Optimize media asset bucket listing

[digital-pro]

Summary

• Align bucket prefixes to match audio/ and visual/ layout
• Limit GCS list calls to the intended prefixes and tighten path validation
• Fix recursive pagination argument order to keep filtering consistent

Issue

Listing could include unrelated paths when bucket prefixes or validation didn’t align with the current bucket structure, causing extra listing work and risk of pulling unrelated assets.

Fix

Bucket name generation and prefix selection now mirror the actual layout, and validation explicitly enforces audio//... and visual//....

Savings

Estimated list-traffic reduction when 6 tasks run:

• Prior worst-case list payload ~0.8 MB per call
• 4 list calls per task → ~24 calls → ~19 MB total list traffic
• With prefix scoping, list traffic should drop to a few MB total (depends on asset counts under each prefix)

Test plan

• Not run (logic-only change)
• Optional: run a task in dev and confirm assets resolve for audio and visual prefixes

[asengupta3]

Looks like the arguments are flipped here. @digital-pro did you run the tasks locally to check if this PR works?

[asengupta3]

I think this:

  if (assetType === 'visual') {
    return `${bucket}/${taskName}`;
  }

is unnecessary since the default return is the same as well.

[asengupta3]

So previously if the language was not present it was adding an extra / and that was causing issues?

[digital-pro]

Sorry, I posted to the -engineering channel when I found the Sentry report over the weekend that this was mostly a placeholder and that someone who knows the code better should figure out if this is the right approach to fixing it. I didn't feel competent to know, but wanted to flag that the issue wasn't the initially blamed large asset, but large object name retrievals.

[asengupta3]

@zwatson2001 do you think we can look into this?

[digital-pro]

It was asking for a list of all assets in the bucket without filtering by language, etc. So the returned list was getting flagged as it is over 800K (just for the list), and doing it multiple times. We talked about it at standup and Zach thinks he can sort it out and presumably run the fix by you. From: asengupta3 ***@***.***> Sent: Monday, February 2, 2026 9:11 AM To: levante-framework/core-tasks ***@***.***> Cc: David Cardinal ***@***.***>; Mention ***@***.***> Subject: Re: [levante-framework/core-tasks] Optimize media asset bucket listing (PR #415) @asengupta3 commented on this pull request.

________________________________ In task-launcher/src/tasks/shared/helpers/getBucketName.ts<https://urldefense.com/v3/__https:/github.com/levante-framework/core-tasks/pull/415*discussion_r2755390806__;Iw!!G92We9drHetJ8EofZw!dHsdAlLSGai08CSFWJUxgSGA3kJDc441YRnqK7qLvXYgJpztx5G5T75ekbxSV3JMQNjuM1tT7yJYrqajX6QXQn3Av9Wi$>:

@@ -10,5 +10,13 @@ export function getBucketName(

? TASK_BUCKET_NAMES_DEV[assetType as keyof typeof TASK_BUCKET_NAMES_DEV] : TASK_BUCKET_NAMES_PROD[assetType as keyof typeof TASK_BUCKET_NAMES_PROD]; - return `${bucket}/${assetType === 'audio' ? language : taskName}`; + if (assetType === 'audio') { + return language ? `${bucket}/${language}` : bucket; So previously if the language was not present it was adding an extra / and that was causing issues? — Reply to this email directly, view it on GitHub<https://urldefense.com/v3/__https:/github.com/levante-framework/core-tasks/pull/415*pullrequestreview-3740815292__;Iw!!G92We9drHetJ8EofZw!dHsdAlLSGai08CSFWJUxgSGA3kJDc441YRnqK7qLvXYgJpztx5G5T75ekbxSV3JMQNjuM1tT7yJYrqajX6QXQpRGSfry$>, or unsubscribe<https://urldefense.com/v3/__https:/github.com/notifications/unsubscribe-auth/ABJ4N3FLD77Y4MGHKO3GWM34J6AJBAVCNFSM6AAAAACTTZFUYCVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTONBQHAYTKMRZGI__;!!G92We9drHetJ8EofZw!dHsdAlLSGai08CSFWJUxgSGA3kJDc441YRnqK7qLvXYgJpztx5G5T75ekbxSV3JMQNjuM1tT7yJYrqajX6QXQm9YgSvE$>. You are receiving this because you were mentioned.Message ID: ***@***.******@***.***>>