Skip to content

[Feature]: Support seamless recovery / egress rotation when public browsing or stream requests hit IP-level bot checks #428

Open

Description

@Aquarius-Situla

Type of request

Enhancement (improve existing UI/UX)

What problem would this solve?

GluetunEgressRotation and ProviderSessionRenewal already help a lot when YouTube playback sessions are put on hold. However, when running TypeType behind a rotating proxy/VPN (such as Gluetun or WARP), I still run into two UX gaps when an outbound IP gets flagged (429 / bot check):

  1. Public browsing paths don't recover automatically: Endpoints backed by PublicExtractionCache / PublicCacheBackoff (such as home/subscription feed refreshes, channels, search, and playlists) as well as non-SABR-fallback stream paths (such as /streams/audio-only) do not notify egress recovery when an IP is blocked on page load. If no video playback is active yet, browsing feeds keeps failing or serving stale/backoff responses (30s–600s in PublicCacheBackoff).
  2. Users have to refresh manually after an IP change: Even when the outbound IP rotates quickly, the in-flight request that triggered the block still returns provider_access_blocked / backoff immediately, and cached failure entries in StreamFailureMemory or PublicCacheBackoff can stay sticky. As a result, I often have to refresh the page several times—or back out and re-open playback—before the new IP starts working.

What would you like to see?

I would love a cleaner, more seamless way for TypeType-Server to recover from IP-level blocks—ideally making outbound IP switching largely invisible to the end user across both browsing and playback.

Depending on what fits best with your architecture (for example, extending GluetunEgressRotation or adding an optional generic rotation hook), some ideas that could help:

  1. Cover public extraction routes in block detection / recovery: Allow IP-level refusals (ProviderAccessBlocked / RateLimited) on browsing endpoints like /channel, /trending, /search, and /playlist (under cooldown / single-flight guards so concurrent feed requests don't cause a storm) to trigger egress rotation or session recovery as well.
  2. Optional generic egress-rotation hook + transparent single retry: When an operator configures an egress rotation endpoint/command, allow a request that encounters an IP-level refusal to wait (in-flight / single-flight) for the rotation and cache/pool reset to finish, then transparently retry once on the new IP so the user doesn't see errors or have to refresh manually.

I'm completely open to whichever design aligns best with your roadmap for GluetunEgressRotation and ProviderSessionRenewal.

Area

Backend (API / extraction)

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions