Site Requests

Site requests and parsing

How the library works

HDRezka.NET does not launch a browser, execute website JavaScript, or inspect a browser DOM. It sends ordinary requests through HttpClient

Responses use one of two paths:

1. JSON and compact HTML fragments are read from direct website endpoints
2. Complete HTML pages are parsed in memory with AngleSharp when no smaller endpoint provides the required data

AngleSharp builds an in-memory document from the response body. This is still an ordinary HTTP client workflow and does not require Selenium, Playwright, or a browser process

Direct endpoints currently used

Feature	Request	Response
Login	POST /ajax/login/	JSON and authentication cookies
Session verification	GET /favorites/	HTML
Logout	GET /logout/	HTML and expired cookies
Fast search	POST /engine/ajax/search.php	Compact HTML fragment
Seasons and streams	POST /ajax/get_cdn_series/	JSON
Comments	GET /ajax/get_comments/	JSON containing HTML fragments
Playback progress	POST /ajax/send_save/	JSON
Continue watched state	POST /engine/ajax/cdn_saves_view.php	JSON
Continue removal	POST /engine/ajax/cdn_saves_remove.php	JSON
Bookmark changes	POST /ajax/favorites/	JSON

The player endpoint name contains series, but the website itself uses it for both series and films with different action values

The current favorites and profile pages do not expose the account-tier token. Authentication can therefore be verified from cookies while AuthenticationState.AccountTier remains Unknown. Media pages still expose the token and provide an accurate Media.AccountTier

Complete pages currently parsed

• media pages
• account settings
• continue-watching history
• bookmark folders
• catalog sections and dedicated directories
• collections
• full search results

Extended media metadata, recommendations, and series schedules are parsed from the already downloaded media page and do not create extra requests

Discovered read-only sources not yet exposed

These sources can be added when their data is needed:

• /ajax/person_info/ and /person/{id}-{name}/ for person biography and filmography
• /engine/ajax/gettrailervideo.php for trailer metadata and embed code
• /ajax/comments_likes/ for users who liked a comment
• /engine/ajax/quick_content.php for hover-card details
• /engine/ajax/get_newest_slider_content.php for the compact home-page slider
• the home-page hot episode update list
• country, year, genre, and best-rating directory pages

Quick-content and slider endpoints duplicate data already returned by catalog or media pages. Calling them for every card would create an N+1 request pattern, so they should remain explicit opt-in operations instead of enriching every catalog result automatically

Account-changing endpoints

Playback progress, continue-watching state, and bookmark changes are exposed as explicit AccountClient operations

They never run as a side effect of loading a page or resolving a stream

SavePlaybackProgressAsync sends the media and translator identifiers together with optional season, episode, current position, and duration

SetContinueWatchingWatchedAsync uses the state from a loaded ContinueWatchingEntry and only calls the website toggle when a change is needed

Media.SetBookmarkAsync uses the selected folders parsed from the media page and calls the website toggle only when the requested state differs

AccountClient.ToggleBookmarkAsync exposes the raw checkbox behavior and is intentionally named as a toggle because the same add_post action handles both states

Ratings, watched episode schedule state, comment creation, and comment likes remain unexposed until they receive their own validated contracts and tests

Performance model

No dedicated threads are required for network requests. HttpClient waits asynchronously, while bulk operations use MaxConcurrentRequests to overlap independent requests without sending an uncontrolled burst to the website

The first page of a full search determines the total page count, remaining pages load concurrently, and results are restored to page order. Translator catalogs and whole-season streams use the same limited-concurrency model