feat: improve local worker, deployment progress, and WSL support

README.md

@@ -1,6 +1,10 @@
-# LMStack
+<p align="center">
+  <img src="docs/LMStack-light.png" alt="LMStack" height="80">
+</p>
 
-[中文文檔](README_zh-TW.md)
+<p align="center">
+  <a href="README_zh-TW.md">中文文檔</a>
+</p>
 
 LLM Deployment Management Platform - Deploy and manage Large Language Models on distributed GPU workers.
 
@@ -48,29 +52,14 @@ docker compose -f docker-compose.deploy.yml up -d
 - Frontend: http://localhost:3000
 - Backend API: http://localhost:52000
 
-### Start Worker (on GPU machine)
-
-```bash
-docker run -d \
-  --name lmstack-worker \
-  --gpus all \
-  --privileged \
-  -p 52001:52001 \
-  -v /var/run/docker.sock:/var/run/docker.sock \
-  -v ~/.cache/huggingface:/root/.cache/huggingface \
-  -v /:/host:ro \
-  -e BACKEND_URL=http://YOUR_SERVER_IP:52000 \
-  -e WORKER_NAME=gpu-worker-01 \
-  infinirc/lmstack-worker:latest
-```
-
 ### Usage
 
 1. Login with `admin` / `admin` (change password after first login)
-2. Check **Workers** page - workers auto-register
-3. Add model in **Models** page
-4. Create deployment in **Deployments** page
-5. Use OpenAI-compatible API:
+2. Go to **Workers** page and click **Add Worker** to get the Docker command
+3. Run the Docker command on your GPU machine to register a worker
+4. Add model in **Models** page
+5. Create deployment in **Deployments** page
+6. Use OpenAI-compatible API:
 
 ```bash
 curl http://localhost:52000/v1/chat/completions \
@@ -96,21 +85,10 @@ Build and run Docker images locally:
 
 # Run locally built backend + frontend
 docker compose -f docker-compose.local.yml up -d
-
-# Run locally built worker (on GPU machine)
-docker run -d \
-  --name lmstack-worker \
-  --gpus all \
-  --privileged \
-  -p 52001:52001 \
-  -v /var/run/docker.sock:/var/run/docker.sock \
-  -v ~/.cache/huggingface:/root/.cache/huggingface \
-  -v /:/host:ro \
-  -e BACKEND_URL=http://YOUR_SERVER_IP:52000 \
-  -e WORKER_NAME=gpu-worker-01 \
-  infinirc/lmstack-worker:local
 ```
 
+Then go to **Workers** page in the UI to add a worker.
+
 ### Without Docker
 
 ```bash

README_zh-TW.md

@@ -1,6 +1,10 @@
-# LMStack
+<p align="center">
+  <img src="docs/LMStack-light.png" alt="LMStack" height="80">
+</p>
 
-[English](README.md)
+<p align="center">
+  <a href="README.md">English</a>
+</p>
 
 LLM 部署管理平台 - 在分散式 GPU 節點上部署和管理大型語言模型。
 
@@ -23,8 +27,8 @@ LLM 部署管理平台 - 在分散式 GPU 節點上部署和管理大型語言
                    ┌────────────┴────────────┐
                    ▼                         ▼
            ┌──────────────┐          ┌──────────────┐
-           │ Worker Agent │          │ Worker Agent │
-           │  (GPU 節點)  │          │  (GPU 節點)  │
+           │ Worker       │          │ Worker       │
+           │  (GPU 節點)  │           │  (GPU 節點)  │
            └──────────────┘          └──────────────┘
 ```
 
@@ -48,29 +52,14 @@ docker compose -f docker-compose.deploy.yml up -d
 - 前端: http://localhost:3000
 - 後端 API: http://localhost:52000
 
-### 啟動 Worker（在 GPU 機器上）
-
-```bash
-docker run -d \
-  --name lmstack-worker \
-  --gpus all \
-  --privileged \
-  -p 52001:52001 \
-  -v /var/run/docker.sock:/var/run/docker.sock \
-  -v ~/.cache/huggingface:/root/.cache/huggingface \
-  -v /:/host:ro \
-  -e BACKEND_URL=http://你的伺服器IP:52000 \
-  -e WORKER_NAME=gpu-worker-01 \
-  infinirc/lmstack-worker:latest
-```
-
 ### 使用方式
 
 1. 使用 `admin` / `admin` 登入（首次登入後請更改密碼）
-2. 查看 **Workers** 頁面 - Workers 會自動註冊
-3. 在 **Models** 頁面新增模型
-4. 在 **Deployments** 頁面建立部署
-5. 使用 OpenAI 相容 API：
+2. 前往 **Workers** 頁面，點擊 **Add Worker** 取得 Docker 指令
+3. 在 GPU 機器上執行該 Docker 指令以註冊 Worker
+4. 在 **Models** 頁面新增模型
+5. 在 **Deployments** 頁面建立部署
+6. 使用 OpenAI 相容 API：
 
 ```bash
 curl http://localhost:52000/v1/chat/completions \
@@ -96,21 +85,10 @@ curl http://localhost:52000/v1/chat/completions \
 
 # 運行本地構建的 backend + frontend
 docker compose -f docker-compose.local.yml up -d
-
-# 運行本地構建的 worker（在 GPU 機器上）
-docker run -d \
-  --name lmstack-worker \
-  --gpus all \
-  --privileged \
-  -p 52001:52001 \
-  -v /var/run/docker.sock:/var/run/docker.sock \
-  -v ~/.cache/huggingface:/root/.cache/huggingface \
-  -v /:/host:ro \
-  -e BACKEND_URL=http://你的伺服器IP:52000 \
-  -e WORKER_NAME=gpu-worker-01 \
-  infinirc/lmstack-worker:local
 ```
 
+然後前往 UI 中的 **Workers** 頁面新增 Worker。
+
 ### 不使用 Docker
 
 ```bash

backend/Dockerfile

@@ -19,6 +19,12 @@ FROM python:3.11-slim
 
 WORKDIR /app
 
+# Install docker CLI for local worker spawn feature
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    curl \
+    && curl -fsSL https://download.docker.com/linux/static/stable/x86_64/docker-24.0.7.tgz | tar xz --strip-components=1 -C /usr/local/bin docker/docker \
+    && rm -rf /var/lib/apt/lists/*
+
 # Copy installed packages from builder
 COPY --from=builder /usr/local/lib/python3.11/site-packages /usr/local/lib/python3.11/site-packages
 COPY --from=builder /usr/local/bin /usr/local/bin

backend/app/api/apps/deployment.py

@@ -84,12 +84,47 @@ async def pull_image_with_progress(
         Exception: On pull failure
     """
     url = f"http://{worker.address}/images/pull"
+    progress_url = f"http://{worker.address}/images/pull-progress/{app_id}"
 
     set_deployment_progress(app_id, "pulling", 0, f"Pulling image {image}...")
 
     try:
         async with httpx.AsyncClient(timeout=IMAGE_PULL_TIMEOUT) as client:
-            response = await client.post(url, json={"image": image})
+            # Start the pull request in a task with app_id for progress tracking
+            pull_task = asyncio.create_task(
+                client.post(url, json={"image": image, "app_id": app_id})
+            )
+
+            # Poll for progress while waiting
+            while not pull_task.done():
+                try:
+                    progress_resp = await client.get(progress_url, timeout=5.0)
+                    if progress_resp.status_code == 200:
+                        progress_data = progress_resp.json()
+                        status = progress_data.get("status", "")
+                        progress = progress_data.get("progress", 0)
+
+                        if status == "pulling":
+                            set_deployment_progress(
+                                app_id,
+                                "pulling",
+                                progress,
+                                f"Pulling image {image}... ({progress}%)",
+                            )
+                        elif status == "completed":
+                            set_deployment_progress(
+                                app_id,
+                                "pulling",
+                                100,
+                                "Image pulled successfully",
+                            )
+                except Exception:
+                    pass  # Progress polling is best-effort
+
+                await asyncio.sleep(2)
+
+            # Get the final response
+            response = await pull_task
             if response.status_code >= 400:
                 raise Exception(f"Failed to pull image: {response.text}")
 

backend/app/api/apps/routes.py

@@ -179,6 +179,15 @@ async def deploy_app(
     proxy_path = f"/apps/{app_type.value}"
     port = await _find_available_port(db, worker.id)
 
+    # Auto-disable proxy for localhost workers to avoid port conflicts
+    # When worker is on localhost (using --network host), app container binds
+    # directly to host port, so proxy would conflict
+    worker_host = worker.address.split(":")[0]
+    use_proxy = deploy_request.use_proxy
+    if worker_host in ("localhost", "127.0.0.1"):
+        use_proxy = False
+        logger.info(f"Auto-disabled proxy for localhost worker {worker.name}")
+
     # Create app record
     app = App(
         app_type=app_type.value,
@@ -188,7 +197,7 @@ async def deploy_app(
         status=AppStatus.PENDING.value,
         proxy_path=proxy_path,
         port=port,
-        use_proxy=deploy_request.use_proxy,
+        use_proxy=use_proxy,
     )
     db.add(app)
     await db.commit()
@@ -222,7 +231,7 @@ async def deploy_app(
         port=port,
         app_def=app_def,
         lmstack_port=lmstack_port,
-        use_proxy=deploy_request.use_proxy,
+        use_proxy=use_proxy,
     )
 
     return app_to_response(app, request)

backend/app/api/headscale.py

@@ -9,7 +9,7 @@
 from pydantic import BaseModel, Field
 
 from app.api.auth import require_admin
-from app.services.headscale_manager import LMSTACK_USER, get_headscale_manager
+from app.services.headscale_manager import LMSTACK_USER, get_headscale_manager, get_startup_progress
 
 logger = logging.getLogger(__name__)
 router = APIRouter()
@@ -94,6 +94,14 @@ async def get_headscale_status(
         return HeadscaleStatusResponse(enabled=False, running=False)
 
 
+@router.get("/progress")
+async def get_headscale_progress(
+    _: dict = Depends(require_admin),
+):
+    """Get Headscale startup progress."""
+    return get_startup_progress()
+
+
 @router.post("/start", response_model=HeadscaleStatusResponse)
 async def start_headscale(
     request: Request,