@rahularya01/pi-cursor 1.4.7 → 1.4.9
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 +13 -0
- package/dist/index.js +17 -17
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -1,5 +1,18 @@
|
|
|
1
1
|
# Changelog
|
|
2
2
|
|
|
3
|
+
## [1.4.9] - 2026-08-12
|
|
4
|
+
|
|
5
|
+
### Fixed
|
|
6
|
+
|
|
7
|
+
- **Tool continuation recovery after a silent resumed bridge.** Idle-timeout, abort, and transport-retry persistence now retain the mid-pause checkpoint and pending tool-call metadata when a tool result is being resumed, preventing valid checkpoints from being discarded as stale after the 3-minute watchdog fires.
|
|
8
|
+
- Added a recovery guard that refuses to replay tool results into a checkpoint when no matching durable mid-pause snapshot exists.
|
|
9
|
+
|
|
10
|
+
## [1.4.8] - 2026-08-02
|
|
11
|
+
|
|
12
|
+
### Fixed
|
|
13
|
+
|
|
14
|
+
- **`stale_checkpoint` error after tool calls when the bridge is lost.** Root cause traced via the durable journal: `commitStoredCheckpoint` records `checkpointTurnCount = completedTurns + 1` (it includes the just-finished turn). When a tool-result recovery request arrives, `turns.length` is still the pre-tool count (one less), so `discardStaleCheckpointIfNeeded` treated the valid checkpoint as stale, cleared both the checkpoint and mid-pause metadata, and left `planRecovery` with nothing to work with. The staleness check now allows the off-by-one when `midPause` metadata confirms the request is a tool continuation for that exact turn — the checkpoint is preserved for `planRecovery` to use.
|
|
15
|
+
|
|
3
16
|
## [1.4.7] - 2026-08-02
|
|
4
17
|
|
|
5
18
|
### Fixed
|