@flisk/analyze-tracking

Automatically document your analytics setup by analyzing tracking code and generating data schemas from tools like Segment, Amplitude, Mixpanel, and more 🚀

Why Use @flisk/analyze-tracking?

📊 Understand Your Tracking – Effortlessly analyze your codebase for track calls so you can see all your analytics events, properties, and triggers in one place. No more guessing what's being tracked!

🔍 Auto-Document Events – Generates a complete YAML schema that captures all events and properties, including where they're implemented in your codebase.

🕵️‍♂️ Track Changes Over Time – Easily spot unintended changes or ensure your analytics setup remains consistent across updates.

📚 Populate Data Catalogs – Automatically generate structured documentation that can help feed into your data catalog, making it easier for everyone to understand your events.

Quick Start

Run without installation! Just use:

npx @flisk/analyze-tracking /path/to/project [options]

Key Options

• -g, --generateDescription: Generate descriptions of fields (default: false)
• -p, --provider <provider>: Specify a provider (options: openai, gemini)
• -m, --model <model>: Specify a model (ex: gpt-4.1-nano, gpt-4o-mini, gemini-2.0-flash-lite-001)
• -o, --output <output_file>: Name of the output file (default: tracking-schema.yaml)
• -c, --customFunction <function_signature>: Specify the signature of your custom tracking function (see instructions here)
• --format <format>: Output format, either yaml (default) or json. If an invalid value is provided, the CLI will exit with an error.
• --stdout: Print the output to the terminal instead of writing to a file (works with both YAML and JSON)

🔑  Important: If you are using generateDescription, you must set the appropriate credentials for the LLM provider you are using as an environment variable. OpenAI uses OPENAI_API_KEY and Google Vertex AI uses GOOGLE_APPLICATION_CREDENTIALS.

Custom Functions

If you have your own in-house tracker or a wrapper function that calls other tracking libraries, you can specify the function signature with the -c or --customFunction option.

Your function signature should be in the following format:

yourCustomTrackFunctionName(EVENT_NAME, PROPERTIES, customFieldOne, customFieldTwo)

• EVENT_NAME is the name of the event you are tracking. It should be a string or a pointer to a string. This is required.
• PROPERTIES is an object of properties for that event. It should be an object / dictionary. This is optional.
• Any additional parameters are other fields you are tracking. They can be of any type. The names you provide for these parameters will be used as the property names in the output.

For example, if your function has a userId parameter at the beginning, followed by the event name and properties, you would pass in the following:

yourCustomTrackFunctionName(userId, EVENT_NAME, PROPERTIES)

If your function follows the standard format yourCustomTrackFunctionName(EVENT_NAME, PROPERTIES), you can simply pass in yourCustomTrackFunctionName to --customFunction as a shorthand.

You can also pass in multiple custom function signatures by passing in the --customFunction option multiple times or by passing in a space-separated list of function signatures.

npx @flisk/analyze-tracking /path/to/project --customFunction "yourFunc1" --customFunction "yourFunc2(userId, EVENT_NAME, PROPERTIES)"
npx @flisk/analyze-tracking /path/to/project -c "yourFunc1" "yourFunc2(userId, EVENT_NAME, PROPERTIES)"

What's Generated?

A clear YAML schema that shows where your events are tracked, their properties, and more. Here's an example:

version: 1
source:
  repository: <repository_url>
  commit: <commit_sha>
  timestamp: <commit_timestamp>
events:
  <event_name>:
    description: <ai_generated_description>
    implementations:
      - description: <ai_generated_description>
        path: <path_to_file>
        line: <line_number>
        function: <function_name>
        destination: <platform_name>
    properties:
      <property_name>:
        description: <ai_generated_description>
        type: <property_type>

Use this to understand where your events live in the code and how they're being tracked.

Your LLM of choice is used for generating descriptions of events, properties, and implementations.

See schema.json for a JSON Schema of the output.

Supported tracking libraries & languages

| Library | JavaScript/TypeScript | Python | Ruby | Go | Swift | |---------|:---------------------:|:------:|:----:|:--:|:--:| | Google Analytics | ✅ | ❌ | ❌ | ❌ | ✅ | | Google Tag Manager | ✅ | ❌ | ❌ | ❌ | ✅ | | Segment | ✅ | ✅ | ✅ | ✅ | ✅ | | Mixpanel | ✅ | ✅ | ✅ | ✅ | ✅ | | Amplitude | ✅ | ✅ | ❌ | ✅ | ✅ | | Rudderstack | ✅ | ✅ | ✳️ | ✳️ | ✅ | | mParticle | ✅ | ❌ | ❌ | ❌ | ✅ | | PostHog | ✅ | ✅ | ✅ | ✅ | ✅ | | Pendo | ✅ | ❌ | ❌ | ❌ | ✅ | | Heap | ✅ | ❌ | ❌ | ❌ | ✅ | | Snowplow | ✅ | ✅ | ✅ | ✅ | ❌ | | Datadog RUM | ✅ | ❌ | ❌ | ❌ | ❌ | | Custom Function | ✅ | ✅ | ✅ | ✅ | ✅ |

✳️ Rudderstack's SDKs often use the same format as Segment, so Rudderstack events may be detected as Segment events.

SDKs for supported libraries

JavaScript/TypeScript

gtag('event', '<event_name>', {
  '<property_name>': '<property_value>'
});

Swift

Analytics.logEvent("<event_name>", parameters: [
  "<property_name>": "<property_value>"
])

JavaScript/TypeScript

dataLayer.push({
  event: '<event_name>',
  '<property_name>': '<property_value>'
});

// Or via window
window.dataLayer.push({
  event: '<event_name>',
  '<property_name>': '<property_value>'
});

Swift

dataLayer.push(["event": "<event_name>", "<property_name>": "<property_value>"])

JavaScript/TypeScript

analytics.track('<event_name>', {
  '<property_name>': '<property_value>'
});

Python

analytics.track('<event_name>', {
  '<property_name>': '<property_value>'
})

Ruby

Analytics.track(
  event: '<event_name>',
  properties: {
    '<property_name>': '<property_value>'
  }
)

Go

client.Enqueue(analytics.Track{
  UserId: "user-id",
  Event:  "<event_name>",
  Properties: analytics.NewProperties().
    Set("<property_name>", "<property_value>"),
})

Swift

analytics.track(name: "<event_name>", properties: TrackProperties("<property_name>": "<property_value>"))

JavaScript/TypeScript

mixpanel.track('<event_name>', {
  '<property_name>': '<property_value>'
});

Python

mixpanel.track('<event_name>', {
  '<property_name>': '<property_value>'
})

Ruby

tracker.track('<distinct_id>', '<event_name>', {
  '<property_name>': '<property_value>'
})

Go

ctx := context.Background()
mp := mixpanel.NewApiClient("YOUR_PROJECT_TOKEN")
mp.Track(ctx, []*mixpanel.Event{
  mp.NewEvent("<event_name>", "", map[string]any{}{
    "<property_name>": "<property_value>",
  }),
})

Swift

Mixpanel.mainInstance().track(event: "<event_name>", properties: [
  "<property_name>": "<property_value>"
])

JavaScript/TypeScript

amplitude.track('<event_name>', {
  <event_parameters>
});

Python

client.track(
  BaseEvent(
    event_type="<event_name>",
    user_id="<user_id>",
    event_properties={
      "<property_name>": "<property_value>",
    },
  )
)

Go

client.Track(amplitude.Event{
  UserID:    "<user_id>",
  EventType: "<event_name>",
  EventProperties: map[string]any{}{
    "<property_name>": "<property_value>",
  },
})

Swift

amplitude.track(
  eventType: "<event_name>",
  eventProperties: ["<property_name>": "<property_value>"]
)

JavaScript/TypeScript

rudderanalytics.track('<event_name>', {
  <event_parameters>
});

Python

rudder_analytics.track('<event_name>', {
  '<property_name>': '<property_value>'
})

Ruby

analytics.track(
  user_id: '<user_id>',
  event: '<event_name>',
  properties: {
    '<property_name>': '<property_value>'
  }
)

Go

client.Enqueue(analytics.Track{
  UserId: "<user_id>",
  Event:  "<event_name>",
  Properties: analytics.NewProperties().
    Set("<property_name>", "<property_value>"),
})

Swift

RSClient.sharedInstance()?.track("<event_name>", properties: [
  "<property_name>": "<property_value>"
])

JavaScript/TypeScript

mParticle.logEvent('<event_name>', mParticle.EventType.<event_type>, {
  '<property_name>': '<property_value>'
});

Swift

let event = MPEvent(name: "<event_name>", type: .other)
event.customAttributes = [
  "<property_name>": "<property_value>"
]
MParticle.sharedInstance().logEvent(event)

JavaScript/TypeScript

posthog.capture('<event_name>', {
  '<property_name>': '<property_value>'
});

Python

posthog.capture('distinct_id', '<event_name>', {
  '<property_name>': '<property_value>'
})
# Or
posthog.capture(
  'distinct_id',
  event='<event_name>',
  properties={
    '<property_name>': '<property_value>'
  }
)

Ruby

posthog.capture({
  distinct_id: '<distinct_id>',
  event: '<event_name>',
  properties: {
    '<property_name>': '<property_value>'
  }
})

Go

client.Enqueue(posthog.Capture{
  DistinctId: "<distinct_id>",
  Event:      "<event_name>",
  Properties: posthog.NewProperties().
    Set("<property_name>", "<property_value>"),
})

Swift

PostHogSDK.shared.capture("<event_name>", properties: [
  "<property_name>": "<property_value>"
])

JavaScript/TypeScript

pendo.track('<event_name>', {
  <event_parameters>
});

Python

pendo.track('<event_name>', {
  '<property_name>': '<property_value>'
})

Swift

PendoManager.shared().track("<event_name>", properties: [
  "<property_name>": "<property_value>"
])

JavaScript/TypeScript

heap.track('<event_name>', {
  <event_parameters>
});

Python

heap.track('<event_name>', {
  '<property_name>': '<property_value>'
})

Swift

Heap.shared.track("<event_name>", properties: [
  "<property_name>": "<property_value>"
])

JavaScript/TypeScript

datadogRum.addAction('<event_name>', {
  '<property_name>': '<property_value>'
});

// Or via window
window.DD_RUM.addAction('<event_name>', {
  '<property_name>': '<property_value>'
});

// Or via global DD_RUM
DD_RUM.addAction('<event_name>', {
  '<property_name>': '<property_value>'
});

JavaScript/TypeScript

tracker.track(buildStructEvent({
  action: '<event_name>',
  category: '<category>',
  label: '<label>',
  property: '<property>',
  value: <value>
}));

Python

tracker.track(StructuredEvent(
  action="<event_name>",
  category="<category>",
  label="<label>",
  property_="<property>",
  value=<value>,
))

Ruby

tracker.track_struct_event(
  action: '<event_name>',
  category: '<category>',
  label: '<label>',
  property: '<property>',
  value: <value>
)

Go

tracker.TrackStructEvent(sp.StructuredEvent{
  	Action:   sp.NewString("<event_name>"),
  	Category: sp.NewString("<category>"),
  	Label:    sp.NewString("<label>"),
  	Property: sp.NewString("<property>"),
  	Value:    sp.NewFloat64(<value>),
  })

Contribute

We're actively improving this package. Found a bug? Have a feature request? Open an issue or submit a pull request!