PostHog/posthog-js posthog-js@1.410.10

on GitHub

1.410.10

Patch Changes

• #4271 3d4e2fd Thanks @felipeatom! - Fix inline surveys rendering an empty container when a stale persisted question index (left over from a prior completion) points past the last question. When the persisted index is out of range the whole in-progress record is now discarded and the survey starts fresh, instead of clamping the index while keeping the equally-stale responses and visited indices. Restored visited indices are also filtered to valid questions so the Back button can never navigate to a non-existent question and re-empty the container.
  (2026-08-04)

• #4412 5f2b78a Thanks @TueHaulund! - fix(replay): hold fresh interaction-less session recordings until there is evidence someone cares

  A tab that loads but never sees any user interaction (prefetched pages, background tabs, in-app browser preloads) no longer ships a billable recording while it sits untouched. Like rotation-born sessions, a fresh recording epoch is held until there is evidence someone cares about it: a user interaction, an event trigger match, or an explicit override (posthog.startSessionRecording(...)) releases the hold and ships the buffer on the normal flush cadence, so released recordings are playable from the session's start. A clean unload also ships a fresh-start hold, so passive visits (reading, watching a video) are still captured exactly as before; rotation-born holds are discarded on unload as before. A held buffer that reaches the size cap is dropped to bound memory, and a later release takes a fresh full snapshot so the recording resumes playable. (2026-08-04)

• #4410 064874a Thanks @ioannisj! - Fix held rotation-born session replay buffers not flushing when a V2 event trigger matches
  (2026-08-04)

• #4343 83a9b67 Thanks @arnohillen! - Session replay no longer freezes the page re-encoding base64 images that are already small. When canvas recording is enabled, every <img> with a data: URL was synchronously redrawn and re-encoded through canvas.toDataURL during full snapshots and attribute mutations. The encode cost scales with pixel dimensions, not payload size, so a page of base64 lazy-load placeholders (measured: 18 images of 4096x3072 at ~33KB each) blocked the main thread for 7+ seconds to produce outputs that were larger than the inputs. Recompression now skips data URLs under 100KB (where it cannot save meaningful payload), keeps the original when the re-encoded output is not smaller, and memoizes by input so repeated snapshots and src-swapping mutations never pay for the same image twice. Genuinely large base64 images are still recompressed as before.
  (2026-08-04)

• #4339 f865818 Thanks @posthog! - Report privacy-aware dropped-event count, page and session context in the client rate limit warning
  (2026-08-04)

• Updated dependencies [f865818]:

  • @posthog/types@1.400.2