# Components Guidelines & Patterns

For a component to be accepted into the Pipedream registry, it should follow these guidelines below. These guidelines help ensure components are high quality, are intutive for both Pipedream users and component developers to use and extend.

Questions about best practices?

Join the discussion with fellow Pipedream component developers at the #contribute channel (opens new window) in Slack or on Discourse (opens new window).

# Local Checks

When submitting pull requests, the new code will run through a series of automated checks like linting the code. If you want to run those checks locally for quicker feedback you must have pnpm (opens new window) installed and run the following commands at the root of the project:

To install all the project's dependencies (only needed once):
```
pnpm install
```
To install all required dependencies:
```
npx pnpm install -r
```
To run the linter checks against your code (assuming that your changes are located at components/foo for example):
```
npx eslint components/foo
```
Optionally, you can automatically fix any linter issues by running the following command:
```
npx eslint --fix components/foo
```
Keep in mind that not all issues can be automatically fixed by the linter since they could alter the behaviour of the code.

# General

# Components should be ES modules

The Node.js community has started publishing ESM-only (opens new window) packages that do not work with CommonJS modules (opens new window). This means you must import the package. You can't use require.

You also cannot mix ESM with CJS. This will not work:

// ESM
import axios from "axios";

// CommonJS - this should be `export default`
module.exports = {
  ...
}

Therefore, all components should be written as ES modules:

import axios from "axios";

export default {
  ...
}

You'll need to use the .mjs file extension (opens new window) for any components written as ES modules.

You'll notice that many of the existing components are written as CommonJS modules. Please fix these and submit a pull request as you refactor related code. For example, if you're developing new Spotify actions, and you notice the existing event sources use CommonJS, change them to ESM:

Rename the file extension from .js to .mjs using git mv (e.g. git mv source.js source.mjs).
Change all require statements to imports.
Change instances of module.exports to export default.

# Component Scope

Create components to address specific use cases whenever possible. For example, when a user subscribes to a Github webhook to listen for “star” activity, events can be generated when users star or unstar a repository. The “New Star” source filters events for only new star activity so the user doesn’t have to.

There may be cases where it's valuable to create a generic component that provides users with broad latitude (e.g., see the custom webhook (opens new window) event source for GitHub). However, as a general heuristic, we found that tightly scoped components are easier for users to understand and use.

# Required Metadata

Registry components require a unique key and version, and a friendly name and description. Action components require a type field to be set to action (sources will require a type to be set in the future).

export default {
  key: "google_drive-new-shared-drive",
  name: "New Shared Drive",
  description: "Emits a new event any time a shared drive is created.",
  version: "0.0.1",
};

# Component Key Pattern

When publishing components to the Pipedream registry, the key must be unique across registry components and should follow the pattern:

app_name_slug-slugified-component-name

Source keys should use past tense verbs that describe the event that occurred (e.g., linear_app-issue-created-instant). For action keys, use active verbs to describe the action that will occur, (e.g., linear_app-create-issue).

# Versioning

When you first publish a component to the registry, set its version to 0.0.1.

Pipedream registry components try to follow semantic versioning (opens new window). From their site:

Given a version number MAJOR.MINOR.PATCH, increment the:

MAJOR version when you make incompatible API changes,
MINOR version when you add functionality in a backwards compatible manner, and
PATCH version when you make backwards compatible bug fixes.

When you're developing actions locally, and you've incremented the version in your account multiple times, make sure to set it to the version it should be at in the registry prior to submitting your PR. For example, when you add an action to the registry, the version should be 0.0.1. If the action was at version 0.1.0 and you've fixed a bug, change it to 0.1.1 when committing your final code.

# Folder Structure

Registry components are organized by app in the components directory of the pipedreamhq/pipedream repo.

/components
 /[app-name-slug]
  /[app-name-slug].app.js
  /actions
   /[action-name-slug]
    /[action-name-slug].js
  /sources
   /[source-name-slug]
    /[source-name-slug].js

The name of each app folder corresponds with the name slug for each app
The app file should be in the root of the app folder (e.g., /components/[app_slug]/[app_slug].app.js)
Components for each app are organized into /sources and /actions subfolders
Each component should be placed in its own subfolder (with the name of the folder and the name of the js file equivalent to the slugified component name). For example, the path for the "Search Mentions" source for Twitter is /components/twitter/sources/search-mentions/search-mentions.js.
Aside from app_slug, words in folder and file names are separated by dashes (-) (i.e., in kebab case)

You can explore examples in the components directory (opens new window).

# Using APIs vs Client Libraries

If the app has a well-supported Node.js client library, feel free to use that instead of manually constructing API requests.

# `package.json`

Each app should have a package.json in its root folder. If one doesn't exist, run npm init in the app's root folder and customize the file using this package.json (opens new window) as a template.

Each time you change the code for an app file, or change the dependencies for any app component, modify the package version.

Save any dependencies in the component app directory:

npm i --save package
npm i --save-dev package

# Error-handling and input validation

When you use the SDK of a popular API, the SDK might raise clear errors to the user. For example, if the user is asked to pass an email address, and that email address doesn't validate, the library might raise that in the error message.

But other libraries will not raise clear errors. In these cases, you may need to throw your own custom error that wraps the error from the API / lib. See the Airtable components (opens new window) for an example of custom error-handling and input validation.

In general, imagine you are a user troubleshooting an issue. Is the error easy-to-understand? If not, throw a better error.

# `README` files

New actions and sources should include README.md files within the same directory to describe how to use the action or source to users.

Here's an example README.md structure:


# Overview

# Getting Started

# Troubleshooting

These sections will appear within the correponding app, source and action page, along with any subheadings and content.

Here's an example of an app README.md within the discord component on the Pipedream registry (opens new window). That same content is rendered within the Pipedream integration page for the Discord app (opens new window).

You can add additional subheadings to each of the top level Overview, Getting Started and Troubleshooting headings:

# Overview

## Limitations

Perhaps there are some limitations about the API that users should know about.

# Getting Started

## Generating an API key

Instructions on how to generate an API key from within the service's dashboard.

# Troubleshooting

## Required OAuth scopes

Please take note, you'll need to have sufficient privileges in order to complete authentication.

TIP

Only these three top level headings Overview, Getting Starting and Troubleshooting will appear within the corresponding App Marketplace page. All other headings will be ignored.

# Pagination

When making API requests, handle pagination to ensure all data/events are processed. Moreover, if the underlying account experiences and/or generates too much data paginating through the entire collection of records, it might cause out-of-memory or timeout issues (or both!), so as a rule of thumb the pagination logic should:

Be encapsulated as a generator (opens new window) so that the component can start processing records after the very first API call. As an example, you can check the Microsoft OneDrive methods (opens new window) to list files.
Accept a "next token/page/ID" whenever possible, so that API calls do not retrieve the entire collection of records during every execution but rather from a recent point in time. The scanDeltaItems generator method in the example above follows this pattern.
Persist the last page number, token or record ID right after processing, so that following executions of the component process new records to minimize the amount of duplicate events, execution time and delayed events. Following the same Microsoft OneDrive example, check the processEvent method in this component (opens new window) for an example.

# Capturing Sensitive Data

If users are required to enter sensitive data, always use secret props.

# Promoting Reusability

# App Files

App files contain components that declare the app and include prop definitions and methods that may be reused across components. App files should adhere to the following naming convention: [app_name_slug].app.js. If an app file does not exist for your app, please reach out (opens new window).

# Prop Definitions

Whenever possible, reuse existing prop definitions.

If a prop definition does not exist and you are adding an app-specific prop that may be reused in future components, add it as a prop definition to the app file. Prop definitions will also be surfaced for apps the Pipedream marketplace.

# Methods

Whenever possible, reuse methods defined in the app file. If you need to use an API for which a method is not defined and it may be used in future components, define a new method in the app file.

Use the JS Docs (opens new window) pattern for lightweight documentation of each method in the app file. Provide a description and define @params and @returns block tags (with default values if applicable — e.g., [foo=bar]). This data will both help with reusability and will be surfaced in documentation for apps in the Pipedream marketplace. For example:

export default {
  methods: {
    /**
     * Get the most recently liked Tweets for a user
     *
     * @params {Object} opts - An object representing the configuration options
     * for this method
     * @params {String} opts.screenName - The user's Twitter screen name (e.g.,
     * `pipedream`)
     * @params {String} [opts.count=200] - The maximum number of Tweets to
     * return
     * @params {String} [opts.tweetMode=extended] - Use the default of
     * `extended` to return non-truncated Tweets
     * @returns {Array} Array of most recent Tweets liked by the specified user
     */
    async getLikedTweets(opts = {}) {
      const { screenName, count = 200, tweetMode = "extended" } = opts;
      const { data } = await this._makeRequest({
        url: "https://api.twitter.com/1.1/favorites/list.json",
        params: {
          screen_name: screenName,
          count,
          tweet_mode: tweetMode,
        },
      });
      return data;
    },
  },
};

# Testing

Pipedream does not currently support unit tests to validate that changes to app files are backwards compatible with existing components. Therefore, if you make changes to an app file that may impact other sources, you must currently test potentially impacted components to confirm their functionality is not negatively affected. We expect to support a testing framework in the future.

# Common Files (optional)

An optional pattern to improve reusability is to use a common module to abstract elements that are used across to multiple components. The trade-off with this approach is that it increases complexity for end-users who have the option of customizing the code for components within Pipedream. When using this approach, the general pattern is:

The .app.js module contains the logic related to making the actual API calls (e.g. calling axios.get, encapsulate the API URL and token, etc).
The common.js module contains logic and structure that is not specific to any single component. Its structure is equivalent to a component, except that it doesn't define attributes such as version, dedupe, key, name, etc (those are specific to each component). It defines the main logic/flow and relies on calling its methods (which might not be implemented by this component) to get any necessary data that it needs. In OOP terms, it would be the equivalent of a base abstract class.
The component module of each action would inherit/extend the common.js component by setting additional attributes (e.g. name, description, key, etc) and potentially redefining any inherited methods.

See Google Drive (opens new window) for an example of this pattern. When using this approach, prop definitions should still be maintained in the app file.

Please note that the name common is just a convention and depending on each case it might make sense to name any common module differently. For example, the AWS sources (opens new window) contains a common directory instead of a common.js file, and the directory contains several modules that are shared between different event sources.

# Props

As a general rule of thumb, we should strive to only incorporate the 3-4 most relevant options from a given API as props. This is not a hard limit, but the goal is to optimize for usability. We should aim to solve specific use cases as simply as possible.

# Labels

Use prop labels to customize the name of a prop or propDefinition (independent of the variable name in the code). The label should mirror the name users of an app are familiar with; i.e., it should mirror the equivalent label in the app’s UI. This applies to usage in labels, descriptions, etc. E.g., the Twitter API property for search keywords is “q”, but but label is set to “Search Term”.

# Descriptions

Include a description for props if it helps the user understand what they need to do. Use Markdown as appropriate to improve the clarity of the description or instructions. When using Markdown:

Enclose sample input values in backticks (`)
Refer to other props using bold by surrounding with double asterisks (*)
Use Markdown links with descriptive text rather than displaying a full URL.
If the description isn't self-explanatory, link to the API docs of the relevant method to further clarify how the prop works. When the value of the prop is complex (for example, an object with many properties), link to the section of the API docs that include details on this format. Users may pass values from previous steps using expressions, so they'll need to know how to structure the input data.

Examples:

The async option to select an Airtable Base is self-explanatory so includes no description:
The “Search Term” prop for Twitter includes a description that helps the user understand what values they can enter, with specific values highlighted using backticks and links to external content.

# Optional vs Required Props

Use optional props whenever possible to minimize the input fields required to use a component.

For example, the Twitter search mentions source only requires that a user connect their account and enter a search term. The remaining fields are optional for users who want to filter the results, but they do not require any action to activate the source:

# Default Values

Provide default values whenever possible. NOTE: the best default for a source doesn’t always map to the default recommended by the app. For example, Twitter defaults search results to an algorithm that balances recency and popularity. However, the best default for the use case on Pipedream is recency.

# Async Options

Avoid asking users to enter ID values. Use async options (with label/value definitions) so users can make selections from a drop down menu. For example, Todoist identifies projects by numeric IDs (e.g., 12345). The async option to select a project displays the name of the project as the label, so that’s the value the user sees when interacting with the source (e.g., “My Project”). The code referencing the selection receives the numeric ID (12345).

Async options should also support pagination (so users can navigate across multiple pages of options for long lists). See Hubspot (opens new window) for an example of offset-based pagination. See Twitter (opens new window) for an example of cursor-based pagination.

# Dynamic Props

Dynamic props can improve the user experience for components. They let you render props in the Pipedream dynamically, based on the value of other props, and can be used to collect more specific information that can make it easier to use the component. See the Google Sheets example in the linked component API docs.

# Interface & Service Props

In the interest of consistency, use the following naming patterns when defining interface and service props in source components:

Prop	Recommended Prop Variable Name
`$.interface.http`	`http`
`$.interface.timer`	`timer`
`$.service.db`	`db`

Use getters and setters when dealing with $.service.db to avoid potential typos and leverage encapsulation (e.g., see the Search Mentions (opens new window) event source for Twitter).

# Source Guidelines

These guidelines are specific to source development.

# Webhook vs Polling Sources

Create subscription webhooks sources (vs polling sources) whenever possible. Webhook sources receive/emit events in real-time and typically use less compute time from the user’s account. Note: In some cases, it may be appropriate to support webhook and polling sources for the same event. For example, Calendly supports subscription webhooks for their premium users, but non-premium users are limited to the REST API. A webhook source can be created to emit new Calendly events for premium users, and a polling source can be created to support similar functionality for non-premium users.

# Source Name

Source name should be a singular, title-cased name and should start with "New" (unless emits are not limited to new items). Name should not be slugified and should not include the app name. NOTE: Pipedream does not currently distinguish real-time event sources for end-users automatically. The current pattern to identify a real-time event source is to include “(Instant)” in the source name. E.g., “New Search Mention” or “New Submission (Instant)”.

# Source Description

Enter a short description that provides more detail than the name alone. Typically starts with "Emit new". E.g., “Emit new Tweets that matches your search criteria”.

# Emit a Summary

Always emit a summary for each event. For example, the summary for each new Tweet emitted by the Search Mentions source is the content of the Tweet itself.

If no sensible summary can be identified, submit the event payload in string format as the summary.

# Deduping

Use built-in deduping strategies whenever possible (unique, greatest, last) vs developing custom deduping code. Develop custom deduping code if the existing strategies do not support the requirements for a source.

# Surfacing Test Events

In order to provide users with source events that they can immediately reference when building their workflow, we should implement 2 strategies whenever possible:

# Emit Events on First Run:

Polling sources should always emit events on the first run (see the Spotify: New Playlist (opens new window) source as an example)
Webhook-based sources should attempt to fetch existing events in the deploy() hook during source creation (see the Jotform: New Submission (opens new window) source)

Note – make sure to emit the most recent events (considering pagination), and limit the count to no more than 50 events.

# Include a static sample event:

There are times where there may not be any historical events available (think about sources that emit less frequently, like "New Customer" or "New Order", etc). In these cases, we should include a static sample event so users can see the event shape and reference it while building their workflow, even if it's using fake data.

To do this,

Copy the JSON output from the source's emit (what you get from steps.trigger.event) and make sure to remove or scrub any sensitive or personal data (you can also copy this from the app's API docs)
Add a new file called test-event.mjs in the same directory as the component source and export the JSON event via export default (example (opens new window))
In the source component code, make sure to import that file as sampleEmit (example (opens new window))
And finally, export the sampleEmit object (example (opens new window))

This will render a "Generate Test Event" button in the UI for users to emit that sample event:

generate-sample-event

# Polling Sources

# Default Timer Interval

As a general heuristic, set the default timer interval to 15 minutes. However, you may set a custom interval (greater or less than 15 minutes) if appropriate for the specific source. Users may also override the default value at any time.

For polling sources in the Pipedream registry, the default polling interval is set as a global config. Individual sources can access that default within the props definition:

import { DEFAULT_POLLING_SOURCE_TIMER_INTERVAL } from "@pipedream/platform";

export default {
  props: {
    timer: {
      type: "$.interface.timer",
      default: {
        intervalSeconds: DEFAULT_POLLING_SOURCE_TIMER_INTERVAL,
      },
    },
  },
  // rest of component
}

# Rate Limit Optimization

When building a polling source, cache the most recently processed ID or timestamp using $.service.db whenever the API accepts a since_id or "since timestamp" (or equivalent). Some apps (e.g., Github) do not count requests that do not return new results against a user’s API quota.

If the service has a well-supported Node.js client library, it'll often build in retries for issues like rate limits, so using the client lib (when available) should be preferred. In the absence of that, Bottleneck (opens new window) can be useful for managing rate limits. 429s should be handled with exponential backoff (instead of just letting the error bubble up).

# Webhook Sources

# Hooks

Hooks are methods that are automatically invoked by Pipedream at different stages of the component lifecycle. Webhook subscriptions are typically created when components are instantiated or activated via the activate() hook, and deleted when components are deactivated or deleted via the deactivate() hook.

# Helper Methods

Whenever possible, create methods in the app file to manage creating and deleting webhook subscriptions.

Description	Method Name
Method to create a webhook subscription	`createHook()`
Method to delete a webhook subscription	`deleteHook()`

# Storing the 3rd Party Webhook ID

After subscribing to a webhook, save the ID for the hook returned by the 3rd party service to the $.service.db for a source using the key hookId. This ID will be referenced when managing or deleting the webhook. Note: some apps may not return a unique ID for the registered webhook (e.g., Jotform).

# Signature Validation

Subscription webhook components should always validate the incoming event signature if the source app supports it.

# Shared Secrets

If the source app supports shared secrets, implement support transparent to the end user. Generate and use a GUID for the shared secret value, save it to a $.service.db key, and use the saved value to validate incoming events.

# Action Guidelines

# Action Name

Like source name, action name should be a singular, title-cased name, should not be slugified, and should not include the app name.

As a general pattern, articles are not included in the action name. For example, instead of "Create a Post", use "Create Post".

# Use `@pipedream/platform` axios for all HTTP requests

By default, the standard axios package doesn't return useful debugging data to the user when it throws errors on HTTP 4XX and 5XX status codes. This makes it hard for the user to troubleshoot the issue.

Instead, use @pipedream/platform axios.

# Return JavaScript objects

When you return data from an action, it's exposed as a step export for users to reference in future steps of their workflow. Return JavaScript objects in all cases, unless there's a specific reason not to.

For example, some APIs return XML responses. If you return XML from the step, it's harder for users to parse and reference in future steps. Convert the XML to a JavaScript object, and return that, instead.

# "List" actions

# Return an array of objects

To simplify using results from "list"/"search" actions in future steps of a workflow, return an array of the items being listed rather than an object with a nested array. See this example for Airtable (opens new window).

# Handle pagination

For actions that return a list of items, the common use case is to retrieve all items. Handle pagination within the action to remove the complexity of needing to paginate from users. We may revisit this in the future and expose the pagination / offset params directly to the user.

In some cases, it may be appropriate to limit the number of API requests made or records returned in an action. For example, some Twitter actions optionally limit the number of API requests that are made per execution (using a maxRequests prop (opens new window)) to avoid exceeding Twitter's rate limits. See the Airtable components (opens new window) for an example of using a maxRecords prop to optionally limit the maximum number of records to return.

# Use `$.summary` to summarize what happened

Describe what happened when an action succeeds by following these guidelines:

Use plain language and provide helpful and contextually relevant information (especially the count of items)
Whenever possible, use names and titles instead of IDs
Basic structure: Successfully [action performed (like added, removed, updated)] “[relevant destination]”

# Don't export data you know will be large

Browsers can crash when users load large exports (many MBs of data). When you know the content being returned is likely to be large – e.g. files — don't export the full content. Consider writing the data to the /tmp directory and exporting a reference to the file.

← TypeScript components Troubleshooting →