Every release file now carries a title: two to six plain words naming the change a user would remember it for. The generator refuses a release without one, promote carries the title from unreleased.md, and the unreleased template opens with a title line. The changelog-authoring skill and the format README describe how to write it. Claude-Session: https://claude.ai/code/session_01VqV56Hez25hTxXH4ipJfzH
1.7 KiB
1.7 KiB
version, date, title
| version | date | title |
|---|---|---|
| 1.9.2 | 2026-03-31 | Faster worktrees and smoother live sync |
App
New
- Worktrees/Multi-Run: added instant draft-first worktree creation and redesigned the multi-run launcher with a cleaner, faster flow for parallel runs.
- CLI/Server: added
--foregroundfor process-manager deployments, made managed server hostname configurable, and added an explicit--hostoption with safer localhost defaults (thanks to @colinmollenhour, @rapidrabbit76, @yulia-ivashko).
Improvements
- Chat/Performance: rebuilt live session sync and streaming updates to cut render churn, reduce CPU spikes, and keep long-running chats smooth and more stable across runtimes.
- VSCode/UI: polished the extension chat and sidebar with improved spacing, tooltips, a resizable sessions pane, and file-to-chat mention flows from Explorer.
- Models/Providers: improved custom provider model metadata loading and caching (thanks to @ZeppLu).
- Docker/Deployments: improved container defaults, including UID 1000 user behavior, non-fatal SSH key generation, and better localhost detection in container networking (thanks to @yulia-ivashko).
Fixes
- Web/PWA: fixed manifest behavior behind Cloudflare Access (thanks to @arthurfiorette).
VS Code
Improvements
- Chat/Performance: overhauled live sync and streaming updates to reduce re-render churn and keep long-running chats smooth in the extension.
- Sessions/UI: refined sidebar behavior with cleaner spacing, better truncation/tooltips, and a resizable sessions pane for tighter workspace control.
- Chat/Editor integration: improved Explorer file insertion.
- Reliability: startup now queues bridge and stream requests until the API is ready.
- Chat: reasoning content now renders through the markdown pipeline.