fix(channels): close delivery feedback loop so Slack errors reach the LLM session#311
Merged
Merged
Conversation
… LLM session Slack delivery failures (timeouts, generic exceptions, permission errors) were silently lost because PostResult lacked a FailureKind and the channel adapter only reported 3 of 6 failure kinds to the session. The session never learned about the error and could not retry or inform the user, causing the agent to appear non-responsive. Changes: - Assign FailureKind to all PostResult failures (TransportFailure for timeouts, Unknown for generic exceptions) - Widen ShouldNotifySession to report all failure kinds to the session instead of only ContentRejected/MessageTooLarge/UnsupportedContent - Rethrow in NotifyDeliveryFailedAsync so broken pipelines trigger reinit instead of silently swallowing the feedback - Handle non-retryable failures in LlmSessionActor by injecting a nudge visible on the next turn (transport errors can't be fixed by changing output, but the LLM can acknowledge the issue) - Add explicit nudge guidance for TransportFailure and PermissionDenied - Record ErrorOutput post failures in _lastFailedPost for consistency
Aaronontheweb
added a commit
that referenced
this pull request
Mar 20, 2026
…econnect, and init race fix Bump VersionPrefix to 0.7.1 and update PackageReleaseNotes in Directory.Build.props. Add 0.7.1 entry to RELEASE_NOTES.md covering: - Slack delivery failures (timeouts, permission errors) now propagate back to the LLM session (#311) - netclaw stats token counts fixed for OpenAI-compatible providers (#303) - MCP OAuth servers auto-reconnect after authorization; mcp list uses daemon-side statuses for OAuth-protected servers (#301) - netclaw init no longer triggers a restart race on existing installs (#300)
Aaronontheweb
added a commit
that referenced
this pull request
Mar 20, 2026
…econnect, and init race fix (#313) Bump VersionPrefix to 0.7.1 and update PackageReleaseNotes in Directory.Build.props. Add 0.7.1 entry to RELEASE_NOTES.md covering: - Slack delivery failures (timeouts, permission errors) now propagate back to the LLM session (#311) - netclaw stats token counts fixed for OpenAI-compatible providers (#303) - MCP OAuth servers auto-reconnect after authorization; mcp list uses daemon-side statuses for OAuth-protected servers (#301) - netclaw init no longer triggers a restart race on existing installs (#300)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
PostResultlacked aFailureKindand the channel adapter only reported 3 of 6 failure kinds to the sessionD0AC6CKBK5K/1773983518.003979andD0AC6CKBK5K/1773981376.795159)SlackThreadBindingActor: missingFailureKindon timeouts/generic exceptions, selectiveShouldNotifySessiongate, and silent exception swallowing inNotifyDeliveryFailedAsyncChanges
SlackThreadBindingActor.cs:
DeliveryFailureKind.TransportFailureto timeout errors andUnknownto generic exceptions (bothSafePostAsyncandSafeUploadFileAsync)ShouldNotifySessionfrom 3 specific kinds toFailureKind is not null— session decides retryabilityelse ifbranch for non-retryable failures (now all failures have a kind)NotifyDeliveryFailedAsyncso broken pipelines triggerOutputStreamTerminated→ reinitErrorOutputpost failures in_lastFailedPostfor consistency with other output typesLlmSessionActor.cs:
AddSystemNudge()(visible on next user turn) instead of being silently droppedTransportFailure,PermissionDenied, andUnknownTest plan
Transport_failure_injects_nudge_without_triggering_retry— session receives TransportFailure, no LLM retry, nudge visible on next turnUnknown_delivery_failure_injects_nudge_without_triggering_retry— same for Unknown kindTimeout_during_post_sends_transport_failure_feedback_to_session— channel adapter correctly classifies and reports timeoutsGeneric_exception_during_post_sends_unknown_failure_feedback_to_session— channel adapter correctly classifies and reports generic exceptionsdotnet slopwatch analyze— 0 new violations