github yetone/magpie v0.1.2583
magpie v0.1.2583

latest releases: v0.1.2592, v0.1.2591, v0.1.2590...
2 hours ago

Bug Fixes

  • Sessions tab token totals now include cached tokens, matching the Overview (#2591).
  • Balance meter and text now align with the core budget (#3193).
  • Custom provider balance percentages now read as "Left" or "Used," matching the window display.
  • Empty GitHub answers and truncated responses are now reported clearly (#3045).
  • MCP servers signed in within Codex now show the correct status instead of "needs sign-in" (#3508).
  • Overflow memory now keeps compacted and answered cases in one conversation and is read under lock, preventing crashes from concurrent requests (#3453).
  • Chat requests relayed to Gemini’s OpenAI-compatible API no longer include Ollama-specific options (#3097).
  • Command Code streams that end without a finish reason, and responses completed with no output or tokens, now trigger failover (#2471).

Improvements

  • Long conversations now state they apply only to Magpie models (#3440).
  • Providers that refuse a conversation as too long are now asked last instead of first.

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>

Thanks to ggbdpq, n1majne3, pc418, ye4241, hipposonCN, Davyricardo and esp32gallery, whose reports and code went into this release. Everyone who helped: contributors.

Bug Fixes

  • 会话选项卡的 token 总数现在包含缓存 token,与 Overview 一致(#2591)。
  • 余额指示器和文本现在与核心预算对齐(#3193)。
  • 自定义提供方的余额百分比现在显示为“剩余”或“已用”,与窗口显示一致。
  • 空的 GitHub 回答和被截断的响应现在会被清楚地报告(#3045)。
  • 在 Codex 中登录的 MCP 服务器现在显示正确状态,而不是“需要登录”(#3508)。
  • 溢出内存现在将已压缩和已回答的案例保留在同一个会话中,并在锁定状态下读取,防止并发请求导致崩溃(#3453)。
  • 转发到 Gemini 的 OpenAI 兼容 API 的聊天请求不再包含 Ollama 专用选项(#3097)。
  • 在没有结束原因的情况下结束的 Command Code 流,以及在没有输出或 token 的情况下完成的响应,现在会触发故障转移(#2471)。

Improvements

  • 长会话现在会声明它们仅适用于 Magpie 模型(#3440)。
  • 因对话过长而拒绝的提供方现在会被放到最后询问,而不是最先。

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。

感谢 ggbdpq、n1majne3、pc418、ye4241、hipposonCN、Davyricardo 和 esp32gallery:这次更新里有你们的反馈与代码。所有贡献者见贡献墙。


The same release, and the feed magpie updates from: https://github.com/yetone/magpie-releases/releases/tag/v0.1.2583

Don't miss a new magpie release

NewReleases is sending notifications on new releases.