diff --git a/.env.example b/.env.example
index a12b24e7b..de27196af 100644
--- a/.env.example
+++ b/.env.example
@@ -1,10 +1,15 @@
 # This file is intentionally documentation-only.
 #
 # Integrated LexVoice runs should not configure frontend variables here.
-# Use `../lex-voice/.env` as the single source of truth; `lex-voice/run.sh`
-# injects LiveKit, room-input, input-source, role-device, agent, media, and
-# debug settings into the frontend process when it starts `make start_ui`.
+# Use the LexVoice repository `.env` as the single source of truth; its `run.sh`
+# injects LiveKit, room-input, input-source, role-device, agent, media, and debug
+# settings into the frontend process when it starts `make start_ui`.
 #
 # Only create `agent-starter-react/.env.local` for standalone frontend
 # development where this repository is launched directly with `pnpm dev`.
 # In that case, define only the variables needed for that standalone run.
+
+# For sandbox gateway deployment, see the LexVoice repository's
+# `deploy/liveavatar_gateway/env.example`; copy that reference file to
+# `deploy/liveavatar_gateway/.env` inside the LexVoice repository to configure
+# a deployed gateway.
diff --git a/.gitignore b/.gitignore
index 122bf192a..03a7151c0 100644
--- a/.gitignore
+++ b/.gitignore
@@ -13,6 +13,12 @@
 # testing
 /coverage
 
+# Legacy sandbox gateway artifacts from pre-migration local runs
+/.sandbox-gateway/
+/logs/direct-proxy.log
+/logs/sandbox-gateway*.log
+/logs/sandbox-gateway.pid
+
 # next.js
 /.next/
 /out/
diff --git a/README.md b/README.md
index 3cadc5a13..54ebe2f90 100644
--- a/README.md
+++ b/README.md
@@ -57,10 +57,10 @@ Run the following command to automatically clone this template.
 lk app create --template agent-starter-react
 ```
 
-For integrated LexVoice runs, configure `../lex-voice/.env` and start the
-frontend through the LexVoice runtime scripts. `../lex-voice/run.sh` injects
-LiveKit, room-input, input-source, role-device, agent, media, and debug settings
-into this Next.js process.
+For integrated LexVoice runs, configure the LexVoice repository `.env` and start
+the frontend through the LexVoice runtime scripts. The LexVoice repository's
+`run.sh` injects LiveKit, room-input, input-source, role-device, agent, media,
+and debug settings into this Next.js process.
 
 The session lifecycle API keeps start/stop state in memory, so integrated
 deployments should route `/api/session/*` to a single Next.js instance or sticky routing.
@@ -68,6 +68,16 @@ If you replace the custom connection details endpoint, it must echo the requeste
 `sessionId` and derive the same room name so dispatch and stop calls coordinate
 with the connected room.
 
+### LiveAvatar Gateway Deployments
+
+Sandbox-backed public deployments are owned by the LexVoice repository. Set
+`LIVEAVATAR_USE_SANDBOX=1` in the LexVoice repository `.env` and configure
+broker, template, warm pool, and `SANDBOX_ENV_*` values in the LexVoice repository's
+`deploy/liveavatar_gateway/.env`.
+
+This frontend repository only runs the Next.js UI. It does not create, release,
+or warm sandbox sessions.
+
 For standalone frontend development, install dependencies and run the dev
 server directly:
 
@@ -78,8 +88,8 @@ pnpm dev
 
 And open http://localhost:3000 in your browser.
 
-You'll also need a LiveKit server and an agent worker. In this workspace, those
-are normally provided by the sibling `../lex-voice` project.
+You'll also need a LiveKit server and an agent worker. In integrated workspaces,
+those are normally provided by the LexVoice project.
 
 ## Configuration
 
@@ -118,7 +128,7 @@ You can update these values in [`app-config.ts`](./app-config.ts) to customize b
 
 #### Environment Variables
 
-Integrated runs should keep runtime variables in `../lex-voice/.env`; this
+Integrated runs should keep runtime variables in the LexVoice repository `.env`; this
 repository's `.env.example` is documentation-only. Only create
 `agent-starter-react/.env.local` for standalone frontend development launched
 directly with `pnpm dev`.
@@ -130,7 +140,7 @@ LIVEKIT_URL=https://your-livekit-server-url
 ```
 
 The frontend defaults to the browser camera/microphone input when no input
-source is provided. Configure `INPUT_SOURCE` only in `../lex-voice/.env` for
+source is provided. Configure `INPUT_SOURCE` only in the LexVoice repository `.env` for
 integrated backend runs. The LiveKit variables above are required for
 standalone voice agent functionality to work with your LiveKit project.
 
diff --git a/next.config.ts b/next.config.ts
index 597fcaa5a..8564a5b86 100644
--- a/next.config.ts
+++ b/next.config.ts
@@ -1,7 +1,7 @@
 import type { NextConfig } from 'next';
 
 const nextConfig: NextConfig = {
-  allowedDevOrigins: ['liveavatar.lexmount.cn'],
+  allowedDevOrigins: ['liveavatar.local.lexmount.net'],
 };
 
 export default nextConfig;
diff --git a/tests/project-config.test.mjs b/tests/project-config.test.mjs
index fa75dc818..1cb1bf5c1 100644
--- a/tests/project-config.test.mjs
+++ b/tests/project-config.test.mjs
@@ -15,7 +15,9 @@ test('README matches the documented LexVoice environment source', async () => {
   const envExample = await readFile('.env.example', 'utf8');
 
   assert.match(envExample, /documentation-only/);
-  assert.match(readme, /\.\.\/lex-voice\/\.env/);
+  assert.match(readme, /LexVoice repository `\.env`/);
+  assert.doesNotMatch(readme, /\blex-voice\b/);
+  assert.doesNotMatch(envExample, /\blex-voice\b/);
   assert.match(readme, /single Next\.js instance or sticky routing/);
   assert.match(readme, /custom connection details endpoint/);
   assert.match(readme, /sessionId/);