@rahularya01/pi-cursor 1.4.32 → 1.4.33
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +7 -1
- package/dist/index.js +39 -39
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -1,6 +1,12 @@
|
|
|
1
1
|
# Changelog
|
|
2
2
|
|
|
3
|
-
## [
|
|
3
|
+
## [1.4.33] - 2026-09-08
|
|
4
|
+
|
|
5
|
+
### Fixed
|
|
6
|
+
|
|
7
|
+
- **A missing `getBlobArgs` is no longer answered as an empty blob.** Cursor treats that empty result as valid history, then aborts with Connect `internal: Error`. The provider now refuses the round-trip, invalidates the checkpoint, and fails the generation so the next turn rebuilds from Pi history. Fixes [#26](https://github.com/Rahularya01/pi-cursor/issues/26).
|
|
8
|
+
- **Connect `internal` / `unavailable` / `deadline_exceeded` / GOAWAY is no longer labeled as wire-drift.** Unknown envelope fields still accumulate on `/cursor.doctor`; they are not stapled onto a transport abort.
|
|
9
|
+
- **Tool continuation after a stale checkpoint no longer dies with `skipReason=stale_checkpoint`.** `discardStaleCheckpointIfNeeded` clears mid-pause metadata along with the checkpoint, and a `synthesized_after_idle` rebuild keys the next pause to rewritten wire history. Rebuild then required that snapshot and hard-skipped even when Pi's in-flight tool ids matched the results. Full-history rebuild now pins to the current request's in-flight turn when the snapshot was wiped. A recorded pending-exec list is still required to be covered, even if its fingerprint no longer matches. Fixes [#28](https://github.com/Rahularya01/pi-cursor/issues/28).
|
|
4
10
|
|
|
5
11
|
## [1.4.32] - 2026-09-06
|
|
6
12
|
|