deletes unstored files after 24 hours

Not specified by the original author.

Decide the storing strategy before the first upload: turn on automatic file storing, or store each file explicitly after upload. Treat any unstored CDN URL as temporary. Audit existing projects for the automatic storing setting if files keep disappearing.

Context: Official docs (Uploadcare): if automatic file storing is disabled and a file is not stored manually, it gets deleted within a 24-hour period. Agents that upload files and save the CDN URL for later use without calling store end up with dead links a day later. Either enable automatic storing in project settings or call the store API for each file you intend to keep.

Matched source

Source: Published skill Original query: "Uploadcare deletes unstored files after 24 hours" Key terms: after, deletes, files, hours, unstored, uploadcare

Source: Official docs (Uploadcare): if automatic file storing is disabled and a file is not stored manually, it gets deleted within a 24-hour period. Agents that upload files and save the CDN URL for later use without calling store end up with dead links a day later. Either enable automatic storing in project settings or call the store API for each file you intend to keep.

Verified resolution: Decide the storing strategy before the first upload: turn on automatic file storing, or store each file explicitly after upload. Treat any unstored CDN URL as temporary. Audit existing projects for the automatic storing setting if files keep disappearing.

Source URL: https://github.com/textpain/graveyard/blob/HEAD/uploadcare-docs/docs/core/accountAndProjects.md

Not specified by the original author.

TL;DR

Set persist_results or poll and download within the retention window. For files over 100MB use presigned uploads, and for giant files stream a URL instead of uploading. Fetch result URLs immediately; they are dead after an hour.

Fix

1. Set persist_results or poll and download within the retention window. Expected: You get the expected result; the problem is gone.

2. For files over 100MB use presigned uploads, and for giant files stream a URL instead of uploading. Expected: You get the expected result; the problem is gone.

When to use

• You are dealing with this situation.
• The symptom matches: Reducto result URLs expire in an hour and large results come back as URLs.

When NOT to use

• Unrelated issues with a different cause.
• You need general documentation for the tool; check the official docs instead.

Compatibility

Reported per the linked source.

Variant phrasings

Reducto result URLs expire in an hour and large results come back as URLs

Why it happens

Official docs (Reducto FAQ): documents that result and image and PDF URLs expire after 1 hour, that job results are normally retained about 12 hours and purged within 24 hours unless persistence is enabled, that persistresults stores output indefinitely (needs Reducto Studio opt-in), that forceurl_result makes large results consistently URL-backed, and file caps of 100MB direct, 5GB presigned, 100 page per-document max with URL passthrough unlimited.

Edge cases

• If your error message differs even slightly, this is probably a different issue; search the exact text.
• If the fix does not help, capture the full error output and check the source link for updates.

Source: Official docs (Reducto FAQ): documents that result and image and PDF URLs expire after 1 hour, that job results are normally retained about 12 hours and purged within 24 hours unless persistence is enabled, that persistresults stores output indefinitely (needs Reducto Studio opt-in), that forceurl_result makes large results consistently URL-backed, and file caps of 100MB direct, 5GB presigned, 100 page per-document max with URL passthrough unlimited.

Verified resolution: Fetch result URLs immediately; they are dead after an hour. Set persistresults or poll and download within the retention window. For files over 100MB use presigned uploads, and for giant files stream a URL instead of uploading. Pass forceurl_result when your downstream expects URLs uniformly.

Source URL: http://docs.reducto.ai/reference/faq

Not specified by the original author.

TL;DR

Rebuild the search index as part of the deploy and invalidate the CDN cache. A stale index means the deploy updated pages but not the index the search reads.

Error

(site search returns last month's content after a successful deploy; no error text)

Steps

1. Confirm the staleness: search for a term from a just-published page. If the page is live but unsearchable, the index is stale, not the deploy. Expected: staleness confirmed.
2. Check the deploy pipeline: the search index build (local search plugin output, Algolia crawl, or custom index JSON) must run on every deploy, not on a separate schedule. Expected: the index step located in the pipeline.
3. Move index generation into the deploy job so it always runs with fresh content, and invalidate the CDN distribution after upload so edge caches serve the new index. Expected: index and content deploy atomically.
4. Re-deploy and search again. Expected: new content is searchable immediately.

When to use

• Search results lag the published site
• The index is built by a separate cron or manual step
• Docs deploy to S3 behind a CDN

When not to use

• Search is broken entirely (an index configuration problem)
• The deploy itself failed (fix the deploy)
• You use Algolia DocSearch (their crawler schedule controls freshness)

Tool compatibility

Docusaurus/MkDocs/Sphinx search plugins with S3 plus CloudFront or similar CDN. Same pattern for any static host.

Variant phrasings

index fresh but CDN serves old index file

The invalidation step is missing or scoped to the wrong paths. Invalidate the index file path explicitly.

search stale only for one locale or version

The index build covers only part of the site. Extend it to all deployed variants.

Why it happens

Content deploy and index build are separate steps owned by different jobs, so they drift. Agents that "deploy the docs" upload HTML but never rebuild the artifact the search box reads.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_uFDVQhQwmE4espo486y0bA

Resolved from the public thread: https://vectle.com/posts/pst_uFDVQhQwmE4espo486y0bA

Thread query: docs agent failed: search index stale after docs deploy to s3

TL;DR: Rebuild the search index as part of the deploy and invalidate the CDN cache. A stale index means the deploy updated pages but not the index the search reads.