From b9a160bf1c24e63e7d164094d9426b6b06cbe086 Mon Sep 17 00:00:00 2001
From: tactino <18781106300@163.com>
Date: Fri, 18 Sep 2026 17:30:48 -0400
Subject: [PATCH] docs: the quickstart on the home page could not be run

It opened two terminals and ran the CLIs with no install step above them. Neither package is on PyPI, which get_started.md says and the home page only linked to from its last section, so a reader following the front page in order met command not found. The four lines are copied from that page so the two recipes cannot drift.

Two smaller things, both languages: a line after the intro now says a full-size pi0.5 ran end to end and that the RL result is negative, linking to that section, because the page led with a sentence any framework could write; and the opening note is an admonition like every other callout on the page rather than a bare blockquote.
---
 docs/index.md    | 25 ++++++++++++++++++++++++-
 docs/index.zh.md | 23 ++++++++++++++++++++++-
 2 files changed, 46 insertions(+), 2 deletions(-)

diff --git a/docs/index.md b/docs/index.md
index 4952221..72adfc3 100644
--- a/docs/index.md
+++ b/docs/index.md
@@ -2,7 +2,15 @@
 
 PlugRL is an RL infrastructure for distributed experiments with a clean split between training and environment execution.
 
-> Note: PlugRL is a suite of Python packages. You can keep envs, policies, and algorithms in your own packages and import them on the side that uses them.
+!!! note
+
+    PlugRL is a suite of Python packages. You can keep envs, policies and
+    algorithms in your own packages and import them on the side that uses
+    them.
+
+A full-size pi0.5 has run end to end across that split on LIBERO, and the
+reinforcement learning result is negative:
+[A real VLA, end to end](#a-real-vla-end-to-end).
 
 ## Quickstart
 
@@ -10,6 +18,21 @@ Two processes: a training server that holds the policy, and an env client
 that runs environments and asks it for actions. This pair actually learns -
 FPO on HalfCheetah-v5, CPU only, no GPU and no assets to download.
 
+Neither package is on PyPI, so install both from source first:
+
+```bash
+git clone https://github.com/PlugRL/plugrl-server.git
+git clone https://github.com/PlugRL/plugrl-env-client.git
+
+cd plugrl-server     && uv sync && cd ..
+cd plugrl-env-client && uv sync --extra mujoco && cd ..
+```
+
+The env client has one extra per environment family, and `mujoco` is the one
+this run needs. Full steps: [Get Started](user_guide/get_started.md).
+
+Then, in two terminals:
+
 ```bash
 # Terminal 1 - the training server
 plugrl-run-server fpo-policy default fpo default \
diff --git a/docs/index.zh.md b/docs/index.zh.md
index 3c7e61e..1f5baf7 100644
--- a/docs/index.zh.md
+++ b/docs/index.zh.md
@@ -2,13 +2,34 @@
 
 PlugRL 是一套面向分布式强化学习实验的基础设施。训练端与环境端通过统一协议解耦。
 
-> Note: PlugRL 是一组可插拔的 Python 包。你的 env、policy、algorithm 可以放在自己的包里，只要在使用侧 import 并完成注册。
+!!! note
+
+    PlugRL 是一组可插拔的 Python 包。你的 env、policy、algorithm 可以放在自己的
+    包里，只要在使用侧 import 并完成注册。
+
+一个完整尺寸的 pi0.5 已经通过这条边界在 LIBERO 上端到端跑通，而强化学习的结果是
+负面的：见[真实 VLA，端到端](#vla)。
 
 ## 快速开始
 
 两个进程：训练端持有策略，环境端跑环境并向它请求动作。下面这一对**真的会学** ——
 FPO + HalfCheetah-v5，纯 CPU，不需要 GPU，也不需要下载任何资源文件。
 
+两个包都不在 PyPI 上，先从源码安装：
+
+```bash
+git clone https://github.com/PlugRL/plugrl-server.git
+git clone https://github.com/PlugRL/plugrl-env-client.git
+
+cd plugrl-server     && uv sync && cd ..
+cd plugrl-env-client && uv sync --extra mujoco && cd ..
+```
+
+env client 按环境家族划分 extra，这次要用的是 `mujoco`。完整步骤见
+[快速开始](user_guide/get_started.zh.md)。
+
+然后开两个终端：
+
 ```bash
 # 终端 1 —— 训练端
 plugrl-run-server fpo-policy default fpo default \