afucontext 0.1.24 → 0.1.25

package/README.md

@@ -32,6 +32,7 @@ Supported on macOS (Apple Silicon) and Linux x64. Start a new session after inst
 
 ¹ Windows: Codex CLI, Pi and DSH are implemented but not yet tested on Windows hardware; this release is a preview, and Windows support becomes official in the next release. Run the commands in PowerShell; `npx.cmd` and the quotes around the package name are required there. Claude Code on Windows is not supported yet.
 
+ See [Installation notes](https://code.alipay.com/med_ai/afu_context_liwu/blob/master/docs/install-notes.md) for verification scope, versions and mirrors, startup after reboot, Pi / DSH options and known issues.
 
 <a id="results"></a>
 
@@ -144,8 +145,9 @@ Run `afuctx status`, or the status command inside a host session (see [Commands]
 - **Session detail**: the context drop after each compaction, how long each one took and what it saved, and cache hit rates. A conversation the host split into segments (sub-agents, title requests, its own compaction) lists them here.
 - **Status**: integrations, versions and issues.
 
-Each compaction is recorded in three steps: **request observed → summary ready → summary applied**. For Claude Code and Codex, "applied" means a request carrying the summary received a complete successful response; for Pi it means a model request was sent with the summary in place of the history it covers; for DSH it means the host committed the summary to session history. Summaries run in the background and never delay the conversation. Unreported cache usage is shown as unknown; savings are calibrated against the token counts the model service reports for requests sent without a summary, and like costs they are estimates, not provider bills. To estimate cost, supply per-session input, output and cache prices.
+Each compaction is recorded in three steps: **request observed → summary ready → summary applied**. For Claude Code and Codex, "applied" means a request carrying the summary received a complete successful response; for Pi it means a model request was sent with the summary in place of the history it covers; for DSH it means the host committed the summary to session history. Summaries run in the background and never delay the conversation. Unreported cache usage is shown as unknown; savings are calibrated against the token counts the model service reports for requests sent without a summary, and like costs they are estimates, not provider bills. To estimate cost, supply per-session input, output and cache prices; see [Dashboard and costs](https://code.alipay.com/med_ai/afu_context_liwu/blob/master/docs/local-dashboard.md).
 
+![AFU local dashboard overview (dark)](https://code.alipay.com/med_ai/afu_context_liwu/raw/master/docs/assets/dashboard-overview.en.jpg)
 
 *Demo data.*
 

package/README.zh-CN.md

@@ -32,6 +32,7 @@
 
 ¹ Windows：Codex CLI、Pi、DSH 已实现，但尚未在 Windows 实机测试，本版本仅供预览，下一版本正式提供。请在 PowerShell 中执行，命令里的 `npx.cmd` 和包名两侧的双引号都不能省。Claude Code 暂不支持 Windows。
 
+验证范围、版本与镜像、重启后的启动方式、Pi / DSH 参数和已知问题见 [安装说明](https://code.alipay.com/med_ai/afu_context_liwu/blob/master/docs/install-notes.zh-CN.md)。
 
 <a id="results"></a>
 
@@ -144,8 +145,9 @@ AFU 连续几次连不上模型服务时，会自动重新查找并验证一条
 - **会话详情**：上下文在每次压缩后的回落、每次压缩的耗时和节省量，以及缓存命中率。宿主把一个对话分成多段时（子代理、起标题请求、宿主自己压缩后），这里会列出各段。
 - **运行状态**：接入情况、版本和异常记录。
 
-每次压缩分三步记录：**请求已接入 → 摘要已生成 → 摘要已应用**。Claude Code 和 Codex 的“已应用”表示替换历史后的请求已收到完整成功响应；Pi 表示已有模型请求用摘要替换了它覆盖的历史；DSH 表示宿主已把摘要写入会话历史。摘要在后台生成，不影响对话。模型未返回的缓存用量显示为未知；节省量按模型服务对未压缩请求报告的实际 token 数校准，与费用一样都是估算，不是供应商账单。需要估算费用时，可按会话提供输入、输出和缓存单价。
+每次压缩分三步记录：**请求已接入 → 摘要已生成 → 摘要已应用**。Claude Code 和 Codex 的“已应用”表示替换历史后的请求已收到完整成功响应；Pi 表示已有模型请求用摘要替换了它覆盖的历史；DSH 表示宿主已把摘要写入会话历史。摘要在后台生成，不影响对话。模型未返回的缓存用量显示为未知；节省量按模型服务对未压缩请求报告的实际 token 数校准，与费用一样都是估算，不是供应商账单。需要估算费用时，可按会话提供输入、输出和缓存单价，见[面板与费用](https://code.alipay.com/med_ai/afu_context_liwu/blob/master/docs/local-dashboard.zh-CN.md)。
 
+![AFU 本地面板的总览页（暗色）](https://code.alipay.com/med_ai/afu_context_liwu/raw/master/docs/assets/dashboard-overview.jpg)
 
 *截图为演示数据。*
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "afucontext",
-  "version": "0.1.24",
+  "version": "0.1.25",
   "type": "module",
   "license": "Apache-2.0",
   "description": "Local asynchronous context compression and diagnostics",

package/release.json

@@ -1,7 +1,7 @@
 {
-  "version": "0.1.24",
+  "version": "0.1.25",
   "name": "afucontext",
-  "build": "2b1963d458412b44c9c74988d55c83a74b8e3d46425283f41ba52b1d56b3bff6",
-  "wheel": "afu_context_local-0.1.24-py3-none-any.whl",
+  "build": "4279398341d27332b588cdca520e7a5b1196be5d6004e6d5a60f437ff0f5255f",
+  "wheel": "afu_context_local-0.1.25-py3-none-any.whl",
   "prebuilt": false
 }