v1.3.2 — Release Notes

New

• Codex OAuth provider — first-party auth flow for OpenAI Codex; models are always fetched live, failed fetches surface as errors
• MiniMax M2.7 models — MiniMax-M2.7 and MiniMax-M2.7-Highspeed added to the MiniMax provider (contributed by @luckystars0612)
• Chat sidebar — multi-tab session list with status badges (Running / Approval / Queued / Error / Cancelled), search, and fork/export/delete actions
• "Don't auto-load chats" setting — skips replaying saved sessions on startup; chats are listed in the sidebar and materialized on demand
• Full GFM markdown in chat — tables with alignment, nested blockquotes and lists, task checkboxes, strikethrough, bold-italic, h5/h6, bare-URL autolinking
• Paged bulk renamer — function loading in the bulk renamer is now paged; improved tool paging and data reads in the agent

Fixes

• Abort provider streams on cancel (#48)
• Codex / session files now read as UTF-8 on Windows, fixing crashes on non-ASCII content and stale model lists
• Session restore I/O moved off the UI thread; typewriter buffer smooths bursty network at ~33fps; scroll-up intent respected during streaming
• Fixed dark-on-dark text in message cards; assistant bubble now warm taupe instead of cold grey; WCAG contrast maintained
• Fixed empty vertical band in assistant bubble caused by stale height cache
• Fixed markdown double-spacing (collapsed
  runs against block elements)
• Fixed IDA theme: native dock colors respected, eliminated Base=white flash; fixed recursion crash on palette change
• Detect Binary Ninja Python on Windows; fix IDA batch script unicode symbols (#46)
• Sync plugin version metadata across all manifests (#45)