Latest release
🚀 Vibe Sensei v0.5.31
What's new
Providers — no more false "update Claude" nag, proactive reconnect
- Removed the false "Opus 5 / Update Claude CLI" nag — the provider panel no longer shows a spurious "a newer Opus model is available, update your Claude CLI" prompt that fired even when the CLI was current. The top-model currency check now compares versions correctly instead of nagging on a false positive (#1594, #1592)
- Proactive 401 reconnect MODAL — when your Claude OAuth token is expired or revoked, the app now surfaces a blocking reconnect MODAL at the points you'd hit the failure — app open, the orchestrator chat, and the New Home — instead of letting requests silently fail (#1594)
- One-click headless Claude re-auth — the reconnect flow drives a REAL re-login on an expired/revoked Claude OAuth session via headless
claude /login, so you get signed back in without leaving Vibe Sensei (VSAI-325, #1594, #1586)
Orchestrator chat — boots straight in, rich markdown
- App now boots into the orchestrator chat — instead of landing on a neutral screen, the app opens directly in the orchestrator chat so you can start delegating immediately (#1594)
- Orchestrator chat renders with the prompt UI's rich markdown — orchestrator messages now use the same rich markdown renderer as the prompt UI, so code blocks, lists, and formatting render properly in the chat (#1591)
Overview / tokens — exact reset time, idle auto-refresh
- Exact session reset time at ≥95% usage — the SESSION 5H row on the Overview now shows the precise time your session quota resets once you cross 95% usage, so you know exactly when you're unblocked (#1595)
- Idle auto-refresh for sidebar usage cards — the sidebar token-usage cards now refresh themselves every 5 minutes while idle, keeping the displayed quota current without a manual reload (#1590)
BridgeMind GAP — Phase 1
- Per-workspace
.vibe-memory store + auto build_context injection — each workspace now gets its own .vibe-memory store, and prior context is auto-injected via build_context so agents start grounded (VSAI-30, #1581)
- Visual Kanban board over todo_manager tasks — a visual Kanban board now renders the todo_manager task list, giving a drag-and-drop view of work in flight (VSAI-29, #1582)
- BridgeSwarm auto-decompose —
/swarm <goal> decomposes a goal into parallel-safe Kanban cards ready for delegation (VSAI-31, #1583)
- Command-block segment parser for the PTY — a new parser segments PTY output into command blocks (VSAI-32 GAP-4, #1584)
- External loopback TCP transport exposing
todo_manager — the embedded MCP server now exposes todo_manager over an external loopback TCP transport (VSAI-34 GAP-6, #1590)
- On-device Whisper model-size selection + checksummed download — voice input lets you pick the on-device Whisper model size, with a checksum-verified model download (VSAI-33, #1587)
- Jira create path — issueType + parent/Epic Link support — the Jira create action now supports selecting an issue type and linking a parent / Epic (#1585)
Internal / maintenance
- BridgeBench Phase 1 task-suite spec — docs for the BridgeBench Phase 1 task suite (VSAI-35, #1593)
- Competitive research: PH vibe-coding top-20 insights + Story backlog (VSAI-364)
- Marketing: publisher & social GTM strategy (VSAI-353)
Download on GitHub →