Improvements
- Automatically condense long conversations at 272,000 tokens by default for models with larger context windows, matching the behavior of OpenAI Codex and Anthropic Claude Code. This prevents very large conversations from being resent in full each turn and reduces waiting times.
- Added a Long conversations setting with Compact at 272K and Full window options.
Bug Fixes
- Fixed delays of 70–90 seconds before the first response when using 1M-token models in long tasks, caused by resending the entire conversation every turn.
Install
Download from usemagpie.ai, or curl -fsSL https://usemagpie.ai/install.sh | sh. An installed magpie updates itself.
- macOS (Apple Silicon):
magpie-darwin-arm64.dmg - macOS (Intel):
magpie-darwin-amd64.dmg - Windows:
magpie-windows-amd64.exe(ARM:magpie-windows-arm64.exe) - Linux:
magpie-linux-amd64/magpie-linux-arm64(needs GTK 3 and WebKitGTK 4.1) - Terminal only:
magpie-cli-<os>-<arch>
改进
- 对于具有更大上下文窗口的模型,默认会在对话达到 272,000 个 token 时自动压缩长对话,与 OpenAI Codex 和 Anthropic Claude Code 的行为保持一致。这可避免每轮都重新发送超长对话,并减少等待时间。
- 新增了 Long conversations(长对话)设置,提供 Compact at 272K(在 272K 时压缩)和 Full window(完整窗口)选项。
错误修复
- 修复了在使用 1M-token 模型处理长任务时,因每轮都重新发送整个对话而导致首次响应前出现 70–90 秒延迟的问题。
安装
从 usemagpie.ai 下载,或运行 curl -fsSL https://usemagpie.ai/install.sh | sh。已安装的 magpie 会自动更新。