fix: chat stream stuck thinking, dashboard UX improvements, directory…#127
Merged
Conversation
… deliverable preview - Fix SSE stream: resolve messageStream promise on 'done' event instead of waiting for HTTP close - Fix send() race condition: guard cleanup so interrupted stream doesn't kill new stream's state - Add safety net: force-end streaming if agent becomes idle while sending is still true - Dashboard: blocked tasks use amber (waiting state), only stuck-blocked shown as red alert - Dashboard: agent ranking uses activity-weighted health score (idle agents ranked lower) - Dashboard: "working now" metric uses neutral blue color instead of red-green threshold - Deliverable preview: directories show file listing with navigation, render-or-reveal per file Co-authored-by: Cursor <cursoragent@cursor.com>
Existing loop detectors require exact hash matching on args/results, missing cases where an agent calls the same tool consecutively with varying arguments (e.g., different shell commands to debug a failing server). The new detector catches single-tool dominance patterns regardless of argument or result variation. Co-authored-by: Cursor <cursoragent@cursor.com>
… to in_progress When acceptTask transitions a task to completed, it cancels pending HITL approvals which resolve with approved=false, causing notifyHumanReviewer's callback to erroneously set the task back to in_progress. Guard the callback by checking the task is still in review status before acting. Co-authored-by: Cursor <cursoragent@cursor.com>
- Sidebar defaults to minimum draggable width, shows text labels when collapsed - Replace bell icon with sidebar nav notification button that opens floating panel - Remove redundant collapse button from sidebar header - Widen team chat list with Feishu-inspired dark mode styling - Unify chat list item brightness: selected state only changes background Co-authored-by: Cursor <cursoragent@cursor.com>
…eliable billing data - Fix top card label: "LLM Token (本月)" → "累计" since data is all-time, not monthly - Remove Math.max mixing of all-time agent metrics with monthly billing in /api/usage - Remove broken period-specific cost/token display from Cost Overview (billing service is in-memory only, lost on restart, showing ~1.6M vs actual 241M tokens) - Hide Storage card when 0 B (billing never records storage events) - Use actual model catalog pricing (e.g. Claude Opus $5/$25, DeepSeek $0.14/$0.28) instead of hardcoded $3/$15 estimate across AgentMetricsCollector, report-service, and all 8 emitAudit call sites in agent.ts - Add getModelCost() to LLMRouter, track per-call cost with input/output token split - Surface costToday through API and frontend for accurate daily cost display - Backward-compatible: falls back to old estimate for pre-migration persisted metrics Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
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.
… deliverable preview