Skip to content
/ pender Public

URL parsing, archiving and rendering service for Meedan Check, a collaborative media annotation platform

License

Notifications You must be signed in to change notification settings

meedan/pender

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Pender

Test Coverage: Overview Test Coverage: Issues test

Pender is a service for link parsing, archiving and rendering. It is one of the services that supports Check, an open source platform for collaborative fact-checking and media annotation.

General Info

The url is visited, parsed and the data found is used to create a media and its attributes. The data can be obtained by API or parsing directly the HTML.

These are the specific parsers supported:

  • Twitter profiles
  • Twitter posts
  • YouTube profiles (users and channels)
  • YouTube videos
  • Facebook profiles (users and pages)
  • Facebook posts (from pages and users)
  • Instagram posts
  • Instagram profiles
  • TikTok posts
  • TikTok profiles
  • Dropbox links

Besides the specific parsers Pender can parse any link with an oEmbed endpoint or metatags.

Archivers supported

  • Archive.org
    • This archiver requires archive_org_access_key and archive_org_secret_key on config/config.yml file to be enabled. Get your account’s keys at https://archive.org/account/s3.php
  • Perma.cc
    • This archiver requires a perma_cc_key on config/config.yml file or the requesting API key to be enabled. Get your account key at https://perma.cc

Setup

To run Pender, follow these steps:

$ git clone https://github.com/meedan/pender.git
$ cd pender
$ find -name '*.example' | while read f; do cp "$f" "${f%%.example}"; done
$ docker-compose build
$ docker-compose up --abort-on-container-exit

Open http://localhost:3200/api-docs/index.html to access Pender API directly.

Setting Cookies for Requests

We send cookies with certain requests that require logged-in users (e.g. Instagram, TikTok).

In development To provide these for development, log in on your browser and copy the cookie information to config/cookies.txt. The location of this file can also be configured as cookies_file_path in config.yml

To do this easily in Chrome:

  1. Install the Get cookies.txt browser extension
  2. Log into the website (e.g. instagram.com)
  3. Using the browser extension, export cookies on the page you want to view
  4. Replace the entries in config/cookies.txt with the downloaded cookies.txt

Note: If you do install this extension, consider doing it on a limited Chrome profile since it requires read and write permission for all websites.

In deployed environments Deployed environment cookies are stored in S3. To update them, use steps 1-3 above and then update the remote file in AWS. The path to this file can be found for each environment in SSM.

API

To make requests to the API, you must set a request header with the value of the configuration option authorization_header - by default, this is X-Pender-Token. The value of that header should be the API key that you have generated using bundle exec rake lapis:api_keys:create, or any API key that was given to you.

GET /api/about

Use this method in order to get the archivers enabled on this application

Parameters

Response

200: Information about the application

{
  "type": "about",
  "data": {
    "name": "Keep",
    "version": "v0.68.0",
    "archivers": [
      {
        "key": "archive_org",
        "label": "Archive.org"
      }
    ]
  }
}

401: Access denied

{
  "type": "error",
  "data": {
    "message": "Unauthorized",
    "code": 1
  }
}

GET /api/medias.format

Get parseable data for a given URL, that can be a post or a profile, from different providers. format can be one of the following, see responses below:

  • html
  • js
  • json

Parameters

  • url: URL to be parsed/rendered (required)
  • refresh: boolean to indicate that Pender should re-fetch and re-parse the URL if it already exists in its cache (optional)
  • archivers: list of archivers to target. Possible values:
    • empty: the URL will be archived in all available archivers
    • none: the URL will not be archived
    • string with a list of archives separated by commas: the URL will be archived only on specified archivers

Request Example

curl \
-H 'X-Pender-Token: <your_token>' \
-H 'Content-type: application/json' \
http://localhost:3200/api/medias.json?url=<your_url>&refresh=1

Response

HTML

A card-representation of the URL, like the ones below:

YouTube Facebook Twitter

JavaScript

An embed code for the item, which should be called this way:

<script src="http://pender.host/api/medias.js?url=https%3A%2F%2Fwww.youtube.com%2Fchannel%2FUCEWHPFNilsT0IfQfutVzsag"></script>

JSON

200: Parsed data

{
  "type": "media",
  "data": {
    "url": "https://www.youtube.com/user/MeedanTube",
    "provider": "youtube",
    "type": "profile",
    "title": "MeedanTube",
    "description": "",
    "published_at": "2009-03-06T00:44:31.000Z",
    "picture": "https://yt3.ggpht.com/-MPd3Hrn0msk/AAAAAAAAAAI/AAAAAAAAAAA/I1ftnn68v8U/s88-c-k-no/photo.jpg",
    "username": "MeedanTube",
    "author_url": "https://www.youtube.com/user/MeedanTube",
    "author_name": "MeedanTube",
    "raw": {
      "metatags": [],
      "oembed": {},
      "api": {}
    },
    "schema": {},
    "html": "",
    "embed_tag": "<embed_tag>"
  }
}

400: URL not provided

{
  "type": "error",
  "data": {
    "message": "Parameters missing",
    "code": 2
  }
}

401: Access denied

{
  "type": "error",
  "data": {
    "message": "Unauthorized",
    "code": 1
  }
}

408: Timeout

{
  "type": "error",
  "data": {
    "message": "Timeout",
    "code": 10
  }
}

429: API limit reached

{
  "type": "error",
  "data": {
    "message": 354, // Waiting time in seconds
    "code": 11
  }
}

409: Conflict

{
  "type": "error",
  "data": {
    "message": "This URL is already being processed. Please try again in a few seconds.",
    "code": 9
  }
}

POST /api/medias

Create background jobs to parse each URL and notify the caller with the result

Parameters

  • url: URL(s) to be parsed. Can be an array of URLs, a single URL or a list of URLs separated by a commas (required)
  • refresh: Force a refresh from the URL instead of the cache. Will be applied to all URLs
  • archivers: List of archivers to target. Can be empty, none or a list of archives separated by commas. Will be applied to all URLs

Response

200: Enqueued URLs

{
  "type": "success",
  "data": {
    "enqueued": [
      "https://www.youtube.com/user/MeedanTube",
      "https://twitter.com/meedan"
    ],
    "failed": [

    ]
  }
}

401: Access denied

{
  "type": "error",
  "data": {
    "message": "Unauthorized",
    "code": 1
  }
}

DELETE|PURGE /api/medias

Clears the cache for the URL(s) passed as parameter.

Parameters

  • url: URL(s) to be deleted, either as an array or a string with one URL or multiple URLs separated by a space (required)

Response

200: Success

{
  "type": "success",
}

401: Access denied

{
  "type": "error",
  "data": {
    "message": "Unauthorized",
    "code": 1
  }
}

Webhook Notification

The archiving feature uses asynchronous events. Pender can notify your application after it sends URLs for archiving.

Pender sends the url, type and the information associated with the event. The webhook endpoint should have an associated URL (e.g., http://api:3000/api/webhooks/keep) and a token. These information should be added to API key's application_settings: api_key.application_settings = {:webhook_url=>"http://api:3000/api/webhooks/keep", :webhook_token=>"somethingsecret"}

Rake tasks

There are rake tasks for a few tasks (besides Rails' default ones). Run them this way: bundle exec rake <task name>

  • test:coverage: Run all tests and calculate test coverage
  • application=<application name> lapis:api_keys:create: Create a new API key for an application
  • lapis:api_keys:delete_expired: Delete all expired keys
  • lapis:error_codes: List all error codes that this application can return
  • lapis:licenses: List the licenses of all libraries used by this project
  • lapis:client:ruby: Generate a client Ruby gem, that allows other applications to communicate and test this service
  • lapis:client:php: Generate a client PHP library, that allows other applications to communicate and test this service
  • lapis:docs: Generate the documentation for this API, including models and controllers diagrams, Swagger, API endpoints, licenses, etc.
  • lapis:docker:run: Run the application in Docker
  • lapis:docker:shell: Enter the Docker container

How to add a new parser

  • Add a new file at app/models/concerns/parser/<provider>_<type>.rb (example... provider could be facebook and type could be post or profile)
  • Include the class in the PARSERS array in app/models/media.rb
  • It should return at least published_at, username, title, description and picture
  • If type is item, it should also return the author_url and author_picture
  • The skeleton should look like this:
module Parser
  class <Provider><Type> < Base
    class << self
      def type
        '<provider>_<type>'.freeze
      end

      def patterns
        # A list of regex that tell us when we've landed on a URL for this parser, eg facebook.com
        [<list of URL patterns>]
      end

      def ignored_urls
        # Optional method to specify disallowed URLs. We generally use this to detect
        # when we've been redirected to a dead end, like a login page.
        #
        # Should return an array in format:
        # [
        #   {
        #     pattern: /^https:\/\/www\.instagram\.com\/accounts\/login/,
        #     reason: :login_page
        #   },
        # ]
      end
    end

    private    

    def parse_data_for_parser(doc, original_url)
      # Populate `@parsed_data` with information and return parsed_data at the end of the function
      # `@parsed_data` is a hash whose key is the attribute and the value is... the value
    end

    def oembed_url(doc)
      # Optional method to define an Oembed URL, will default to looking in HTML in Parser::Base
      # Passed to OembedItem
    end
  end
end

If shared behavior is needed between parsers of the same provider, make a provider class as a concern and include it in the class. See ProviderInstagram, ProviderYoutube, ProviderFacebook, ProviderTwitter, or ProviderTiktok for examples.

How to add a new archiver

  • Add a new file at app/models/concerns/media_<name>_archiver.rb
  • Include the class in app/models/media.rb
  • It should have a method archive_to_<name>
  • It should call method Media.declare_archiver, saying the URL patterns it supports (using the only modifier) or the URL patterns it doesn't support (using the except modifier)
  • The skeleton should look like this:
module Media<Name>Archiver
  extend ActiveSupport::Concern

  included do
    Media.declare_archiver('<name>', [<list of URL patterns as regular expressions>], :only) # Or :except instead of :only
  end

  def archive_to_<name>
    # Archive and then update cache (if needed) and call webhook (if needed)
    Media.notify_webhook_and_update_cache(<name>, url, data, key_id)
  end
end

Profiling

It's possible to profile Pender in order to look for bottlenecks, slownesses, performance issues, etc. To profile a Rails application it is vital to run it using production like settings (cache classes, cache view lookups, etc.). Otherwise, Rail's dependency loading code will overwhelm any time spent in the application itself. The best way to do this is create a new Rails environment. So, follow the steps below:

  • Copy config/environments/profile.rb.example to config/environments/profile.rb
  • Make sure you have a profile environment setup on config/config.yml and config/database.yml
  • Run bundle exec rake db:migrate RAILS_ENV=profile (only needed at the first time)
  • Create an API key for the profile environment: bundle exec rake lapis:api_keys:create RAILS_ENV=profile
  • Start the server in profile mode: bundle exec rails s -e profile -p 3005
  • Make a request you want to profile using the key you created before: curl -XGET -H 'X-Pender-Token: <API key>' 'http://localhost:3005/api/medias.json?url=https://twitter.com/meedan/status/773947372527288320'
  • Check the results at tmp/profile

Everytime you make a new request, the results on tmp/profile are overwritten

We can also run performance tests. It calculates the amount of time taken to validate, instantiate and parse a link for each of the supported types/providers. In order to do that, run: bundle exec rake test:performance. It will generate a CSV at tmp/performance.csv, so that you can compare the time take for each provider.

Error reporting

We use Sentry for tracking exceptions in our application.

By default we unset sentry_dsn in the config.yml, which prevents information from being reported to Sentry. If you would like to see data reported from your local machine, set sentry_dsn to the value provided for Pender in the Sentry app.

Additional configuration:

In config.yml

  • sentry_dsn - the secret that allows us to send information to Sentry, available in the Sentry web app. Scoped to a service (e.g. Pender)
  • sentry_environment - the environment reported to Sentry (e.g. dev, QA, live)
  • sentry_traces_sample_rate - not currently used, since we don't use Sentry for tracing. Set to 0 in config as result.

In 02_sentry.rb

  • config.excluded_exceptions - a list of exception classes that we don't want to send to Sentry

Observability

We use Honeycomb for monitoring information about our application. It is currently configured to suppress Honeycomb reporting when the Open Telemetry required config is unset, which we would expect in development; however it is possible to report data from your local environment to either console or remotely to Honeycomb for troubleshooting purposes.

If you would like to see data reported from your local machine, do the following:

Local console

  1. Make sure that the otlp_exporter prefixed values are set in config.yml following config.yml.example. The values provided in config.yml.example can be used since we don't need a real API key.
  2. In lib/pender/open_telemetry_config.rb, uncomment the line setting exporter to 'console'. Warning: this is noisy!
  3. Restart the server
  4. View output in local server logs

On Honeycomb

  1. Make sure that the otlp_exporter prefixed values are set in config.yml following config.yml.example
  2. In the config key otel_exporter_otlp_headers, set x-honeycomb-team to a Honeycomb API key for the Development environment (a sandbox where we put anything). This can be found in the Honeycomb web interface. To track your own reported info, be sure to set the otel_resource_attributes.developer.name key in config.yml to your own name or unique identifier (e.g. christa). You will need this to filter information on Honeycomb.
  3. Restart the server
  4. See reported information in Development environment on Honeycomb

Configuring sampling

To enable sampling for Honeycomb, set the following configuration (either in config.yml locally, or via environment for deployments):

  • otel_traces_sampler to a supported sampler. See the Open Telemetry documentaiton for supported values.
  • otel_custom_sampling_rate to an integer value. This will be used to calculate and set OTEL_TRACES_SAMPLER_ARG (1 / <sample_rate>) and to append sampler-related value to OTEL_RESOURCE_ATTRIBUTES (as SampleRate=<sample_rate>).

Note: If sampling behavior is changed in Pender, we will also need to update the behavior to match in any other application reporting to Honeycomb. More here

URL Parameters Normalization

Some service providers include URL parameters for tracking purposes that can be safely removed. Pender parsers can define a list of such parameters to be removed during the URL normalization process.

To define URL parameters to be removed, a parser class should implement the urls_parameters_to_remove method, which returns an array of strings representing the parameters to be stripped. For example:

def urls_parameters_to_remove
  ['ighs']
end

#### Environment overrides

Often for rake tasks or background jobs, we will either want none of the data (skip reporting) or all of the data (skip sampling). For these cases we can set specific environment variables:

* To skip reporting to Honeycomb, set `PENDER_SKIP_HONEYCOMB` to `true`
* To skip sampling data we want to report to Honeycomb, set `PENDER_SKIP_HONEYCOMB_SAMPLING` to `true`

## Credits

Meedan (hello@meedan.com)