ltcai 3.6.0 → 4.0.0

package/setup_wizard.py

@@ -0,0 +1,1142 @@
+"""
+Smart Setup Wizard — Environment Scanner, Recommender & Auto-Installer
+Detects hardware, tools, and API keys; returns tailored recommendations;
+streams SSE installation progress.
+
+Formerly the root ``setup.py``; renamed in v4 so it no longer collides with
+the setuptools build entrypoint and actually ships in the wheel
+(``pyproject.toml`` py-modules). Packaging is owned entirely by
+``pyproject.toml`` — there is deliberately no root ``setup.py``.
+"""
+
+import asyncio
+import json as _json
+import os
+import platform
+import re
+import shutil
+import subprocess
+import sys
+import time
+from pathlib import Path
+from typing import Any, AsyncIterator, Dict, List, Tuple
+
+# ── Helpers ───────────────────────────────────────────────────────────────────
+
+def _cmd(args: List[str], timeout: int = 10) -> str:
+    try:
+        r = subprocess.run(args, capture_output=True, text=True, timeout=timeout, check=False)
+        return (r.stdout or r.stderr or "").strip()
+    except Exception:
+        return ""
+
+def _sse(data: Dict) -> str:
+    return f"data: {_json.dumps(data, ensure_ascii=False)}\n\n"
+
+OFFICIAL_DOWNLOADS: Dict[str, str] = {
+    "homebrew": "https://brew.sh",
+    "python": "https://www.python.org/downloads/",
+    "node": "https://nodejs.org/en/download",
+    "git": "https://git-scm.com/downloads",
+    "ollama": "https://ollama.com/download",
+    "lmstudio": "https://lmstudio.ai/download",
+    "cuda": "https://developer.nvidia.com/cuda-downloads",
+    "mlx": "https://ml-explore.github.io/mlx/build/html/install.html",
+    "cloudflared": "https://developers.cloudflare.com/cloudflare-one/connections/connect-networks/downloads/",
+    "tesseract": "https://tesseract-ocr.github.io/tessdoc/Installation.html",
+}
+
+COMMON_PATH_DIRS = [
+    "/opt/homebrew/bin",
+    "/usr/local/bin",
+    "/usr/bin",
+    "/bin",
+    str(Path.home() / ".local" / "bin"),
+    str(Path.home() / ".cargo" / "bin"),
+    str(Path.home() / ".latticeai" / "bin"),
+]
+
+if platform.system() == "Windows":
+    _local_appdata = os.environ.get("LOCALAPPDATA", "")
+    _program_files = os.environ.get("ProgramFiles", r"C:\Program Files")
+    _program_files_x86 = os.environ.get("ProgramFiles(x86)", r"C:\Program Files (x86)")
+    COMMON_PATH_DIRS.extend([
+        str(Path(_local_appdata) / "Programs" / "Ollama") if _local_appdata else "",
+        str(Path(_program_files) / "Ollama"),
+        str(Path(_program_files) / "LM Studio"),
+        str(Path(_program_files) / "NVIDIA Corporation" / "NVSMI"),
+        str(Path(_program_files_x86) / "NVIDIA Corporation" / "NVSMI"),
+    ])
+    COMMON_PATH_DIRS = [p for p in COMMON_PATH_DIRS if p]
+
+WINDOWS_BINARY_CANDIDATES: Dict[str, List[str]] = {
+    "ollama": [
+        str(Path(os.environ.get("LOCALAPPDATA", "")) / "Programs" / "Ollama" / "ollama.exe"),
+        str(Path(os.environ.get("ProgramFiles", r"C:\Program Files")) / "Ollama" / "ollama.exe"),
+    ],
+    "lms": [
+        str(Path(os.environ.get("LOCALAPPDATA", "")) / "Programs" / "LM Studio" / "resources" / "app" / ".webpack" / "lms.exe"),
+        str(Path(os.environ.get("ProgramFiles", r"C:\Program Files")) / "LM Studio" / "resources" / "app" / ".webpack" / "lms.exe"),
+    ],
+    "nvidia-smi": [
+        str(Path(os.environ.get("ProgramFiles", r"C:\Program Files")) / "NVIDIA Corporation" / "NVSMI" / "nvidia-smi.exe"),
+        str(Path(os.environ.get("ProgramFiles(x86)", r"C:\Program Files (x86)")) / "NVIDIA Corporation" / "NVSMI" / "nvidia-smi.exe"),
+    ],
+}
+
+PACKAGE_MODULES: Dict[str, str] = {
+    "mlx-vlm": "mlx_vlm",
+    "huggingface_hub[cli]": "huggingface_hub",
+    "openai-whisper": "whisper",
+}
+
+
+def _project_env_file() -> Path:
+    return Path(__file__).resolve().parent / ".env"
+
+
+def _update_env_file(env_file: Path, key: str, value: str) -> None:
+    lines: List[str] = []
+    found = False
+    if env_file.exists():
+        lines = env_file.read_text(encoding="utf-8").splitlines()
+    updated: List[str] = []
+    for line in lines:
+        if line.startswith(f"{key}="):
+            updated.append(f"{key}={value}")
+            found = True
+        else:
+            updated.append(line)
+    if not found:
+        updated.append(f"{key}={value}")
+    env_file.write_text("\n".join(updated) + "\n", encoding="utf-8")
+
+
+def _merge_path_dirs(dirs: List[str]) -> List[str]:
+    current = os.environ.get("PATH", "")
+    parts = [p for p in current.split(os.pathsep) if p]
+    for item in dirs:
+        expanded = str(Path(item).expanduser())
+        if Path(expanded).exists() and expanded not in parts:
+            parts.insert(0, expanded)
+    os.environ["PATH"] = os.pathsep.join(parts)
+    return parts
+
+
+def _persist_extra_path(dirs: List[str]) -> None:
+    existing = [
+        p for p in os.environ.get("LATTICEAI_EXTRA_PATH", "").split(os.pathsep)
+        if p
+    ]
+    merged = existing[:]
+    for item in dirs:
+        expanded = str(Path(item).expanduser())
+        if Path(expanded).exists() and expanded not in merged:
+            merged.append(expanded)
+    if merged:
+        os.environ["LATTICEAI_EXTRA_PATH"] = os.pathsep.join(merged)
+        _update_env_file(_project_env_file(), "LATTICEAI_EXTRA_PATH", os.environ["LATTICEAI_EXTRA_PATH"])
+
+
+def repair_path_for(binary: str | None = None) -> List[str]:
+    before = _which_any(binary) if binary else None
+    paths = _merge_path_dirs(COMMON_PATH_DIRS)
+    if binary and not before and _which_any(binary):
+        _persist_extra_path(COMMON_PATH_DIRS)
+    return paths
+
+
+def _which_any(binary: str) -> str | None:
+    path = shutil.which(binary)
+    if path:
+        return path
+    if platform.system() == "Windows":
+        for candidate in WINDOWS_BINARY_CANDIDATES.get(binary, []):
+            if candidate and Path(candidate).exists():
+                return candidate
+    return None
+
+
+def _which_detail(binary: str) -> Dict[str, Any]:
+    path = _which_any(binary)
+    return {"installed": path is not None, "path": path}
+
+
+def _module_available(module_name: str) -> bool:
+    import importlib.util
+    return importlib.util.find_spec(module_name) is not None
+
+
+def _package_module(package: str) -> str:
+    return PACKAGE_MODULES.get(package, package.replace("-", "_").split("[", 1)[0])
+
+
+def _component_detail(name: str, binary: str | None = None, module: str | None = None) -> Dict[str, Any]:
+    detail: Dict[str, Any] = {"official_url": OFFICIAL_DOWNLOADS.get(name)}
+    if binary:
+        detail.update(_which_detail(binary))
+    if module:
+        detail["module_available"] = _module_available(module)
+        detail["installed"] = bool(detail.get("installed") or detail["module_available"])
+    return detail
+
+
+def _verify_binary(binary: str, version_args: List[str] | None = None, timeout: int = 20) -> Tuple[bool, str]:
+    repair_path_for(binary)
+    found = _which_any(binary)
+    if not found:
+        return False, f"{binary} 실행 파일을 PATH에서 찾지 못했습니다."
+    args = [found, *(version_args or ["--version"])]
+    try:
+        completed = subprocess.run(args, capture_output=True, text=True, timeout=timeout, check=False)
+    except Exception as e:
+        return False, str(e)
+    output = (completed.stdout or completed.stderr or "").strip().splitlines()
+    if completed.returncode == 0:
+        return True, output[0] if output else found
+    return False, (completed.stderr or completed.stdout or f"returncode={completed.returncode}")[-400:]
+
+
+async def _wait_for_binary(binary: str, seconds: int = 300) -> Tuple[bool, str]:
+    deadline = time.time() + seconds
+    while time.time() < deadline:
+        ok, msg = _verify_binary(binary)
+        if ok:
+            return True, msg
+        await asyncio.sleep(2)
+    return False, f"{binary} 설치 완료를 제한 시간 안에 감지하지 못했습니다."
+
+# ── Environment Detection ─────────────────────────────────────────────────────
+
+def _detect_chip() -> Dict[str, Any]:
+    arch = platform.machine()
+    is_apple = arch == "arm64" and platform.system() == "Darwin"
+    name = "Unknown CPU"
+    gen: Any = None
+
+    if is_apple:
+        profiler = _cmd(["system_profiler", "SPHardwareDataType"], timeout=8)
+        m = re.search(r"Chip:\s+(Apple M\S+)", profiler)
+        name = m.group(1) if m else "Apple Silicon"
+        gm = re.search(r"M(\d+)", name)
+        gen = int(gm.group(1)) if gm else 1
+    else:
+        brand = ""
+        if platform.system() == "Darwin":
+            brand = _cmd(["sysctl", "-n", "machdep.cpu.brand_string"])
+        elif platform.system() == "Windows":
+            raw = _cmd(["wmic", "cpu", "get", "Name", "/value"], timeout=5)
+            if "Name=" in raw:
+                brand = raw.split("Name=", 1)[-1].splitlines()[0].strip()
+        elif platform.system() == "Linux":
+            try:
+                for line in Path("/proc/cpuinfo").read_text(encoding="utf-8", errors="replace").splitlines():
+                    if line.lower().startswith("model name"):
+                        brand = line.split(":", 1)[-1].strip()
+                        break
+            except Exception:
+                pass
+        name = brand or platform.processor() or "Unknown CPU"
+
+    return {"name": name, "arch": arch, "is_apple_silicon": is_apple, "gen": gen}
+
+
+def _detect_cpu() -> Dict[str, Any]:
+    flags: List[str] = []
+    physical_cores = os.cpu_count() or 0
+    logical_cores = os.cpu_count() or 0
+    model = _detect_chip()["name"]
+    if platform.system() == "Darwin":
+        flags = [item.lower() for item in _cmd(["sysctl", "-n", "machdep.cpu.features"], timeout=5).split()]
+        try:
+            physical_cores = int(_cmd(["sysctl", "-n", "hw.physicalcpu"], timeout=5) or physical_cores)
+            logical_cores = int(_cmd(["sysctl", "-n", "hw.logicalcpu"], timeout=5) or logical_cores)
+        except ValueError:
+            pass
+    elif platform.system() == "Linux":
+        try:
+            text = Path("/proc/cpuinfo").read_text(encoding="utf-8", errors="replace")
+            for line in text.splitlines():
+                if line.lower().startswith(("flags", "features")):
+                    flags = line.split(":", 1)[-1].strip().lower().split()
+                    break
+        except Exception:
+            pass
+    elif platform.system() == "Windows":
+        raw = _cmd(["wmic", "cpu", "get", "Name,NumberOfCores,NumberOfLogicalProcessors", "/format:list"], timeout=5)
+        for line in raw.splitlines():
+            key, _, value = line.partition("=")
+            if key == "Name" and value.strip():
+                model = value.strip()
+            elif key == "NumberOfCores" and value.strip():
+                try:
+                    physical_cores = int(value.strip())
+                except ValueError:
+                    pass
+            elif key == "NumberOfLogicalProcessors" and value.strip():
+                try:
+                    logical_cores = int(value.strip())
+                except ValueError:
+                    pass
+        try:
+            import ctypes
+            kernel32 = ctypes.windll.kernel32
+            feature_map = {
+                6: "sse",
+                10: "sse2",
+                13: "sse3",
+                19: "neon",
+                28: "rdrand",
+            }
+            flags.extend(name for code, name in feature_map.items() if kernel32.IsProcessorFeaturePresent(code))
+        except Exception:
+            pass
+    interesting = {"avx", "avx2", "avx512f", "fma", "neon", "sse4_2"}
+    if platform.system() == "Windows":
+        interesting.update({"sse", "sse2", "sse3", "rdrand"})
+    return {
+        "model": model,
+        "physical_cores": physical_cores,
+        "logical_cores": logical_cores,
+        "instructions": sorted({flag for flag in flags if flag in interesting}),
+    }
+
+def _detect_ram_gb() -> float:
+    if platform.system() == "Windows":
+        raw = _cmd(["wmic", "ComputerSystem", "get", "TotalPhysicalMemory", "/format:list"], timeout=5)
+        for line in raw.splitlines():
+            if line.startswith("TotalPhysicalMemory="):
+                try:
+                    return round(int(line.split("=", 1)[-1].strip()) / 1_073_741_824, 1)
+                except ValueError:
+                    break
+    raw = _cmd(["sysctl", "-n", "hw.memsize"])
+    if raw:
+        try:
+            return round(int(raw) / 1_073_741_824, 1)
+        except ValueError:
+            pass
+    if platform.system() == "Darwin":
+        profiler = _cmd(["system_profiler", "SPHardwareDataType"], timeout=8)
+        m = re.search(r"Memory:\s+([\d.]+)\s*(TB|GB|MB)", profiler, re.IGNORECASE)
+        if m:
+            value = float(m.group(1))
+            unit = m.group(2).lower()
+            if unit == "tb":
+                return round(value * 1024, 1)
+            if unit == "gb":
+                return round(value, 1)
+            return round(value / 1024, 1)
+        hostinfo = _cmd(["hostinfo"], timeout=5)
+        m = re.search(r"Primary memory available:\s+([\d.]+)\s+gigabytes", hostinfo, re.IGNORECASE)
+        if m:
+            return round(float(m.group(1)), 1)
+    try:
+        with open("/proc/meminfo") as f:
+            for line in f:
+                if line.startswith("MemTotal:"):
+                    return round(int(line.split()[1]) / 1_048_576, 1)
+    except Exception:
+        pass
+    return 0.0
+
+def _detect_disk_free_gb() -> float:
+    try:
+        path = "C:\\" if platform.system() == "Windows" else "/"
+        return round(shutil.disk_usage(path).free / 1_073_741_824, 1)
+    except Exception:
+        return 0.0
+
+
+def _parse_windows_video_controllers(raw: str) -> List[Dict[str, Any]]:
+    controllers: List[Dict[str, Any]] = []
+    if not raw:
+        return controllers
+    try:
+        data = _json.loads(raw)
+        if isinstance(data, dict):
+            data = [data]
+        if isinstance(data, list):
+            for item in data:
+                name = str(item.get("Name") or "").strip()
+                try:
+                    ram_mb = int(item.get("AdapterRAM") or 0) // (1024 * 1024)
+                except Exception:
+                    ram_mb = 0
+                if name:
+                    controllers.append({"name": name, "vram_mb": ram_mb})
+        if controllers:
+            return controllers
+    except Exception:
+        pass
+    current: Dict[str, Any] = {}
+    for line in raw.splitlines():
+        if line.startswith("Name="):
+            if current:
+                controllers.append(current)
+            current = {"name": line.split("=", 1)[-1].strip(), "vram_mb": 0}
+        elif line.startswith("AdapterRAM=") and current:
+            try:
+                current["vram_mb"] = int(line.split("=", 1)[-1].strip()) // (1024 * 1024)
+            except ValueError:
+                current["vram_mb"] = 0
+    if current:
+        controllers.append(current)
+    return controllers
+
+
+def _detect_gpu() -> Dict[str, Any]:
+    devices: List[Dict[str, Any]] = []
+    nvidia_smi = _which_any("nvidia-smi")
+    if nvidia_smi:
+        info = _cmd([nvidia_smi, "--query-gpu=name,memory.total", "--format=csv,noheader,nounits"], timeout=8)
+        for line in [item.strip() for item in info.splitlines() if item.strip()]:
+            try:
+                name, mem = [part.strip() for part in line.split(",", 1)]
+                devices.append({"vendor": "nvidia", "name": name, "vram_mb": int(float(mem)), "backend": "cuda"})
+            except Exception:
+                continue
+
+    if platform.system() == "Windows":
+        shell = _which_any("powershell") or _which_any("pwsh")
+        raw = ""
+        if shell:
+            raw = _cmd([
+                shell, "-NoProfile", "-Command",
+                "Get-CimInstance Win32_VideoController | Select-Object Name,AdapterRAM | ConvertTo-Json -Compress",
+            ], timeout=8)
+        if not raw:
+            raw = _cmd(["wmic", "path", "win32_VideoController", "get", "Name,AdapterRAM", "/format:list"], timeout=8)
+        for item in _parse_windows_video_controllers(raw):
+            if any(existing.get("name") == item["name"] for existing in devices):
+                continue
+            low = item["name"].lower()
+            vendor, backend = "unknown", "cpu"
+            if "nvidia" in low or "geforce" in low or "rtx" in low:
+                vendor, backend = "nvidia", "cuda"
+            elif "amd" in low or "radeon" in low:
+                vendor, backend = "amd", "directml/vulkan"
+            elif "intel" in low or "arc" in low or "iris" in low:
+                vendor, backend = "intel", "directml/vulkan"
+            devices.append({"vendor": vendor, "name": item["name"], "vram_mb": item["vram_mb"], "backend": backend})
+    elif platform.system() == "Darwin":
+        sp = _cmd(["system_profiler", "SPDisplaysDataType"], timeout=8)
+        for line in sp.splitlines():
+            if "Chipset Model" in line:
+                devices.append({"vendor": "apple", "name": line.split(":", 1)[-1].strip(), "vram_mb": 0, "backend": "metal/mlx"})
+                break
+    elif platform.system() == "Linux" and not devices:
+        info = _cmd(["lspci"], timeout=5)
+        for line in info.splitlines():
+            low = line.lower()
+            if not any(token in low for token in ("vga", "3d controller", "display")):
+                continue
+            if "nvidia" in low:
+                devices.append({"vendor": "nvidia", "name": line.strip(), "vram_mb": 0, "backend": "cuda"})
+            elif "amd" in low or "advanced micro devices" in low or "radeon" in low:
+                devices.append({"vendor": "amd", "name": line.strip(), "vram_mb": 0, "backend": "rocm/vulkan"})
+            elif "intel" in low:
+                devices.append({"vendor": "intel", "name": line.strip(), "vram_mb": 0, "backend": "vulkan"})
+
+    primary = max(devices, key=lambda item: int(item.get("vram_mb") or 0), default={})
+    vram_mb = int(primary.get("vram_mb") or 0)
+    return {
+        "devices": devices,
+        "vendor": primary.get("vendor", "none"),
+        "name": primary.get("name", ""),
+        "vram_mb": vram_mb,
+        "vram_gb": round(vram_mb / 1024, 1),
+        "backend": primary.get("backend", "cpu"),
+    }
+
+
+def _detect_cuda() -> Dict[str, Any]:
+    nvidia_smi = _which_any("nvidia-smi")
+    nvcc = _which_any("nvcc")
+    version = ""
+    if nvidia_smi:
+        raw = _cmd([nvidia_smi, "--query-gpu=driver_version", "--format=csv,noheader"], timeout=5)
+        version = raw.splitlines()[0].strip() if raw.splitlines() else ""
+    if nvcc:
+        raw = _cmd([nvcc, "--version"], timeout=5)
+        m = re.search(r"release\s+([\d.]+)", raw)
+        if m:
+            version = m.group(1)
+    return {"available": bool(nvidia_smi or nvcc), "nvidia_smi": nvidia_smi, "nvcc": nvcc, "version": version}
+
+
+def _detect_wsl() -> Dict[str, Any]:
+    if platform.system() != "Linux":
+        return {"is_wsl": False, "version": ""}
+    raw = ""
+    try:
+        raw = Path("/proc/version").read_text(encoding="utf-8", errors="replace")
+    except Exception:
+        pass
+    is_wsl = "microsoft" in raw.lower() or "wsl" in raw.lower()
+    version = "2" if "microsoft-standard" in raw.lower() or "wsl2" in raw.lower() else ("1" if is_wsl else "")
+    return {"is_wsl": is_wsl, "version": version}
+
+
+def _detect_tools() -> Dict[str, bool]:
+    repair_path_for()
+    return {t: _which_any(t) is not None
+            for t in ["brew", "ollama", "python3", "python", "node", "npm", "git", "tesseract", "lms", "nvidia-smi", "nvcc"]}
+
+def _detect_mlx() -> Dict[str, Any]:
+    return {
+        "available": _module_available("mlx"),
+        "mlx_vlm": _module_available("mlx_vlm"),
+    }
+
+def _detect_api_keys() -> Dict[str, bool]:
+    return {
+        "openai":     bool(os.getenv("OPENAI_API_KEY")),
+        "openrouter": bool(os.getenv("OPENROUTER_API_KEY")),
+        "groq":       bool(os.getenv("GROQ_API_KEY")),
+        "together":   bool(os.getenv("TOGETHER_API_KEY")),
+    }
+
+def scan_environment() -> Dict[str, Any]:
+    chip = _detect_chip()
+    cpu = _detect_cpu()
+    gpu = _detect_gpu()
+    cuda = _detect_cuda()
+    wsl = _detect_wsl()
+    tools = _detect_tools()
+    python_binary = "python3" if tools.get("python3") else "python"
+    return {
+        "os":           platform.system(),
+        "os_version":   platform.mac_ver()[0] if platform.system() == "Darwin" else platform.version(),
+        "chip":         chip,
+        "cpu":          cpu,
+        "gpu":          gpu,
+        "cuda":         cuda,
+        "wsl":          wsl,
+        "ram_gb":       _detect_ram_gb(),
+        "disk_free_gb": _detect_disk_free_gb(),
+        "tools":        tools,
+        "components": {
+            "homebrew": _component_detail("homebrew", "brew"),
+            "python": {**_component_detail("python", python_binary), "version": platform.python_version()},
+            "node": _component_detail("node", "node"),
+            "npm": _component_detail("node", "npm"),
+            "git": _component_detail("git", "git"),
+            "ollama": _component_detail("ollama", "ollama"),
+            "lmstudio": _component_detail("lmstudio", "lms"),
+            "cuda": {**_component_detail("cuda", "nvcc"), **cuda},
+            "tesseract": _component_detail("tesseract", "tesseract"),
+            "mlx": _component_detail("mlx", module="mlx"),
+            "mlx_vlm": _component_detail("mlx", module="mlx_vlm"),
+        },
+        "path": {
+            "active": os.environ.get("PATH", ""),
+            "extra": os.environ.get("LATTICEAI_EXTRA_PATH", ""),
+        },
+        "mlx":          _detect_mlx(),
+        "api_keys":     _detect_api_keys(),
+    }
+
+# ── Model Catalog ─────────────────────────────────────────────────────────────
+# (model_id, display_name, size_gb, tag, description, min_ram_gb)
+_MODEL_CATALOG = [
+    ("mlx-community/Qwen3-VL-4B-Instruct-4bit",      "Qwen3-VL 4B",          2.7,  "VLM",     "최신 Qwen 멀티모달 · 저사양",       4),
+    ("mlx-community/gemma-4-e2b-it-4bit",            "Gemma 4 E2B",          3.6,  "VLM",     "Gemma 4 소형 멀티모달",            8),
+    ("mlx-community/Qwen3-VL-8B-Instruct-4bit",      "Qwen3-VL 8B",          4.8,  "VLM",     "최신 Qwen 멀티모달 · 균형 추천",    16),
+    ("mlx-community/gemma-4-12b-it-4bit",            "Gemma 4 12B",          7.6,  "VLM",     "Gemma 4 기본 추천 · 4bit",         16),
+    ("mlx-community/Llama-4-Scout-17B-16E-Instruct-4bit", "Llama 4 Scout",   11.8, "VLM",     "Meta 최신 멀티모달 Scout",          24),
+    ("mlx-community/gemma-4-26b-a4b-it-4bit",        "Gemma 4 26B",         15.6,  "VLM",     "이미지 지원 · 대형 추천",        32),
+    ("mlx-community/gemma-4-31b-it-4bit",            "Gemma 4 31B",         18.4,  "VLM+",    "Gemma 4 최신 31B instruct",      48),
+    ("mlx-community/Qwen3-VL-30B-A3B-Instruct-4bit", "Qwen3-VL 30B A3B",    18.0,  "VLM+",    "최신 Qwen 대형 멀티모달",        48),
+]
+
+_CROSS_PLATFORM_MODEL_CATALOG: Dict[str, List[Tuple[str, str, float, str, str, int]]] = {
+    "ollama": [
+        ("ollama:qwen3-vl:4b", "Qwen3-VL 4B", 2.7, "VLM", "Ollama 멀티모달 · 저사양", 4),
+        ("ollama:qwen3-vl:8b", "Qwen3-VL 8B", 4.8, "VLM", "Ollama 멀티모달 · 균형 추천", 16),
+        ("ollama:hf.co/ggml-org/gemma-4-12B-it-GGUF:Q4_K_M", "Gemma 4 12B Q4", 7.9, "VLM", "Hugging Face GGUF 기반 Gemma 4", 16),
+        ("ollama:hf.co/ggml-org/gemma-4-31B-it-GGUF:Q4_K_M", "Gemma 4 31B Q4", 18.7, "VLM+", "Hugging Face GGUF 기반 Gemma 4", 48),
+        ("ollama:hf.co/ggml-org/Llama-4-Scout-17B-16E-Instruct-GGUF:Q4_K_M", "Llama 4 Scout Q4", 12.0, "VLM", "Meta 최신 멀티모달 Scout", 24),
+    ],
+    "lmstudio": [
+        ("lmstudio:Qwen/Qwen3-VL-4B-Instruct", "Qwen3-VL 4B", 2.7, "VLM", "LM Studio 멀티모달 · 저사양", 4),
+        ("lmstudio:Qwen/Qwen3-VL-8B-Instruct", "Qwen3-VL 8B", 4.8, "VLM", "LM Studio 멀티모달 · 균형 추천", 16),
+        ("lmstudio:ggml-org/gemma-4-12B-it-GGUF", "Gemma 4 12B 4-bit", 7.9, "VLM", "LM Studio GGUF Gemma 4", 16),
+        ("lmstudio:ggml-org/gemma-4-31B-it-GGUF", "Gemma 4 31B 4-bit", 18.7, "VLM+", "LM Studio GGUF Gemma 4", 48),
+        ("lmstudio:Qwen/Qwen3-VL-30B-A3B-Instruct", "Qwen3-VL 30B A3B", 18.0, "VLM+", "대형 Qwen 멀티모달 · 24GB+ VRAM 권장", 32),
+        ("lmstudio:meta-llama/Llama-4-Scout-17B-16E-Instruct", "Llama 4 Scout", 12.0, "VLM", "Meta 최신 멀티모달 Scout", 24),
+    ],
+    "vllm": [
+        ("vllm:Qwen/Qwen3-VL-4B-Instruct", "Qwen3-VL 4B", 2.7, "VLM", "내 컴퓨터 GPU 실행 도구 권장", 4),
+        ("vllm:Qwen/Qwen3-VL-8B-Instruct", "Qwen3-VL 8B", 4.8, "VLM", "내 컴퓨터 NVIDIA 실행 도구 권장", 16),
+        ("vllm:google/gemma-4-12b-it", "Gemma 4 12B", 7.6, "VLM", "Gemma 4 기본 추천 · 4bit", 16),
+        ("vllm:Qwen/Qwen3-VL-30B-A3B-Instruct", "Qwen3-VL 30B A3B", 18.0, "VLM+", "대형 Qwen 멀티모달 · 24GB+ VRAM 권장", 32),
+        ("vllm:suitch/gemma-4-31B-it-4bit", "Gemma 4 31B", 18.7, "VLM+", "Gemma 4 최신 31B instruct", 48),
+        ("vllm:meta-llama/Llama-4-Scout-17B-16E-Instruct", "Llama 4 Scout", 12.0, "VLM", "Meta 최신 멀티모달 Scout", 24),
+    ],
+    "llamacpp": [
+        ("llamacpp:Qwen/Qwen3-VL-4B-Instruct-GGUF", "Qwen3-VL 4B GGUF", 2.7, "GGUF", "CPU/Vulkan 백업 · 멀티모달 GGUF", 4),
+        ("llamacpp:Qwen/Qwen3-VL-8B-Instruct-GGUF", "Qwen3-VL 8B GGUF", 4.8, "GGUF", "CPU/Vulkan 백업 · 균형형", 16),
+        ("llamacpp:ggml-org/gemma-4-12B-it-GGUF", "Gemma 4 12B GGUF", 7.9, "GGUF", "Gemma 4 12B Q4_K_M", 16),
+        ("llamacpp:ggml-org/gemma-4-31B-it-GGUF", "Gemma 4 31B GGUF", 18.7, "GGUF", "Gemma 4 31B Q4_K_M", 48),
+        ("llamacpp:ggml-org/Llama-4-Scout-17B-16E-Instruct-GGUF", "Llama 4 Scout GGUF", 12.0, "GGUF", "Meta 최신 멀티모달 Scout", 24),
+    ],
+}
+
+_VERSIONED_MODEL_PATTERNS = (
+    ("gemma", re.compile(r"\bgemma[-\s]?(\d+(?:\.\d+)?)", re.IGNORECASE)),
+    ("qwen", re.compile(r"\bqwen[-\s]?(\d+(?:\.\d+)?)", re.IGNORECASE)),
+    ("llama", re.compile(r"\bllama[-\s]?(\d+(?:\.\d+)?)", re.IGNORECASE)),
+)
+
+_BEST_MODEL_TIERS: Dict[str, List[Tuple[int, str]]] = {
+    "local_mlx": [
+        (48, "mlx-community/gemma-4-31b-it-4bit"),
+        (32, "mlx-community/gemma-4-26b-a4b-it-4bit"),
+        (16, "mlx-community/gemma-4-12b-it-4bit"),
+        (16, "mlx-community/Qwen3-VL-8B-Instruct-4bit"),
+        (4, "mlx-community/Qwen3-VL-4B-Instruct-4bit"),
+    ],
+    "ollama": [
+        (48, "ollama:hf.co/ggml-org/gemma-4-31B-it-GGUF:Q4_K_M"),
+        (16, "ollama:hf.co/ggml-org/gemma-4-12B-it-GGUF:Q4_K_M"),
+        (16, "ollama:qwen3-vl:8b"),
+        (4, "ollama:qwen3-vl:4b"),
+    ],
+    "lmstudio": [
+        (48, "lmstudio:ggml-org/gemma-4-31B-it-GGUF"),
+        (16, "lmstudio:ggml-org/gemma-4-12B-it-GGUF"),
+        (16, "lmstudio:Qwen/Qwen3-VL-8B-Instruct"),
+        (4, "lmstudio:Qwen/Qwen3-VL-4B-Instruct"),
+    ],
+    "vllm": [
+        (48, "vllm:suitch/gemma-4-31B-it-4bit"),
+        (16, "vllm:google/gemma-4-12b-it"),
+        (16, "vllm:Qwen/Qwen3-VL-8B-Instruct"),
+        (4, "vllm:Qwen/Qwen3-VL-4B-Instruct"),
+    ],
+    "llamacpp": [
+        (48, "llamacpp:ggml-org/gemma-4-31B-it-GGUF"),
+        (16, "llamacpp:ggml-org/gemma-4-12B-it-GGUF"),
+        (16, "llamacpp:Qwen/Qwen3-VL-8B-Instruct-GGUF"),
+        (4, "llamacpp:Qwen/Qwen3-VL-4B-Instruct-GGUF"),
+    ],
+}
+
+
+def _version_tuple(raw: str) -> Tuple[int, ...]:
+    return tuple(int(part) for part in raw.split(".") if part.isdigit())
+
+
+def _catalog_row_family_version(row: Tuple[str, str, float, str, str, int]) -> Tuple[str, Tuple[int, ...]] | None:
+    text = f"{row[0]} {row[1]}"
+    for family, pattern in _VERSIONED_MODEL_PATTERNS:
+        match = pattern.search(text)
+        if match:
+            version = _version_tuple(match.group(1))
+            if version:
+                return family, version
+    return None
+
+
+def _filter_lower_family_versions(
+    rows: List[Tuple[str, str, float, str, str, int]],
+) -> List[Tuple[str, str, float, str, str, int]]:
+    max_versions: Dict[str, Tuple[int, ...]] = {}
+    detected: List[Tuple[Tuple[str, str, float, str, str, int], Tuple[str, Tuple[int, ...]] | None]] = []
+    for row in rows:
+        version_info = _catalog_row_family_version(row)
+        detected.append((row, version_info))
+        if not version_info:
+            continue
+        family, version = version_info
+        if version > max_versions.get(family, (0,)):
+            max_versions[family] = version
+    return [
+        row for row, version_info in detected
+        if not version_info or version_info[1] >= max_versions.get(version_info[0], version_info[1])
+    ]
+
+
+def _best_model_for_engine(engine: str, ram_gb: float, rows: List[Tuple[str, str, float, str, str, int]]) -> str:
+    available_ids = {row[0] for row in rows}
+    for min_ram, model_id in _BEST_MODEL_TIERS.get(engine, []):
+        if ram_gb >= min_ram and model_id in available_ids:
+            return model_id
+    return rows[0][0] if rows else ""
+
+
+# ── Recommendation Logic ──────────────────────────────────────────────────────
+
+def get_recommendations(env: Dict[str, Any]) -> Dict[str, Any]:
+    ram       = env["ram_gb"]
+    chip      = env["chip"]
+    mlx       = env["mlx"]
+    tools     = env["tools"]
+    api_keys  = env["api_keys"]
+    disk_free = env["disk_free_gb"]
+    is_apple  = chip["is_apple_silicon"]
+    gpu       = env.get("gpu", {})
+    cuda      = env.get("cuda", {})
+    wsl       = env.get("wsl", {})
+    cpu       = env.get("cpu", {})
+    os_name   = env.get("os", "")
+
+    max_model_gb = ram * 0.72  # ~28% headroom for OS + apps
+
+    if is_apple:
+        preferred_engine = "local_mlx"
+    elif gpu.get("vendor") == "nvidia" and cuda.get("available") and (os_name == "Linux" or wsl.get("is_wsl")):
+        preferred_engine = "vllm"
+    elif tools.get("lms"):
+        preferred_engine = "lmstudio"
+    elif tools.get("ollama"):
+        preferred_engine = "ollama"
+    else:
+        preferred_engine = "llamacpp"
+
+    apple_catalog = _filter_lower_family_versions(_MODEL_CATALOG)
+    engine_catalog = (
+        []
+        if is_apple
+        else _filter_lower_family_versions(_CROSS_PLATFORM_MODEL_CATALOG[preferred_engine])
+    )
+    best_id = _best_model_for_engine(
+        "local_mlx" if is_apple else preferred_engine,
+        ram,
+        apple_catalog if is_apple else engine_catalog,
+    )
+
+    # ── Engines ──────────────────────────────────────────────────────────────
+    engines: List[Dict] = []
+
+    if is_apple:
+        if mlx["available"] and mlx["mlx_vlm"]:
+            engines.append({
+                "id": "engine_mlx", "name": "MLX",
+                "subtitle": f"{chip['name']} GPU 가속 · MLX-VLM 멀티모달 실행",
+                "status": "installed", "priority": "recommended",
+                "checked": True, "action": None, "badge": "설치됨",
+            })
+        else:
+            engines.append({
+                "id": "engine_mlx", "name": "MLX",
+                "subtitle": f"{chip['name']} 전용 MLX-VLM 멀티모달 실행",
+                "status": "available", "priority": "recommended",
+                "checked": True,
+                "action": {"type": "pip", "packages": ["mlx-vlm"], "verify_modules": ["mlx", "mlx_vlm"]},
+                "badge": "설치 필요",
+            })
+
+    if tools.get("ollama"):
+        engines.append({
+            "id": "engine_ollama", "name": "Ollama",
+            "subtitle": "범용 로컬 LLM 서버 · 크로스 플랫폼",
+            "status": "installed", "priority": "recommended" if preferred_engine == "ollama" else "optional",
+            "checked": preferred_engine == "ollama", "action": None, "badge": "설치됨",
+        })
+    else:
+        hint = "brew install 가능" if (tools.get("brew") or env["os"] == "Darwin") else "수동 설치 필요"
+        engines.append({
+            "id": "engine_ollama", "name": "Ollama",
+            "subtitle": "범용 로컬 LLM 서버 · 크로스 플랫폼",
+            "status": "available", "priority": "recommended" if preferred_engine == "ollama" else "optional",
+            "checked": preferred_engine == "ollama",
+            "action": (
+                {"type": "brew", "package": "ollama", "binary": "ollama", "official_url": OFFICIAL_DOWNLOADS["ollama"]}
+                if tools.get("brew")
+                else {"type": "url", "url": OFFICIAL_DOWNLOADS["ollama"], "binary": "ollama"}
+            ),
+            "badge": hint,
+        })
+
+    if not is_apple:
+        lmstudio_installed = bool(tools.get("lms"))
+        engines.append({
+            "id": "engine_lmstudio", "name": "LM Studio",
+            "subtitle": "Windows/macOS/Linux 데스크톱 GPU 서버 · 모델 다운로드 UI 포함",
+            "status": "installed" if lmstudio_installed else "available",
+            "priority": "recommended" if preferred_engine == "lmstudio" else "optional",
+            "checked": preferred_engine == "lmstudio",
+            "action": None if lmstudio_installed else {"type": "url", "url": OFFICIAL_DOWNLOADS["lmstudio"], "binary": "lms"},
+            "badge": "설치됨" if lmstudio_installed else "설치 필요",
+        })
+        if gpu.get("vendor") == "nvidia":
+            engines.append({
+                "id": "engine_cuda", "name": "CUDA",
+                "subtitle": f"NVIDIA {gpu.get('name') or 'GPU'} · VRAM {gpu.get('vram_gb') or 0} GB",
+                "status": "installed" if cuda.get("available") else "available",
+                "priority": "recommended",
+                "checked": False,
+                "action": None if cuda.get("available") else {"type": "url", "url": OFFICIAL_DOWNLOADS["cuda"], "binary": "nvcc"},
+                "badge": cuda.get("version") or ("감지됨" if cuda.get("available") else "설치 필요"),
+            })
+            engines.append({
+                "id": "engine_vllm", "name": "vLLM",
+                "subtitle": "NVIDIA 서버형 추론 · Windows는 WSL/Linux 권장",
+                "status": "available",
+                "priority": "recommended" if preferred_engine == "vllm" else "optional",
+                "checked": preferred_engine == "vllm",
+                "action": {"type": "pip", "packages": ["vllm", "huggingface_hub[cli]"], "verify_modules": ["vllm", "huggingface_hub"]},
+                "badge": "WSL/Linux 권장" if os_name == "Windows" and not wsl.get("is_wsl") else "설치 가능",
+            })
+        elif gpu.get("vendor") in {"amd", "intel"}:
+            engines.append({
+                "id": "engine_vulkan_directml", "name": "Vulkan/DirectML",
+                "subtitle": f"{gpu.get('vendor', '').upper()} GPU 감지 · LM Studio 또는 llama.cpp 백엔드 권장",
+                "status": "available",
+                "priority": "recommended" if preferred_engine in {"lmstudio", "llamacpp"} else "optional",
+                "checked": False,
+                "action": None,
+                "badge": gpu.get("backend") or "GPU",
+            })
+
+    components: List[Dict] = []
+    component_specs = [
+        ("homebrew", "Homebrew", "macOS 패키지 관리자 · 자동 설치 기반", "brew", None, "recommended"),
+        ("git", "Git", "저장소 · 확장 · MCP 도구 연동에 필요", "git", "git", "recommended"),
+        ("node", "Node.js", "npm 패키지와 VS Code 확장 개발에 필요", "node", "node", "optional"),
+        ("tesseract", "Tesseract OCR", "이미지/PDF OCR 기능에 필요", "tesseract", "tesseract", "optional"),
+    ]
+    for cid, name, subtitle, binary, brew_pkg, priority in component_specs:
+        installed = bool(tools.get(binary))
+        if cid == "homebrew" and env["os"] != "Darwin":
+            continue
+        if installed:
+            components.append({
+                "id": f"component_{cid}", "name": name,
+                "subtitle": subtitle, "status": "installed",
+                "priority": priority, "checked": False, "action": None,
+                "badge": "설치됨",
+            })
+            continue
+        if cid == "homebrew":
+            action = {"type": "url", "url": OFFICIAL_DOWNLOADS["homebrew"], "binary": "brew"}
+        elif tools.get("brew") and brew_pkg:
+            action = {"type": "brew", "package": brew_pkg, "binary": binary, "official_url": OFFICIAL_DOWNLOADS.get(cid)}
+        else:
+            action = {"type": "url", "url": OFFICIAL_DOWNLOADS.get(cid, ""), "binary": binary}
+        components.append({
+            "id": f"component_{cid}", "name": name,
+            "subtitle": subtitle, "status": "available",
+            "priority": priority, "checked": priority == "recommended",
+            "action": action, "badge": "설치 필요",
+        })
+
+    python_ok = sys.version_info >= (3, 11)
+    if not python_ok:
+        components.insert(0, {
+            "id": "component_python", "name": "Python 3.11+",
+            "subtitle": "Lattice AI 서버 실행에 필요한 Python 런타임",
+            "status": "available", "priority": "recommended", "checked": True,
+            "action": {"type": "url", "url": OFFICIAL_DOWNLOADS["python"], "binary": "python3"},
+            "badge": "업데이트 필요",
+        })
+
+    for provider, has_key in api_keys.items():
+        if has_key:
+            engines.append({
+                "id": f"engine_{provider}", "name": provider.title(),
+                "subtitle": f"{provider.upper()}_API_KEY 감지됨 · 클라우드 API",
+                "status": "ready", "priority": "optional",
+                "checked": False, "action": None, "badge": "준비됨",
+            })
+
+    # ── Models ───────────────────────────────────────────────────────────────
+    models: List[Dict] = []
+
+    if is_apple:
+        for mid, mname, size_gb, tag, desc, min_ram in apple_catalog:
+            fits    = ram >= min_ram and size_gb <= max_model_gb and disk_free >= size_gb + 2
+            is_best = mid == best_id
+            models.append({
+                "id":       f"model_{mid.replace('/', '__').replace('-', '_')}",
+                "model_id": mid,
+                "name":     mname,
+                "subtitle": desc,
+                "size_gb":  size_gb,
+                "tag":      tag,
+                "fits":     fits,
+                "priority": "recommended" if is_best else "optional",
+                "checked":  is_best and fits,
+                "disabled": not fits,
+                "badge":    f"{size_gb} GB",
+                "action":   {"type": "load_model", "model_id": mid} if fits else None,
+            })
+    else:
+        vram_gb = float(gpu.get("vram_gb") or 0)
+        gpu_budget_gb = vram_gb * 1.15 if gpu.get("vendor") in {"nvidia", "amd", "intel"} and vram_gb else max_model_gb
+        model_budget_gb = min(max_model_gb, gpu_budget_gb)
+        for mid, mname, size_gb, tag, desc, min_ram in engine_catalog:
+            fits = ram >= min_ram and size_gb <= model_budget_gb and disk_free >= size_gb + 2
+            is_best = mid == best_id
+            models.append({
+                "id":       f"model_{mid.replace('/', '__').replace(':', '__').replace('-', '_')}",
+                "model_id": mid,
+                "name":     mname,
+                "subtitle": desc,
+                "size_gb":  size_gb,
+                "tag":      tag,
+                "fits":     fits,
+                "priority": "recommended" if is_best else "optional",
+                "checked":  is_best and fits,
+                "disabled": not fits,
+                "badge":    f"{size_gb} GB · {preferred_engine}",
+                "action":   {"type": "load_model", "model_id": mid} if fits else None,
+            })
+    if models and not any(item.get("checked") for item in models):
+        for item in models:
+            if not item.get("disabled"):
+                item["priority"] = "recommended"
+                item["checked"] = True
+                break
+
+    # ── MCPs ─────────────────────────────────────────────────────────────────
+    mcps: List[Dict] = [
+        {
+            "id": "mcp_files", "name": "Workspace Files",
+            "subtitle": "파일 읽기/쓰기 · 코드 생성 · 미리보기",
+            "status": "active", "priority": "recommended",
+            "checked": True, "action": None,
+            "badge": "기본 탑재", "needs_auth": False,
+        },
+        {
+            "id": "mcp_presentations", "name": "Presentations",
+            "subtitle": "PPTX · 슬라이드 자동 생성",
+            "status": "active", "priority": "optional",
+            "checked": False, "action": None,
+            "badge": "기본 탑재", "needs_auth": False,
+        },
+        {
+            "id": "mcp_github", "name": "GitHub",
+            "subtitle": "저장소 · PR · 이슈 · CI 연동",
+            "status": "available", "priority": "optional",
+            "checked": False,
+            "action": {"type": "auth", "url": "https://github.com/apps", "mcp_id": "github"},
+            "badge": "인증 필요", "needs_auth": True,
+        },
+        {
+            "id": "mcp_googledrive", "name": "Google Drive",
+            "subtitle": "Docs · Sheets · Drive 파일 연동",
+            "status": "available", "priority": "optional",
+            "checked": False,
+            "action": {"type": "auth", "url": "https://chatgpt.com/connectors", "mcp_id": "google-drive"},
+            "badge": "인증 필요", "needs_auth": True,
+        },
+        {
+            "id": "mcp_slack", "name": "Slack",
+            "subtitle": "팀 채널 공유 · 알림 워크플로",
+            "status": "available", "priority": "optional",
+            "checked": False,
+            "action": {"type": "auth", "url": "https://chatgpt.com/connectors", "mcp_id": "slack"},
+            "badge": "인증 필요", "needs_auth": True,
+        },
+    ]
+
+    return {
+        "components": components,
+        "engines": engines,
+        "models":  models,
+        "mcps":    mcps,
+        "summary": {
+            "chip":            chip["name"],
+            "cpu_cores":       cpu.get("logical_cores"),
+            "cpu_instructions": cpu.get("instructions", []),
+            "gpu":             gpu.get("name") or gpu.get("vendor"),
+            "gpu_vendor":      gpu.get("vendor"),
+            "vram_gb":         gpu.get("vram_gb"),
+            "cuda":            cuda.get("available"),
+            "cuda_version":    cuda.get("version"),
+            "wsl":             wsl,
+            "preferred_engine": preferred_engine,
+            "ram_gb":          ram,
+            "disk_free_gb":    disk_free,
+            "is_apple_silicon": is_apple,
+            "max_model_gb":    round(max_model_gb, 1),
+        },
+    }
+
+# ── Installation Stream ───────────────────────────────────────────────────────
+
+def _verify_action(action: Dict[str, Any]) -> Tuple[bool, str]:
+    atype = action.get("type")
+    if atype == "pip":
+        modules = action.get("verify_modules") or [_package_module(pkg) for pkg in action.get("packages", [])]
+        missing = [module for module in modules if not _module_available(module)]
+        if missing:
+            return False, "Python 모듈 감지 실패: " + ", ".join(missing)
+        return True, "Python 모듈 import 테스트 통과"
+    binary = action.get("binary")
+    if binary:
+        return _verify_binary(binary)
+    return True, "검증 항목 없음"
+
+
+async def _repair_action(action: Dict[str, Any]) -> Tuple[bool, str]:
+    binary = action.get("binary")
+    if binary:
+        repair_path_for(binary)
+        ok, msg = _verify_binary(binary)
+        if ok:
+            return True, f"PATH 자동 보정 완료: {msg}"
+    if action.get("type") == "pip":
+        packages = action.get("packages", [])
+        if packages:
+            for pkg in packages:
+                success, err = await _pip_install(pkg)
+                if not success:
+                    return False, err
+            return _verify_action(action)
+    return False, "자동 복구 방법을 찾지 못했습니다."
+
+
+async def install_stream(items: List[Dict], router: Any) -> AsyncIterator[str]:
+    for item in items:
+        item_id    = item.get("id", "unknown")
+        name       = item.get("name", item_id)
+        action     = item.get("action") or {}
+        atype      = action.get("type")
+
+        if not atype:
+            yield _sse({"id": item_id, "status": "skipped", "msg": f"{name} — 이미 준비됨"})
+            await asyncio.sleep(0.04)
+            continue
+
+        yield _sse({"id": item_id, "status": "starting", "msg": f"{name} 준비 중..."})
+
+        if atype == "pip":
+            packages = action.get("packages", [])
+            ok = True
+            for pkg in packages:
+                yield _sse({"id": item_id, "status": "running", "msg": f"pip install {pkg} ..."})
+                success, err = await _pip_install(pkg)
+                if success:
+                    yield _sse({"id": item_id, "status": "progress", "msg": f"{pkg} 설치 완료"})
+                else:
+                    yield _sse({"id": item_id, "status": "error", "msg": f"{pkg} 실패: {err[:400]}"})
+                    ok = False
+                    break
+            if ok:
+                yield _sse({"id": item_id, "status": "running", "msg": f"{name} 동작 테스트 중..."})
+                verified, detail = _verify_action(action)
+                if verified:
+                    yield _sse({"id": item_id, "status": "done", "msg": f"{name} 설치 · 검증 완료 ✅\n{detail}"})
+                else:
+                    yield _sse({"id": item_id, "status": "running", "msg": f"검증 실패 — 자동 복구 중...\n{detail}"})
+                    repaired, repair_msg = await _repair_action(action)
+                    yield _sse({"id": item_id, "status": "done" if repaired else "error", "msg": repair_msg[:500]})
+
+        elif atype == "brew":
+            pkg = action.get("package", "")
+            yield _sse({"id": item_id, "status": "running", "msg": f"brew install {pkg} ..."})
+            success, err = await _brew_install(pkg)
+            if success:
+                yield _sse({"id": item_id, "status": "running", "msg": "설치 완료 감지 · PATH 보정 중..."})
+                binary = action.get("binary")
+                if binary:
+                    repair_path_for(binary)
+                verified, detail = _verify_action(action)
+                if verified:
+                    yield _sse({"id": item_id, "status": "done", "msg": f"{name} 설치 · 연결 · 검증 완료 ✅\n{detail}"})
+                else:
+                    yield _sse({"id": item_id, "status": "running", "msg": f"검증 실패 — 자동 복구 중...\n{detail}"})
+                    repaired, repair_msg = await _repair_action(action)
+                    yield _sse({"id": item_id, "status": "done" if repaired else "error", "msg": repair_msg[:500]})
+            else:
+                url = action.get("official_url") or action.get("url")
+                if url:
+                    yield _sse({"id": item_id, "status": "auth", "msg": f"자동 설치 실패 — 공식 다운로드 페이지를 엽니다.\n{err[:240]}", "auth_url": url})
+                    open_url(url)
+                yield _sse({"id": item_id, "status": "error", "msg": f"실패: {err[:400]}"})
+
+        elif atype == "load_model":
+            model_id = action.get("model_id", "")
+            yield _sse({"id": item_id, "status": "running",
+                        "msg": f"모델 다운로드 · 로딩 중...\n{model_id}\n(용량에 따라 수 분 소요)"})
+            try:
+                msg = await router.load_model(model_id)
+                yield _sse({"id": item_id, "status": "done", "msg": f"{name} 로드 완료 ✅"})
+            except Exception as e:
+                yield _sse({"id": item_id, "status": "error", "msg": f"로드 실패: {str(e)[:400]}"})
+
+        elif atype == "auth":
+            url = action.get("url", "")
+            yield _sse({"id": item_id, "status": "auth",
+                        "msg": "브라우저에서 인증 페이지를 엽니다...", "auth_url": url})
+            open_url(url)
+            yield _sse({"id": item_id, "status": "waiting",
+                        "msg": "브라우저에서 인증 완료 후 계속하세요"})
+
+        elif atype == "url":
+            url = action.get("url", "")
+            yield _sse({"id": item_id, "status": "auth",
+                        "msg": "설치 페이지를 브라우저에서 엽니다...", "auth_url": url})
+            open_url(url)
+            binary = action.get("binary")
+            if binary:
+                yield _sse({"id": item_id, "status": "waiting",
+                            "msg": f"{binary} 설치 완료를 자동 감지하는 중입니다..."})
+                ok, detail = await _wait_for_binary(binary)
+                if ok:
+                    repair_path_for(binary)
+                    yield _sse({"id": item_id, "status": "done",
+                                "msg": f"{name} 설치 · PATH 연결 · 검증 완료 ✅\n{detail}"})
+                else:
+                    yield _sse({"id": item_id, "status": "error",
+                                "msg": f"{detail}\n공식 페이지에서 설치 후 다시 시도하세요."})
+            else:
+                yield _sse({"id": item_id, "status": "waiting",
+                            "msg": "브라우저에서 설치 또는 인증을 완료한 뒤 다시 시도하세요"})
+
+        else:
+            yield _sse({"id": item_id, "status": "error", "msg": f"알 수 없는 액션: {atype}"})
+
+    yield _sse({"status": "complete", "msg": "모든 항목 처리 완료!"})
+
+
+async def _pip_install(package: str) -> Tuple[bool, str]:
+    try:
+        proc = await asyncio.create_subprocess_exec(
+            sys.executable, "-m", "pip", "install", "--upgrade", package,
+            stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE,
+        )
+        _, stderr = await asyncio.wait_for(proc.communicate(), timeout=600)
+        if proc.returncode == 0:
+            return True, ""
+        return False, stderr.decode(errors="replace")
+    except asyncio.TimeoutError:
+        return False, "설치 시간 초과 (10분)"
+    except Exception as e:
+        return False, str(e)
+
+
+async def _brew_install(package: str) -> Tuple[bool, str]:
+    brew = shutil.which("brew")
+    if not brew:
+        return False, "Homebrew 미설치 — https://brew.sh 에서 설치하세요"
+    try:
+        proc = await asyncio.create_subprocess_exec(
+            brew, "install", package,
+            stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE,
+        )
+        _, stderr = await asyncio.wait_for(proc.communicate(), timeout=300)
+        if proc.returncode == 0:
+            return True, ""
+        return False, stderr.decode(errors="replace")
+    except asyncio.TimeoutError:
+        return False, "설치 시간 초과 (5분)"
+    except Exception as e:
+        return False, str(e)
+
+
+def open_url(url: str) -> None:
+    try:
+        system = platform.system()
+        if system == "Darwin":
+            subprocess.Popen(["open", url])
+        elif system == "Windows":
+            subprocess.Popen(["start", "", url], shell=True)
+        else:
+            subprocess.Popen(["xdg-open", url])
+    except Exception:
+        pass