Seems that regardless if the channel of the recommended video is verified or not, the verified field is always true on youtube recommendations

• should it be removed ?
• should we inject our recommendation on top of it?
• should be infer the present of "anything" and delete whatever youtube might put there?

and whatever choice we can take: should it be an switch in the popup?

As for now the youtube dom selectors we use are hard coded into the extension, so it would require a new release when YT changes some of them.

To resolve this problem we need to add an API request that returns the updated selectors and re-fetch it every hour or so.
Then, the selectors are stored in local storage and communicated from the background to content_script

Research on the method, issues and task to be done, to port the web-extension on the aforementioned platforms

As discussed with @djfm and @ascariandrea, the OGP api when receive a new URL recommendation, MIGHT save in the database empty values (null) but MUST return valid value to comply with typescript enforcing.

This open to three conditions:

1. URL don't have a title: what can we return? should we accept it? (my suggestion is that, if title do not exist, we should replace it with the bare URL)
2. URL don't have a description: but that's ok, it can be an empty string.
3. URL don't have a picture: we'll produce some default one, and would be served from youchoose.ai server.

Is it correct? thanks!

When the user adds a new recommendation the API doesn't check if there is already a recommendation with the given URL and adds it anyway.

We can change the behavior to check if there's already a recommendation with the given url in db and return it when found.

This will prevent duplicated recommendations that produces dis-aggregated stats.

Chrome: https://chrome.google.com/webstore/detail/yttrex/kbbgjcgdcibilpahljnlejefcehbljnd (unlisted)
Firefox: https://addons.mozilla.org/en-US/firefox/addon/yttrex/ (experimental)

still we should check if exist the same ID with different href

-        const id = utils.hash({
+        const metadataId = utils.hash({
             publicKey: headers.publickey,
+            randomUUID: body.randomUUID,
+            href: body.href,
+        });
+        const id = utils.hash({
+            metadataId,
             size: _.size(body.element),
             contenthash: body.contenthash,
-            randomUUID: body.randomUUID,
+            href: body.href,
             i,
         });
-        const metadataId = utils.hash({
-            publicKey: headers.publickey,
-            randomUUID: body.randomUUID,
-        });

This issue keep tracks of the API to be tested and updated for the new personal page, @lc-d

• add totalEvidences in personal API, it is necessary for the paging
• the fields observed in should contains only the fields: { videoId, id, title, author, when, relative }
• the evidence removal method: DELETE /api/v1/personal/:publicKey/evidence/:id

• export of email (similar to questionnaire)
• statistics of email and questionairre
• access to advertising by personal access token
• access to advertising by content producer
• statistics on advertising

Please, it is possible when the popup is open, any interaction in youtube window might close it?
I'm not sure if is better to bind any click in or if there is a better approach.

Searches evidence collection have been restored and enhanced. Still missing:

• make the searches API works again with the new format
• test webpages and what they offer
• remove al mentions to the 'campaigns' concept

Originally posted by @vecna in #101 (comment)

The image above shows the look of poTREX web extension, it's a nice one since it uses the pornhub.tracking.exposed color palette and since there is the Contact Us hyperlink. It would be great if we could render the yTREX pop-up with the same aesthetic style - using the yTREX color palette - and create a hyperlink to send an email to the yTREX team in case of necessity. Finally, it might worth considering adding an additional hyperlink to an Ethics page, like this one

The injected box in YT need to be styled properly following the mock up.

This is how it actually looks like in dark mode,

python3 bin/autowatcher.py config/prova.txt
Profile directory found! prova
Opening https://www.youtube.com/watch?v=qaM80BjvLuA

Video Status: 1 playing
Returning snaps/prova/prova-1-snap-1.png
Error in checkStatus Message: unknown command: session/cf810a5967bf8948f86160b0f3871db6/element/0.41869062236446575-1/screenshot

Opening https://www.youtube.com/watch?v=6McuxV0Krxw

After running python3 bin/autowatcher.py config/prova.txt
we get the following error:

" Video Status: 1 playing
Returning snaps/prova/prova-2-snap-1.png
Video Status: 1 playing
Returning snaps/prova/prova-2-snap-1.png
Error in checkStatus Message: unknown command: session/cf810a5967bf8948f86160b0f3871db6/element/0.42571308286020115-1/screenshot
Test completed: closing "

the field evidences is useless because recommendationOrder is more expressive, and has the same informative value.

the field recommendedLengthText is useless, it was there only as double check

• add interruption to your personal CSV file
• mark the belonging to a specific group

Currently the Image component uses a brute force strategy to blindly try 2 possible ways of loading an image and avoid failures due to content security policy.

If it tries the wrong way first then the image will very briefly appear as not found and errors will be seen in the console. This is especially visible and annoying in the drag and drop.

A more subtle way to proceed would be to send an OPTIONS request first and then choose the appropriate image loading strategy.

Several different patches and developments are needed for version 1.8, which implements necessary and missing features for Guardoni, privacy settings, overall step forward to scraping accuracy, and a small architectural upgrade by defining server-side objects in typescript.

In pull request #100, the three changes are to be tested and reviewed. A few of these belong to previously separated branches, but they are displaying interconnection, so now all are merged in this development branch. To be completed or reviewed:

• client-side privacy settings: by default, it must send non-link-able information. This is the setting that will initially be used on youchoose, and should be a boolean selector eventually controlled by popup.
• search and advertising scraping belonged to an experimental research that works differently from the standard. The extension now also should collect offset X and Y from the HTML elements in the DOM.
• the extension has an internal cache optimization (before a lot of useless data came into the "labels" collection and the addressing IDs were still to be understood).
• The parser and logic have been named 'leafs'. A text circulated in our internal discussion platform better explain the backstory in the scraping approach. leafs might be the first things we want to document (perhaps wirh the API v3) because the new design might speed up our ability to support new platforms if properly supported by tooling.
• the new way to keep track of experiment in guardoni, by 1) uploading a given CSV in the server and receive and experimentId 2) use this experimentId to share the same settings 3) mark you execution with experimentags.
• gradual migration / coverage of typescript in progress. A fews API (youchoose.js and the v3) uses object decode to ensure data with the client. In this branch Bluebird and nodemon have been removed as old dependencies.

This issue is meant to collect updates and suggestion on this query list.
Changes should reflect on this configuration file.

And especially, we should cache all the thumbnails in order to don't trigger multiple reload of the same picture at every click on 'Lab'

A link by mistake would force the content creator to repeat the signup process, instead should offer any form of second thought.

TODO: add a modal dialog (not an alert()) with text "Are you sure you want to unlink your channel? If you unlink your channel you will have to authenticate again. Your recommendations will remain visible on the platform." and Yes/No buttons.

It begun here c9a6fb4

use case: a reearcher want to download all the data related to a CHIARO

we need to switch dependency from puppeteer to puppeteer core because:

• chromium do now support some video codecs https://www.quora.com/How-does-the-lack-of-some-video-codecs-in-the-Chromium-web-browser-affect-its-end-users
• chrome might be found locally in computer or can be installed as external dependency for who wants to use guardoni

solution:

• we need the default installation path of MacOS, Linux, Windows, hardcoded in guardoni.js
• we need a quick test with fs.exist API to spot if any chrome path is found or not
• this would replace the nconf.get('chrome') path

1. add a trash bin icon near the list of video submitted
2. implement an API who takes the 'id' as argument and delete that entry.

• removed all the pug
• updated README
• imported stats from fbtrex
• imported mirror from fbtrex

Tue, 05 Oct 2021 15:40:00 GMT pubtimeAPI Relative time string missing? |Estreou há 4 horas|

• create a new variable in the csv with the time used for each Guardoni script (3000, end, small....).
• fix experimentnumber
• create a new variable for all the csv in yttrex, with the ReccomendedVideoCharset.
• recognize the ad before play the video: play the ad but not keep playing and stop it every few seconds.

With the prospective of MechanicalKurd

• 1) show thumbnail
• 2) a way to add a qualitative category to the csv for each video

Automated profiles should automatic disabling autoplay when watching videos

When runing npm run watch in /yttrex/backend, the server starts but shows the following warnings:

(node:55537) Warning: Accessing non-existent property 'count' of module exports inside circular dependency
(Use `node --trace-warnings ...` to show where the warning was created)
(node:55537) Warning: Accessing non-existent property 'findOne' of module exports inside circular dependency
(node:55537) Warning: Accessing non-existent property 'remove' of module exports inside circular dependency
(node:55537) Warning: Accessing non-existent property 'updateOne' of module exports inside circular dependency

Two quick fixes to not forget:

• password protected https://youchoose.tracking.exposed/adoption/
• API for new channel subscribed, which also link to tracking-exposed/YCAI#66

from https://developers.facebook.com/tools/debug/sharing/?q=https%3A%2F%2Fyoutube.tracking.exposed

but if you check backend/section/index.pug you'll see it contains all the requied tags, and by additional inspection:

unclear why the meta property aren't accepted as expected.

(In order of priority)

• Videos that have been customized should be seen in the UX
• As only the last 30 videos can be imported, crucial is to allow content creator to add new videos by specifing the URL (still missing the backend part, it should be considered a 1 day taks)
• Indicate how many independent observations and/or observed advertising is available per video
• Compare / Related API might be integrated once the right UX is found

In the configuration file, you can point to any "discussion" URL, a forum, an instant messaging channel. Whatever works fine, as long as someone makes a pull request on the configuration file containing all the query term association for the different experiments running.

Currently the query list is made by innocent English animal names:

    "monkey",
    "sloth",
    "cappuccino monkey",
    "mountain goats"

Sun, 27 Sep 2020 10:20:52 GMT yttrex:label:OVERFLOW <NOT> 21 documents Sun, 27 Sep 2020 10:20:52 GMT yttrex:searches Processed 21 entries, effective searches 0, total searches video entry 0 Sun, 27 Sep 2020 10:20:58 GMT yttrex:label:OVERFLOW <NOT> 21 documents Sun, 27 Sep 2020 10:20:58 GMT yttrex:searches [+] 21 start a new cicle, 21 took: a few seconds and now process 21 htmls Sun, 27 Sep 2020 10:20:58 GMT yttrex:searches Processed 21 entries, effective searches 0, total searches video entry 0

This cause a problem because the tool never get quieter

• unique name and/or find its uniqueness combo
• verify if videos match with the directive list, because with homepage and searches it doesn't work

• removed existing "default-opt-in-extension" branch
• fork from a stable 1.8.x
• change versioning and extension/src/chrome/background/account.js
• draft a release
• update guardoni executable to match the new default extension
• rebuild guardoni executable

The user should be able to delete everything connected to him from our DB.

API

DELETE https://youchoose.tracking.exposed/api/v3/creator/unlink -H "x-authorization: ACTKf606****"

Step to implement this feature:

• define endpoint in yytrex
• use derived API client definition to define a command
• invoke the command when the user hit the "unlink" button

• Currently the use the information fetch from open graph protocol, to define the recommendation. But they might be not the best description a creator wants. We should discuss a simple click-edit interface to allow creators in editing title/description.
• An opengraph do not have any size limit, so even description 500 chars long might be acquired. They end us breaking or ruining a bit the experience of the watcher, so we should have a char limit,
• We should also decide what to do with the additional chars (show more? force the text cutting? reduce the font ?)

Qualcosa che ho mangiato | Ben 10 Italia | Cartoon Network by Ben 10 Italiano 2 years ago 4 minutes, 28 seconds 16,258,114 views,

The regexp use numbers to spot the beginning of the date. We should pass the source name, also because it should be available OOB.