Merge dev → main: recapture body-fetch fix + diagnostics (#842) #104

Merged
bvandeusen merged 2 commits from dev into main 2026-06-14 22:31:41 -04:00

2 Commits

Author SHA1 Message Date
bvandeusen 3df191e255 fix(patreon): full-fetch fallback when sparse fieldset returns null content (#842)
CI / lint (push) Successful in 3s
CI / frontend-build (push) Successful in 25s
CI / backend-lint-and-test (push) Successful in 1m0s
CI / integration (push) Successful in 3m27s
Operator-flagged: 9 StickySpoodge posts had empty bodies in FC despite the body
plainly existing + being accessible (creds refresh didn't help). All 9 are
body-only / poll / embed / announcement posts with no downloadable gallery
media — Patreon's detail endpoint returns content:null for these under the
sparse fields[post]=content request even though the body exists.

fetch_post_detail_content now re-fetches the FULL post resource once when the
sparse request comes back empty: recovers the body when the sparse fieldset was
the cause, and logs post_type when even the full resource is empty (body lives
elsewhere). Only the empty cases pay the extra GET; the 126 already-working
posts keep the fast sparse path.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-14 22:26:56 -04:00
bvandeusen b999480db5 feat(ingest): per-post body-capture + recapture diagnostics logging
CI / lint (push) Successful in 3s
CI / frontend-build (push) Successful in 20s
CI / backend-lint-and-test (push) Successful in 33s
CI / integration (push) Successful in 3m30s
Operator-flagged: a recapture 'caught nothing' for a post and there were no
logs explaining why. Three silent spots now log, so a recapture's per-post
outcome is diagnosable (retention bounds the volume):

- patreon_client.fetch_post_detail_content: the 200-OK-but-null-content branch
  was silent — now logs 'fetched N chars' on success AND 'empty/null content
  (tier-gated or no text)' on the empty case (the most common silent miss).
- patreon_downloader.write_post_record: logs each post's FINAL body outcome
  (captured N chars / NO body) read off the memoized attrs after detail-fetch.
- ingest_core summary: appends post-record + relinked counts to the run summary
  (surfaces on the event stdout the operator already reads).
- download_service phase3: logs how many on-disk images got source_filehash
  relinked (N/total) per recapture.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-14 22:10:59 -04:00