OConnector v1.2.0

v1.2.0

New Features

• Side panel — swipe left anywhere on screen to open a file browser and model info panel. Navigate project directories, view model/provider details, and track context window usage.
• File preview — tap any .md or .txt file in the file browser to preview its content inline without leaving the chat.
• Todo task panel — overlay panel shows active AI task progress. Badge count in toolbar. Auto-dismisses and sends notification when all tasks complete.
• Ask/Confirm bubbles — when the AI needs permission or asks a question, inline bubbles appear. Permission bubble has Allow Once / Always Allow / Reject. Question bubble has selectable options and custom text input.
• Sequential question answering — multiple questions presented one at a time with Back/Next navigation. Progress bar shows current step.
• Sequential permission confirmation — multiple permission requests queued and presented one at a time.
• Multi-select questions — questions with multiple: true allow selecting several options at once.
• Undo/Redo — undo the last user message (soft-hides + rolls back file changes). Redo restores. Undone text restored to input box.
• Per-prompt model selection — choose a different AI model for each message. Nested { providerID, modelID } selector sent with each prompt.
• Context usage display — shows current context token consumption (e.g. "32K") alongside model selector in input bar.
• Manual update check button — refresh button on connection page to trigger update check anytime.
• Update check mirror sources — GitHub + gh-proxy.com CDN mirrors checked concurrently. If any source responds, update is detected.
• Version filtering — skips pre-release versions (e.g. v1_pr1); only stable vX.Y.Z considered.
• Help page expansion — new sections for Model Selection, Context Usage, Undo/Redo, and Update Check (EN/ZH).
• Child session filtering — optional toggle to hide sub-sessions from project list.
• Comprehensive API documentation — OPENCODE_API.md documents all REST endpoints, SSE events, and data models.

Fixed

• Streaming message flicker/disappear — unified placeholder uses real message ID so LazyColumn updates in-place with zero visual discontinuity.
• Streaming stalls and blocks — watchdog timer detects and recovers stalled SSE streams. Stale state abort on re-entry.
• Model selector not persisting — fixed ModelRef nested DTO format for server compatibility.
• Context usage inaccurate — reverse scan extracts only latest assistant message's input tokens.
• Session.idle race condition — delayed cleanup with retry-based message fetch and verification.
• Update check fails in restricted networks — concurrent mirror sources with Channel racing.

Changed

• Chat input bar redesign — three-section piano-key style bar with model selector, context usage, and expandable text input.
• SSE pipeline optimization — 256 buffer, 16ms coalescing, 45s heartbeat timeout.
• Smart auto-scroll — only scrolls when user is near bottom.
• ExpandableSegment default expanded — code/text segments show full content by default.

---

Full Changelog: v1.1.2...v1.2.0