YOLO

1.5.1.8 🌳 Conversation Branching & Agent Reliability

✨ New Features & Workflow Efficiency

• 🌿 Conversation Branching: Added support for Conversation Branching (History Chat Copy). You can now create copies of existing conversations to explore different prompt directions or model responses without losing the original context (#141).
• ⚡ One-click Insert: Replaced the statistics button in the message action bar with a new "One-click Insert" feature, making it much faster to bring AI-generated content into your active document (#179, #145).
• 📜 Intuitive Scrolling: Optimized the auto-follow (auto-scroll) mechanism in the sidebar chat. The scrolling behavior is now more natural and aligns better with user expectations during long generations.

🤖 Agent & Model Compatibility

• 🛡️ Gemini Tooling Fix: Resolved an issue where Gemini-based tool calls would prematurely terminate in Agent mode (#186).
• 🧠 Robust Reasoning Logic:
  • Fixed a bug where "Reasoning-only" rounds were misidentified as task completion, preventing the Agent from stopping too early.
  • Optimized the auto-loop and tool execution judgment logic to reduce empty cycles and false positives for completion.
• ⚙️ Gateway Compatibility: Improved parameter compatibility and conflict handling for OpenAI-compatible API gateways.
• 🧹 Toolchain Cleanup: Further optimized tool call stability and reduced overhead by removing legacy file operation tools.

🎨 UI/UX Refinements

• ✨ Visual Alignment: Refined the styling of the "Reasoning Content" block. The left vertical line is now perfectly aligned with the centerline of the theme-colored dot for a more polished look.
• 🧹 Selection Fix: Fixed a bug where selection highlights would occasionally fail to clear in certain edge cases.

1.5.1.8 🌳 会话分支与 Agent 逻辑强化

✨ 新功能与效率提升

• 🌿 对话分支功能：新增对话分支（历史聊天副本）功能。现在支持基于现有对话创建副本，方便用户在保留原始上下文的同时，探索不同的提示词方向或模型路径 (#141)。
• ⚡ 一键插入优化：移除了消息底部的统计按钮，新增 “一键插入” 功能。大幅提升了将对话内容快速应用至文档的操作效率 (#179, #145)。
• 📜 智能滚动跟随：优化了侧边栏对话的自动跟随机制，使流式生成时的页面滚动交互更符合直觉。

🤖 Agent 与模型兼容性

• 🛡️ Gemini 渠道修复：解决了 Gemini 在 Agent 模式下工具调用可能提前终止的问题 (#186)。
• 🧠 推理判定逻辑优化：
  • 修复了仅包含推理内容（Reasoning-only）的轮次被误判为任务完成导致提早结束的问题。
  • 优化了 Agent 自动循环与工具执行的判定逻辑，降低了空轮次误判为任务完成的概率。
• ⚙️ 网关兼容性增强：完善了对 OpenAI-compatible 网关参数的兼容性逻辑，并优化了参数冲突的处理机制。
• 🧹 工具链瘦身：进一步优化了工具调用的稳定性与资源消耗，移除了已废弃的旧版文件操作工具。

🎨 UI/UX 细节优化

• ✨ 视觉对齐打磨：优化了推理内容展示样式，将左侧垂直引导线与主题色圆点的中心线对齐，提升视觉精致度。
• 🧹 选区高亮修复：修复了在特定场景下选区高亮样式无法被正常清除的残留问题。