`XClIdGen.create()` fails via Couldn't get `XClientTxId` indices script error

[jonsuh]

Issue

XClIdGen.create() fails via a Couldn't get XClientTxId indices script error.

2026-07-20 14:09:11.744 | WARNING  | twscrape.queue_client:get:40 - XClIdGen creation attempt 1/3 failed: Exception: Couldn't get XClientTxId indices script
2026-07-20 14:09:14.660 | WARNING  | twscrape.queue_client:get:40 - XClIdGen creation attempt 2/3 failed: Exception: Couldn't get XClientTxId indices script
2026-07-20 14:09:17.809 | WARNING  | twscrape.queue_client:get:40 - XClIdGen creation attempt 3/3 failed: Exception: Couldn't get XClientTxId indices script

Cause

Anonymous/logged-out requests get served a web app version that no longer includes the script INDICES_FILE_RE is looking for; whereas the authenticated sessions include it—it’s needed to compute the x-client-transaction-id header.

[Y3ETCatFace]

Thank goodness someone's also spotted this exact same error I was encountering. You are my goat bro!!

[JAAVAAD]

I made the changes you suggested but it didn't work for me.@jonsuh

[fractefactos]

It works for me, thanks @jonsuh

[MantasVaistaras]

It works for me too! Great work @jonsuh 🏋️‍♀️

[Nickiller]

Confirming the root cause and that the cookie-threading approach fixes it, on stock twscrape 0.19.1 with a single authenticated account.

Symptom: every GraphQL request 404s; the fetch log repeats XClIdGen creation attempt N/3 failed: Exception: Couldn't get XClientTxId indices script, and new_fetched is 0 for all accounts. Easy to misread as cookie expiry, but the cookie is valid (GET https://x.com/ → 200 with react-root, not redirected to login).

Root cause (verified): X no longer serves the x-client-transaction indices script (sign.o-*.js / ondemand.s-*.js) to anonymous sessions. XClIdGen.create() fetches it with a cookieless client, so _find_indices_url finds nothing and raises at xclid.py:263.

Verification of the fix: backported the cookie-threading approach from #321 / gepard-me@5586aa8 as a runtime patch on 0.19.1

Thanks @jonsuh !

[jonsuh]

Issue fixed via #321 and should come with the next release