smashimo/oh-my-pi
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(session): resume Cursor turns after HTTP/2 stream reset

Browse Source
This commit is contained in:
José Otávio Rizzatti Ferreira
2026-08-14 01:44:37 -04:00
parent ad318c7572
commit 63ac830b50
8 changed files with 268 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/non-compaction-retry-policy.md
+2 -2
View File
@@ -54,7 +54,7 @@ Current retryable categories include:
The normalized classifier recognizes the transient categories above from structured flags/status and provider-aware text patterns. Classifier refusals remain a separate typed `stopDetails` decision.
Beyond `isRetryableError(...)`, empty generic aborts may enter the same retry engine when no user, dispose, or streaming-edit-guard abort is in progress. An interrupted turn whose tool calls already have matching results can also be continued safely: the failed assistant/tool-result sequence is preserved so completed side effects are not replayed. Resolved stream stalls use the same preserve-and-continue path.
Beyond `isRetryableError(...)`, empty generic aborts may enter the same retry engine when no user, dispose, or streaming-edit-guard abort is in progress. An interrupted turn whose tool calls already have matching results can also be continued safely: the failed assistant/tool-result sequence is preserved so completed side effects are not replayed. Resolved stream stalls and HTTP/2 stream resets (`NGHTTP2_INTERNAL_ERROR`, `NGHTTP2_REFUSED_STREAM`, `HTTP2StreamReset`) use the same preserve-and-continue path. Cursor idle-stall recovery still requires the exec-resolved marker (the Connect stream may still be open); an HTTP/2 RST does not, because the stream is already dead.
Retry state is owned by `TurnRecovery`:
@@ -233,4 +233,4 @@ A new retry chain can still start later on a future retryable error after counte
- Retry strips the failing assistant error from **runtime context** before re-continue, but session history still keeps that error entry.
- `RpcSessionState` currently exposes `autoCompactionEnabled` but not an `autoRetryEnabled` field; RPC callers must track their own toggle state or query settings through other APIs.
- Model fallback changes append temporary `model_change` entries and may later restore the primary model when its cooldown expires, depending on `retry.fallbackRevertPolicy`.
- Usage-aware fallback runs before a provider request when both `retry.modelFallback` and `retry.usageAwareFallback` are enabled. Unknown/unmapped usage fails open. At the reserve threshold, `"confirm"` asks interactive sessions and keeps the current model when declined; sessions without a confirmation UI automatically apply an eligible configured fallback. `"auto"` applies an eligible fallback without asking. `"fail-closed"` rejects reserve or depleted usage instead of spending it or selecting a fallback. Depleted usage under the other policies applies an eligible fallback without a reserve confirmation.
- Usage-aware fallback runs before a provider request when both `retry.modelFallback` and `retry.usageAwareFallback` are enabled. Unknown/unmapped usage fails open. At the reserve threshold, `"confirm"` asks interactive sessions and keeps the current model when declined; sessions without a confirmation UI automatically apply an eligible configured fallback. `"auto"` applies an eligible fallback without asking. `"fail-closed"` rejects reserve or depleted usage instead of spending it or selecting a fallback. Depleted usage under the other policies applies an eligible fallback without a reserve confirmation.