adminguide.md
    - title: Changelog
      path: docs/changelog.md

Testing Storm Packages

It is highly recommended that any production Storm Packages use development “best practices” including version control and unit testing. For the acme-hello example, we have included a test that you can use as an example to expand on.

test_acme_hello.py:

import os

import synapse.tests.utils as s_test

dirname = os.path.abspath(os.path.dirname(__file__))

class AcmeHelloTest(s_test.StormPkgTest):

    assetdir = os.path.join(dirname, 'testassets')
    pkgprotos = (os.path.join(dirname, 'acme-hello.yaml'),)

    async def test_acme_hello(self):

        async with self.getTestCore() as core:

            msgs = await core.stormlist('acme.hello.sayhi')
            self.stormIsInPrint('hello storm!', msgs)
            self.stormHasNoWarnErr(msgs)

With the file test_acme_hello.py located in the same directory as acme-hello.yaml you can use the standard pytest invocation to run the test:

python -m pytest -svx test_acme_hello.py

Advanced Features

Using divert to implement --yield

The --yield option is typically used to allow a Storm command which takes nodes as input to optionally output the new nodes it added rather than the nodes it received as input. The divert command was added to Storm to simplify implementing this convention.

To implement a command with a --yield option is typically accomplished via the following pattern:

commands:

  - name: acme.hello.mayyield
    descr: |
         Take in an FQDN and make DNS A records to demo --yield

         inet:fqdn=vertex.link | acme.hello.mayyield

    cmdargs:

      - - --yield
        - default: false
          action: store_true
          help: Yield the newly created inet:dns:a records rather than the input inet:fqdn nodes.

Then within storm/commands/acme.hello.mayyield.storm:

function nodeGenrFunc(fqdn) {
    // Fake a DNS lookup and make a few inet:dns:a records...
    [ inet:dns:a=($fqdn, 1.2.3.4) ]
    [ inet:dns:a=($fqdn, 123.123.123.123) ]
}

divert --yield $cmdopts.yield $nodeGenrFunc($node)

When executed, the acme.hello.mayyield command will output the nodes received as inputs which is useful for pipelining enrichments. If the user specifies --yield the command will output the resulting inet:dns:a nodes constructed by the nodeGenrFunc() function.

Optic Actions

If you have access to the Synapse commercial UI Optic you may find it helpful to embed Optic actions within your Storm Package. These actions will be presented to users in the context-menu when they right-click on nodes within Optic.

To define Optic actions, you declare them in the Storm Package YAML file:

optic:
    actions:
      - name: Hello Omgopts
        storm: acme.hello.omgopts --debug
        descr: This description is displayed as the tooltip in the menu
        forms: [ inet:ipv4, inet:fqdn ]

By specifying the forms: key, you can control which node actions will be presented on different forms. For example, if you are writing a DNS power-up, you may want to limit the specified actions to inet:ipv4, inet:ipv6, and inet:fqdn nodes.

When selected, the query specified in the storm: key will be run with the currently selected nodes as input. For example, if you right-click on the node inet:fqdn=vertex.link and select actions -> acme-hello -> Hello Omgopts it will execute the specified query as though it were run like this:

inet:fqdn=vertex.link | acme.hello.omgopts --debug

Any printed output, including warnings, will be displayed in the Optic Console Tool.