Add runtime environment readiness checks

2026-06-30 14:28:49 +08:00
parent 442b521705
commit d9ea249ff0
12 changed files with 603 additions and 18 deletions
--- a/README.md
+++ b/README.md
@@ -17,6 +17,7 @@ core, then adds:
 Seg_Data_Server_Net/
  backend/            FastAPI API, job runner, module wrappers
  frontend/           React + Vite operator UI
+  envs/               conda environment specs for task and MMSeg runtimes
  scripts/            helper scripts for running services and syncing weights
  weights/            copied model weights and manifest.json
 ```
@@ -27,6 +28,9 @@ Seg_Data_Server_Net/
 cd Seg_Data_Server_Net
 cp .env.example .env

+# Create or repair the two runtime environments, then verify imports.
+scripts/bootstrap_conda_envs.sh
+
 # Backend. The deployment env is seg_smp so the API and most task wrappers
 # share the same segmentation dependency stack. MMSeg jobs default to the
 # separate SEG_MMSEG_CONDA_ENV because full mmcv wheels must match torch/CUDA.
@@ -59,6 +63,10 @@ The coverage panel calls `GET /api/coverage` and verifies that the user-facing
 scripts from the existing `Seg/` workspace are mapped to web jobs. MMSeg
 vendored internals, docs, build outputs, converters, and config templates are
 classified as supporting artifacts rather than direct web actions.
+The runtime panel calls `GET /api/system/readiness` and verifies the conda
+imports required for the backend/task environment and the full MMSeg/mmcv
+environment. Command-line verification is available with
+`PYTHONPATH=backend conda run -n seg_smp python scripts/verify_runtime_envs.py --refresh`.

 The same panel can run `POST /api/acceptance/smoke`, a lightweight live smoke
 that creates an upload dataset, uploads a label, downloads it through the
@@ -81,14 +89,13 @@ Current `seg_smp` uses `mmcv-lite` because no `torch 2.6/cu124` full `mmcv`
 wheel is available on this machine and `nvcc` is not installed for source
 builds. A dedicated `seg_mmcv` environment is used for MMSeg tasks and has
 `torch 2.1.2+cu121`, `mmcv 2.1.0`, `mmsegmentation 1.2.2`, and NumPy 1.26.
-If rebuilding the environment, keep these versions aligned:
+The reproducible specs live in `envs/seg_smp.yml` and `envs/seg_mmcv.yml`;
+the bootstrap script uses the same pinned package sources:

 ```bash
-conda create -n seg_mmcv python=3.10 -y
-conda run -n seg_mmcv python -m pip install -U pip
-conda run -n seg_mmcv python -m pip install torch==2.1.2 torchvision==0.16.2 --index-url https://download.pytorch.org/whl/cu121
-conda run -n seg_mmcv python -m pip install mmengine==0.10.7 mmsegmentation==1.2.2 'mmcv==2.1.0' -f https://download.openmmlab.com/mmcv/dist/cu121/torch2.1/index.html
-conda run -n seg_mmcv python -m pip install 'numpy<2' 'opencv-python<4.12' ftfy regex matplotlib pandas scikit-learn scipy seaborn tqdm tensorboard
+scripts/bootstrap_conda_envs.sh all
+scripts/bootstrap_conda_envs.sh task
+scripts/bootstrap_conda_envs.sh mmseg
 ```

 ## Weight Sync
@@ -149,8 +156,9 @@ PYTHONPATH=backend conda run -n seg_smp python scripts/run_agents.py --build
 ```

 The validation agent checks catalog coverage, the `seg_smp` task env, the
-`seg_mmcv` MMSeg env, GPU visibility, no-weight Git safety, backend tests,
-frontend build, and live backend/frontend endpoints when the services are
-running. With live validation enabled it also runs the lightweight acceptance
-smoke above. By default it also runs the deep training acceptance; set
-`SEG_VALIDATE_DEEP=0` when a quick non-training validation pass is needed.
+`seg_mmcv` MMSeg env, runtime import readiness, GPU visibility, no-weight Git
+safety, backend tests, frontend build, and live backend/frontend endpoints
+when the services are running. With live validation enabled it also runs the
+lightweight acceptance smoke above. By default it also runs the deep training
+acceptance; set `SEG_VALIDATE_DEEP=0` when a quick non-training validation
+pass is needed.
--- a/backend/app/agents/evaluation_agent.py
+++ b/backend/app/agents/evaluation_agent.py
@@ -43,9 +43,13 @@ def evaluate_project() -> dict:
        "dataset_quality_ui": "DatasetQuality" in frontend_text and "generateSelectedYoloYaml" in frontend_text,
        "loss_result_ui": "loss" in frontend_text.lower() and "heatmap" in frontend_text.lower() and "CurvePanel" in frontend_text,
        "job_progress_ui": "JobProgressBar" in frontend_text and "progressTrack" in frontend_text,
+        "runtime_readiness_ui": "runtimeReadiness" in frontend_text and "环境就绪" in frontend_text,
        "dataset_api": "/api/datasets" in backend_text and "api_upload_dataset_files" in backend_text,
        "dataset_quality_api": "/api/datasets/{dataset_name}/validate" in backend_text and "/api/datasets/{dataset_name}/yolo-yaml" in backend_text,
        "job_progress_api": "progress_from_log_path" in backend_text and '"progress"' in backend_text,
+        "runtime_readiness_api": "/api/system/readiness" in backend_text,
+        "runtime_bootstrap_scripts": (settings.project_root / "scripts" / "bootstrap_conda_envs.sh").exists()
+        and (settings.project_root / "scripts" / "verify_runtime_envs.py").exists(),
        "curve_api": "/api/results/curves" in backend_text,
        "deep_acceptance_api": "/api/acceptance/deep" in backend_text,
        "deep_acceptance_ui": "runDeepAcceptance" in frontend_text and "深度训练" in frontend_text,
--- a/backend/app/agents/validation_agent.py
+++ b/backend/app/agents/validation_agent.py
@@ -13,7 +13,7 @@ from ..catalog import get_catalog
 from ..config import settings
 from ..coverage import get_coverage_report
 from ..modules.results.service import scan_training_curves
-from ..modules.system.service import get_conda_envs, get_gpus
+from ..modules.system.service import get_conda_envs, get_gpus, get_runtime_readiness
 from ..modules.weights.service import load_manifest
 from ..progress import parse_progress

@@ -64,6 +64,8 @@ def validate_project(run_build: bool = False) -> dict:
    env_names = [item["name"] for item in get_conda_envs().get("envs", [])]
    checks.append({"name": "task_env_exists", "passed": settings.task_conda_env in env_names, "detail": {"env": settings.task_conda_env}})
    checks.append({"name": "mmseg_env_exists", "passed": settings.mmseg_conda_env in env_names, "detail": {"env": settings.mmseg_conda_env}})
+    runtime_readiness = get_runtime_readiness(force=True)
+    checks.append({"name": "runtime_env_readiness", "passed": runtime_readiness["passed"], "detail": runtime_readiness})

    smoke = _run(
        [
@@ -102,6 +104,7 @@ def validate_project(run_build: bool = False) -> dict:
        health = _fetch(f"{backend_url}/api/health")
        datasets = _fetch(f"{backend_url}/api/datasets")
        live_jobs = _fetch(f"{backend_url}/api/jobs")
+        live_readiness = _fetch(f"{backend_url}/api/system/readiness")
        live_coverage = _fetch(f"{backend_url}/api/coverage")
        live_curves = _fetch(f"{backend_url}/api/results/curves")
        frontend = _fetch(frontend_url)
@@ -116,6 +119,11 @@ def validate_project(run_build: bool = False) -> dict:
            "passed": live_jobs["passed"] and isinstance(live_job_items, list) and (not live_job_items or "progress" in live_job_items[0]),
            "detail": live_jobs,
        })
+        checks.append({
+            "name": "live_runtime_readiness_api",
+            "passed": live_readiness["passed"] and '"passed":true' in live_readiness.get("body", "").replace(" ", ""),
+            "detail": live_readiness,
+        })
        checks.append({"name": "live_coverage_api", "passed": live_coverage["passed"] and '"task_build_passed":true' in live_coverage.get("body", "").replace(" ", ""), "detail": live_coverage})
        checks.append({"name": "live_training_curves_api", "passed": live_curves["passed"] and live_curves.get("body", "").lstrip().startswith("["), "detail": live_curves})
        checks.append({"name": "live_frontend_index", "passed": frontend["passed"] and "Seg Data Server" in frontend.get("body", ""), "detail": frontend})
--- a/backend/app/main.py
+++ b/backend/app/main.py
@@ -15,7 +15,7 @@ from .config import settings
 from .coverage import get_coverage_report
 from .jobs import cancel_job, create_job
 from .modules.results.service import scan_results, scan_training_curves
-from .modules.system.service import disk_usage, get_conda_envs, get_gpus
+from .modules.system.service import disk_usage, get_conda_envs, get_gpus, get_runtime_readiness
 from .modules.dataset.service import create_dataset, generate_yolo_dataset_yaml, list_uploaded_datasets, save_upload, validate_dataset
 from .modules.weights.service import load_manifest, sync_weights, verify_weights
 from .agents.evaluation_agent import evaluate_project
@@ -71,6 +71,11 @@ def api_envs() -> dict:
    return get_conda_envs()


+@app.get("/api/system/readiness")
+def api_runtime_readiness(refresh: bool = False) -> dict:
+    return get_runtime_readiness(force=refresh)
+
+
@app.get("/api/catalog")
 def api_catalog() -> dict:
    return get_catalog()
--- a/backend/app/modules/system/service.py
+++ b/backend/app/modules/system/service.py
@@ -1,11 +1,70 @@
 from __future__ import annotations

+import json
 import shutil
 import subprocess
+import threading
+import time
+from contextlib import contextmanager
 from pathlib import Path
+from typing import Any
+
+import fcntl

 from ...config import settings

+READINESS_CACHE_SECONDS = 300
+_readiness_cache: tuple[float, dict[str, Any]] | None = None
+_readiness_thread_lock = threading.Lock()
+
+
+@contextmanager
+def readiness_probe_lock():
+    settings.project_root.joinpath("var").mkdir(parents=True, exist_ok=True)
+    lock_path = settings.project_root / "var" / "runtime_readiness.lock"
+    with _readiness_thread_lock, lock_path.open("w") as lock_file:
+        fcntl.flock(lock_file, fcntl.LOCK_EX)
+        try:
+            yield
+        finally:
+            fcntl.flock(lock_file, fcntl.LOCK_UN)
+
+
+def runtime_environment_specs() -> list[dict[str, Any]]:
+    return [
+        {
+            "role": "backend_task",
+            "name": settings.task_conda_env,
+            "label": "Backend, SegModel, YOLO",
+            "env_file": "envs/seg_smp.yml",
+            "required_imports": [
+                {"module": "fastapi", "package": "fastapi"},
+                {"module": "uvicorn", "package": "uvicorn"},
+                {"module": "torch", "package": "torch"},
+                {"module": "cv2", "package": "opencv-python"},
+                {"module": "segmentation_models_pytorch", "package": "segmentation-models-pytorch"},
+                {"module": "ultralytics", "package": "ultralytics"},
+                {"module": "albumentations", "package": "albumentations"},
+                {"module": "mmengine", "package": "mmengine"},
+                {"module": "mmseg", "package": "mmsegmentation"},
+            ],
+        },
+        {
+            "role": "mmseg_full",
+            "name": settings.mmseg_conda_env,
+            "label": "MMSeg full mmcv runtime",
+            "env_file": "envs/seg_mmcv.yml",
+            "required_imports": [
+                {"module": "torch", "package": "torch"},
+                {"module": "cv2", "package": "opencv-python"},
+                {"module": "mmcv", "package": "mmcv"},
+                {"module": "mmcv._ext", "package": "mmcv"},
+                {"module": "mmengine", "package": "mmengine"},
+                {"module": "mmseg", "package": "mmsegmentation"},
+            ],
+        },
+    ]
+

 def parse_nvidia_smi_csv(output: str) -> list[dict]:
    gpus: list[dict] = []
@@ -63,6 +122,152 @@ def get_conda_envs() -> dict:
    return {"available": True, "envs": envs, "task_default": settings.task_conda_env, "mmseg_default": settings.mmseg_conda_env}


+def probe_code(required_imports: list[dict[str, str]]) -> str:
+    imports_json = json.dumps(required_imports, ensure_ascii=False)
+    return f"""
+import importlib
+import importlib.metadata as metadata
+import json
+import platform
+import sys
+
+required = {imports_json}
+checks = []
+for item in required:
+    module_name = item["module"]
+    package_name = item.get("package") or module_name
+    try:
+        module = importlib.import_module(module_name)
+        version = getattr(module, "__version__", None)
+        if version is None:
+            try:
+                version = metadata.version(package_name)
+            except Exception:
+                version = None
+        checks.append({{
+            "module": module_name,
+            "package": package_name,
+            "passed": True,
+            "version": str(version) if version is not None else None,
+        }})
+    except Exception as exc:
+        checks.append({{
+            "module": module_name,
+            "package": package_name,
+            "passed": False,
+            "error": f"{{type(exc).__name__}}: {{exc}}",
+        }})
+
+extra = {{"python": sys.version.split()[0], "platform": platform.platform()}}
+try:
+    import torch
+    extra["torch_cuda_available"] = bool(torch.cuda.is_available())
+    extra["torch_cuda"] = getattr(torch.version, "cuda", None)
+except Exception:
+    pass
+print(json.dumps({{"checks": checks, "extra": extra}}, ensure_ascii=False))
+"""
+
+
+def parse_probe_stdout(stdout: str) -> dict[str, Any]:
+    for line in reversed(stdout.splitlines()):
+        text = line.strip()
+        if not text or not text.startswith("{"):
+            continue
+        return json.loads(text)
+    raise ValueError("probe did not emit JSON")
+
+
+def inspect_conda_env(env_name: str, required_imports: list[dict[str, str]], timeout: int = 45, retries: int = 1) -> dict[str, Any]:
+    command = ["conda", "run", "-n", env_name, "python", "-c", probe_code(required_imports)]
+    attempts = []
+    result: subprocess.CompletedProcess[str] | None = None
+    for attempt in range(retries + 1):
+        result = subprocess.run(command, capture_output=True, text=True, timeout=timeout)
+        attempts.append({"attempt": attempt + 1, "returncode": result.returncode})
+        if result.returncode == 0 or result.returncode not in {139, -11} or attempt >= retries:
+            break
+        time.sleep(2)
+    if result is None:
+        raise RuntimeError("probe did not run")
+    detail = {
+        "command": command[:4] + ["..."],
+        "returncode": result.returncode,
+        "attempts": attempts,
+        "stdout_tail": result.stdout[-2000:],
+        "stderr_tail": result.stderr[-2000:],
+    }
+    if result.returncode != 0:
+        return {"passed": False, "checks": [], "extra": {}, "detail": detail}
+    try:
+        parsed = parse_probe_stdout(result.stdout)
+    except Exception as exc:
+        return {"passed": False, "checks": [], "extra": {}, "detail": {**detail, "error": str(exc)}}
+    checks = parsed.get("checks", [])
+    return {
+        "passed": bool(checks) and all(item.get("passed") for item in checks),
+        "checks": checks,
+        "extra": parsed.get("extra", {}),
+        "detail": detail,
+    }
+
+
+def get_runtime_readiness(force: bool = False) -> dict[str, Any]:
+    global _readiness_cache
+    now = time.time()
+    if not force and _readiness_cache and now - _readiness_cache[0] < READINESS_CACHE_SECONDS:
+        cached = dict(_readiness_cache[1])
+        cached["cached"] = True
+        return cached
+
+    with readiness_probe_lock():
+        now = time.time()
+        if not force and _readiness_cache and now - _readiness_cache[0] < READINESS_CACHE_SECONDS:
+            cached = dict(_readiness_cache[1])
+            cached["cached"] = True
+            return cached
+
+        conda = get_conda_envs()
+        env_paths = {item["name"]: item["path"] for item in conda.get("envs", [])}
+        envs: list[dict[str, Any]] = []
+        for spec in runtime_environment_specs():
+            env_name = spec["name"]
+            exists = env_name in env_paths
+            env_report: dict[str, Any] = {
+                "role": spec["role"],
+                "name": env_name,
+                "label": spec["label"],
+                "env_file": spec["env_file"],
+                "path": env_paths.get(env_name),
+                "exists": exists,
+                "passed": False,
+                "checks": [],
+                "extra": {},
+            }
+            if exists:
+                probe = inspect_conda_env(env_name, spec["required_imports"])
+                env_report.update(probe)
+            envs.append(env_report)
+
+        payload = {
+            "available": bool(conda.get("available")),
+            "passed": bool(conda.get("available")) and all(item["passed"] for item in envs),
+            "generated_at": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime(now)),
+            "cache_seconds": READINESS_CACHE_SECONDS,
+            "cached": False,
+            "envs": envs,
+            "specs": {
+                "bootstrap_script": "scripts/bootstrap_conda_envs.sh",
+                "verify_script": "scripts/verify_runtime_envs.py",
+                "env_files": [spec["env_file"] for spec in runtime_environment_specs()],
+                "task_default": settings.task_conda_env,
+                "mmseg_default": settings.mmseg_conda_env,
+            },
+        }
+        _readiness_cache = (now, payload)
+        return payload
+
+
 def disk_usage() -> dict:
    usage = shutil.disk_usage(settings.source_root)
    return {
--- a/backend/tests/test_system_service.py
+++ b/backend/tests/test_system_service.py
@@ -1,4 +1,5 @@
-from app.modules.system.service import parse_nvidia_smi_csv
+from app.modules.system import service
+from app.modules.system.service import get_runtime_readiness, parse_nvidia_smi_csv, parse_probe_stdout


 def test_parse_nvidia_smi_csv():
@@ -16,3 +17,57 @@ def test_parse_nvidia_smi_csv():
        }
    ]

+
+def test_parse_probe_stdout_uses_last_json_line():
+    parsed = parse_probe_stdout(
+        "warning before json\n"
+        '{"checks":[{"module":"torch","passed":true,"version":"2.6.0"}],"extra":{"python":"3.11"}}\n'
+    )
+
+    assert parsed["checks"][0]["module"] == "torch"
+    assert parsed["extra"]["python"] == "3.11"
+
+
+def test_runtime_readiness_marks_missing_env(monkeypatch):
+    monkeypatch.setattr(service, "_readiness_cache", None)
+    monkeypatch.setattr(service, "get_conda_envs", lambda: {"available": True, "envs": []})
+
+    readiness = get_runtime_readiness(force=True)
+
+    assert readiness["passed"] is False
+    assert all(not item["exists"] for item in readiness["envs"])
+
+
+def test_runtime_readiness_aggregates_probe_results(monkeypatch):
+    monkeypatch.setattr(service, "_readiness_cache", None)
+    specs = [
+        {"role": "task", "name": "seg_smp", "label": "task", "env_file": "envs/seg_smp.yml", "required_imports": []},
+        {"role": "mmseg", "name": "seg_mmcv", "label": "mmseg", "env_file": "envs/seg_mmcv.yml", "required_imports": []},
+    ]
+    monkeypatch.setattr(service, "runtime_environment_specs", lambda: specs)
+    monkeypatch.setattr(
+        service,
+        "get_conda_envs",
+        lambda: {
+            "available": True,
+            "envs": [
+                {"name": "seg_smp", "path": "/envs/seg_smp"},
+                {"name": "seg_mmcv", "path": "/envs/seg_mmcv"},
+            ],
+        },
+    )
+    monkeypatch.setattr(
+        service,
+        "inspect_conda_env",
+        lambda name, imports: {
+            "passed": True,
+            "checks": [{"module": "torch", "passed": True}],
+            "extra": {"python": "3.11"},
+        },
+    )
+
+    readiness = get_runtime_readiness(force=True)
+
+    assert readiness["passed"] is True
+    assert readiness["envs"][0]["path"] == "/envs/seg_smp"
+    assert readiness["specs"]["env_files"] == ["envs/seg_smp.yml", "envs/seg_mmcv.yml"]
--- a/envs/seg_mmcv.yml
+++ b/envs/seg_mmcv.yml
@@ -0,0 +1,27 @@
+name: seg_mmcv
+channels:
+  - conda-forge
+  - defaults
+dependencies:
+  - python=3.10
+  - pip
+  - pip:
+      - --index-url https://download.pytorch.org/whl/cu121
+      - torch==2.1.2
+      - torchvision==0.16.2
+      - --extra-index-url https://pypi.org/simple
+      - mmengine==0.10.7
+      - mmsegmentation==1.2.2
+      - -f https://download.openmmlab.com/mmcv/dist/cu121/torch2.1/index.html
+      - mmcv==2.1.0
+      - numpy<2
+      - opencv-python<4.12
+      - ftfy
+      - regex
+      - matplotlib
+      - pandas
+      - scikit-learn
+      - scipy
+      - seaborn
+      - tqdm
+      - tensorboard
--- a/envs/seg_smp.yml
+++ b/envs/seg_smp.yml
@@ -0,0 +1,31 @@
+name: seg_smp
+channels:
+  - conda-forge
+  - defaults
+dependencies:
+  - python=3.11
+  - pip
+  - pip:
+      - --extra-index-url https://download.pytorch.org/whl/cu124
+      - fastapi>=0.110
+      - uvicorn[standard]>=0.27
+      - pydantic>=2
+      - python-multipart>=0.0.9
+      - pytest>=8
+      - torch==2.6.0
+      - torchvision==0.21.0
+      - opencv-python<4.12
+      - numpy<2
+      - albumentations
+      - segmentation-models-pytorch
+      - ultralytics
+      - mmengine
+      - mmsegmentation==1.2.2
+      - mmcv-lite
+      - matplotlib
+      - pandas
+      - scikit-learn
+      - scipy
+      - seaborn
+      - tqdm
+      - tensorboard
--- a/frontend/src/main.tsx
+++ b/frontend/src/main.tsx
@@ -157,6 +157,42 @@ type GpuPayload = {
  }>;
 };

+type RuntimeCheck = {
+  module: string;
+  package?: string;
+  passed: boolean;
+  version?: string | null;
+  error?: string;
+};
+
+type RuntimeEnv = {
+  role: string;
+  name: string;
+  label: string;
+  env_file: string;
+  path?: string;
+  exists: boolean;
+  passed: boolean;
+  checks: RuntimeCheck[];
+  extra: Record<string, unknown>;
+};
+
+type RuntimeReadinessPayload = {
+  available: boolean;
+  passed: boolean;
+  generated_at: string;
+  cached: boolean;
+  cache_seconds: number;
+  envs: RuntimeEnv[];
+  specs: {
+    bootstrap_script: string;
+    verify_script: string;
+    env_files: string[];
+    task_default: string;
+    mmseg_default: string;
+  };
+};
+
 async function api<T>(path: string, init?: RequestInit): Promise<T> {
  const res = await fetch(`${API_BASE}${path}`, {
    headers: { "Content-Type": "application/json" },
@@ -229,13 +265,15 @@ function useData() {
  const [coverage, setCoverage] = useState<CoveragePayload | null>(null);
  const [acceptance, setAcceptance] = useState<AcceptancePayload | null>(null);
  const [deepAcceptance, setDeepAcceptance] = useState<DeepAcceptancePayload | null>(null);
+  const [runtimeReadiness, setRuntimeReadiness] = useState<RuntimeReadinessPayload | null>(null);
  const [error, setError] = useState<string>("");

  async function refresh() {
    try {
-      const [catalogNext, gpusNext, jobsNext, resultsNext, curvesNext, datasetsNext, coverageNext, acceptanceNext, deepAcceptanceNext] = await Promise.all([
+      const [catalogNext, gpusNext, readinessNext, jobsNext, resultsNext, curvesNext, datasetsNext, coverageNext, acceptanceNext, deepAcceptanceNext] = await Promise.all([
        api<Catalog>("/api/catalog"),
        api<GpuPayload>("/api/system/gpus"),
+        api<RuntimeReadinessPayload>("/api/system/readiness"),
        api<Job[]>("/api/jobs"),
        api<ResultItem[]>("/api/results"),
        api<TrainingCurve[]>("/api/results/curves"),
@@ -246,6 +284,7 @@ function useData() {
      ]);
      setCatalog(catalogNext);
      setGpus(gpusNext);
+      setRuntimeReadiness(readinessNext);
      setJobs(jobsNext);
      setResults(resultsNext.slice(0, 80));
      setCurves(curvesNext.slice(0, 12));
@@ -277,7 +316,7 @@ function useData() {
    return () => window.clearInterval(timer);
  }, []);

-  return { catalog, gpus, jobs, results, curves, datasets, datasetValidations, coverage, acceptance, deepAcceptance, error, refresh };
+  return { catalog, gpus, runtimeReadiness, jobs, results, curves, datasets, datasetValidations, coverage, acceptance, deepAcceptance, error, refresh };
 }

 function StatusPill({ status }: { status: string }) {
@@ -300,7 +339,7 @@ function JobProgressBar({ progress }: { progress?: JobProgress }) {
 }

 function App() {
-  const { catalog, gpus, jobs, results, curves, datasets, datasetValidations, coverage, acceptance, deepAcceptance, error, refresh } = useData();
+  const { catalog, gpus, runtimeReadiness, jobs, results, curves, datasets, datasetValidations, coverage, acceptance, deepAcceptance, error, refresh } = useData();
  const [taskType, setTaskType] = useState("mock.echo");
  const [params, setParams] = useState(JSON.stringify(defaultParams["mock.echo"], null, 2));
  const [selectedJob, setSelectedJob] = useState<Job | null>(null);
@@ -492,6 +531,7 @@ function App() {
          <a href="#jobs"><Terminal size={18} />任务</a>
          <a href="#datasets"><Boxes size={18} />数据集</a>
          <a href="#gpu"><Cpu size={18} />GPU</a>
+          <a href="#runtime"><ShieldCheck size={18} />环境</a>
          <a href="#coverage"><ClipboardCheck size={18} />覆盖</a>
          <a href="#weights"><HardDrive size={18} />权重</a>
          <a href="#results"><BarChart3 size={18} />结果</a>
@@ -768,7 +808,7 @@ function App() {
          </div>
        </section>

-        <section className="grid three">
+        <section className="grid four">
          <div className="panel" id="gpu">
            <div className="panelHead">
              <div>
@@ -789,6 +829,37 @@ function App() {
            ))}
          </div>

+          <div className="panel" id="runtime">
+            <div className="panelHead">
+              <div>
+                <p className="eyebrow">Runtime</p>
+                <h2>环境就绪</h2>
+              </div>
+              <ShieldCheck size={22} />
+            </div>
+            <div className="envList">
+              {(runtimeReadiness?.envs ?? []).map((env) => (
+                <div key={env.role} className={`envCard ${env.passed ? "ok" : "bad"}`}>
+                  <div className="envHead">
+                    <div>
+                      <strong>{env.name}</strong>
+                      <small>{env.label}</small>
+                    </div>
+                    <span>{env.passed ? "READY" : env.exists ? "CHECK" : "MISSING"}</span>
+                  </div>
+                  <div className="envChecks">
+                    {env.checks.slice(0, 8).map((check) => (
+                      <span key={check.module} className={check.passed ? "ok" : "bad"} title={check.error ?? check.package}>
+                        {check.module}{check.version ? ` ${check.version}` : ""}
+                      </span>
+                    ))}
+                  </div>
+                </div>
+              ))}
+            </div>
+            <p className="muted">{runtimeReadiness?.passed ? "runtime imports ready" : "run scripts/bootstrap_conda_envs.sh"} · {runtimeReadiness?.generated_at ?? "not checked"}</p>
+          </div>
+
          <div className="panel" id="weights">
            <div className="panelHead">
              <div>
--- a/frontend/src/styles.css
+++ b/frontend/src/styles.css
@@ -219,6 +219,11 @@ h2 {
  margin-bottom: 16px;
 }

+.grid.four {
+  grid-template-columns: repeat(4, minmax(0, 1fr));
+  margin-bottom: 16px;
+}
+
 .panel {
  padding: 18px;
  min-width: 0;
@@ -748,6 +753,87 @@ meter {
  accent-color: var(--green);
 }

+.envList {
+  display: grid;
+  gap: 10px;
+}
+
+.envCard {
+  min-width: 0;
+  display: grid;
+  gap: 8px;
+  padding: 10px;
+  border-radius: 7px;
+  border: 1px solid var(--line);
+  background: #101310;
+}
+
+.envCard.ok {
+  border-color: rgba(157, 226, 111, 0.32);
+}
+
+.envCard.bad {
+  border-color: rgba(240, 113, 103, 0.55);
+}
+
+.envHead {
+  min-width: 0;
+  display: grid;
+  grid-template-columns: minmax(0, 1fr) auto;
+  gap: 8px;
+  align-items: start;
+}
+
+.envHead strong,
+.envHead small {
+  display: block;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.envHead small {
+  color: var(--muted);
+  margin-top: 2px;
+}
+
+.envHead > span {
+  color: var(--green);
+  font-size: 11px;
+  font-weight: 760;
+}
+
+.envCard.bad .envHead > span {
+  color: var(--red);
+}
+
+.envChecks {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 5px;
+}
+
+.envChecks span {
+  max-width: 100%;
+  padding: 4px 6px;
+  border-radius: 5px;
+  border: 1px solid rgba(238, 242, 232, 0.1);
+  color: var(--muted);
+  background: #080a08;
+  font-size: 11px;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.envChecks span.ok {
+  color: var(--green);
+}
+
+.envChecks span.bad {
+  color: var(--red);
+}
+
 .bigNumber {
  font-size: 54px;
  font-weight: 760;
--- a/scripts/bootstrap_conda_envs.sh
+++ b/scripts/bootstrap_conda_envs.sh
@@ -0,0 +1,58 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+TASK_ENV="${SEG_TASK_CONDA_ENV:-seg_smp}"
+MMSEG_ENV="${SEG_MMSEG_CONDA_ENV:-seg_mmcv}"
+
+env_exists() {
+  conda env list | awk '{print $1}' | grep -Fxq "$1"
+}
+
+create_task_env() {
+  if ! env_exists "${TASK_ENV}"; then
+    conda create -n "${TASK_ENV}" python=3.11 -y
+  fi
+  conda run -n "${TASK_ENV}" python -m pip install -U pip
+  conda run -n "${TASK_ENV}" python -m pip install -r "${ROOT_DIR}/backend/requirements.txt"
+  conda run -n "${TASK_ENV}" python -m pip install \
+    torch==2.6.0 torchvision==0.21.0 \
+    'numpy<2' 'opencv-python<4.12' albumentations segmentation-models-pytorch ultralytics \
+    mmengine mmsegmentation==1.2.2 mmcv-lite \
+    matplotlib pandas scikit-learn scipy seaborn tqdm tensorboard
+}
+
+create_mmseg_env() {
+  if ! env_exists "${MMSEG_ENV}"; then
+    conda create -n "${MMSEG_ENV}" python=3.10 -y
+  fi
+  conda run -n "${MMSEG_ENV}" python -m pip install -U pip
+  conda run -n "${MMSEG_ENV}" python -m pip install \
+    torch==2.1.2 torchvision==0.16.2 \
+    --index-url https://download.pytorch.org/whl/cu121
+  conda run -n "${MMSEG_ENV}" python -m pip install \
+    mmengine==0.10.7 mmsegmentation==1.2.2 'mmcv==2.1.0' \
+    -f https://download.openmmlab.com/mmcv/dist/cu121/torch2.1/index.html
+  conda run -n "${MMSEG_ENV}" python -m pip install \
+    'numpy<2' 'opencv-python<4.12' ftfy regex matplotlib pandas scikit-learn scipy seaborn tqdm tensorboard
+}
+
+case "${1:-all}" in
+  all)
+    create_task_env
+    create_mmseg_env
+    PYTHONPATH="${ROOT_DIR}/backend" conda run -n "${TASK_ENV}" python "${ROOT_DIR}/scripts/verify_runtime_envs.py" --refresh
+    ;;
+  task)
+    create_task_env
+    echo "Created or repaired ${TASK_ENV}. Run '$0 all' for full runtime verification."
+    ;;
+  mmseg)
+    create_mmseg_env
+    echo "Created or repaired ${MMSEG_ENV}. Run '$0 all' for full runtime verification."
+    ;;
+  *)
+    echo "usage: $0 [all|task|mmseg]" >&2
+    exit 2
+    ;;
+esac
--- a/scripts/verify_runtime_envs.py
+++ b/scripts/verify_runtime_envs.py
@@ -0,0 +1,27 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+import argparse
+import json
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT / "backend"))
+
+from app.modules.system.service import get_runtime_readiness  # noqa: E402
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Verify Seg Data Server runtime conda environments.")
+    parser.add_argument("--refresh", action="store_true", help="ignore the backend readiness cache")
+    args = parser.parse_args()
+
+    report = get_runtime_readiness(force=args.refresh)
+    print(json.dumps(report, ensure_ascii=False, indent=2))
+    if not report.get("passed"):
+        raise SystemExit(1)
+
+
+if __name__ == "__main__":
+    main()