dsh-scnet 0.6.7 → 0.6.9

package/README.md

@@ -4,15 +4,17 @@
 
 DSH-SCNet (`dsh-scnet` on npm) is a community-maintained DSH bundle for operating Supercomputing Network (SCNet). It packages the source [`scnet-hpc`](https://github.com/lql341/scnet-hpc) Skill, profile-aware utilities, and bounded tools for SSH, Slurm, OpenAPI files, regional resources, and Notebook discovery.
 
-Current release: **0.6.7**
+Current release: **0.6.9**
 
 <!-- scnet-release:start -->
-## 0.6.7 highlights
-
-- Desktop hosts can inject a verified Python 3 interpreter through `DSH_SCNET_PYTHON` or
-  `SCNET_PYTHON`; local entrypoints no longer require the command name `python3`.
-- Shared shell helpers reject Python 2 and invalid configured interpreters before running
-  OpenAPI or local job-generation code.
+## 0.6.9 highlights
+
+- Default CLI output is compact across OpenAPI and SSH, with `--details` and `--raw` for
+  progressively richer responses and bounded internal job waiting to reduce repeated output.
+- OpenAPI completion is confirmed against accounting before it becomes terminal, and SSH job
+  lookup falls back to accounting after a job leaves the live queue.
+- First-use setup defaults to OpenAPI, retains existing backend preferences, and preserves
+  unrelated configuration fields when settings are updated.
 <!-- scnet-release:end -->
 
 > This is an independent community project. It is compatible with DeepSeek Harness but is not an official DeepSeek product and does not imply endorsement, partnership, or authorization by DeepSeek.

package/README_CN.md

@@ -4,14 +4,15 @@
 
 <p align="center">在 DeepSeek Harness 里操作超算互联网（SCNet）：配置 SSH、管理 Slurm 作业、查询 OpenAPI 区域和文件、检查 Notebook 资源与实例，并保留海光 DCU/DTK 诊断能力。</p>
 
-当前版本：**0.6.7**
+当前版本：**0.6.9**
 
 <!-- scnet-release:start -->
-## 0.6.7 更新
+## 0.6.9 更新
 
-- Desktop 主机可通过 `DSH_SCNET_PYTHON` 或 `SCNET_PYTHON` 注入经过校验的 Python 3
-  解释器，本地入口不再强制要求命令名为 `python3`。
-- 共享 shell 辅助函数会在运行 OpenAPI 或本地作业生成前拒绝 Python 2 和无效解释器路径。
+- OpenAPI 与 SSH 的 CLI 默认使用精简输出；需要更多信息时使用 `--details` 或 `--raw`，
+  作业等待改为程序内部有界轮询，减少重复输出。
+- OpenAPI 在作业状态终结前通过账本确认，SSH 作业离开实时队列后自动回退到账本查询。
+- 首次配置默认选择 OpenAPI，已有用户保留原 backend 偏好，更新设置时保留无关配置字段。
 <!-- scnet-release:end -->
 
 > 本项目是独立维护的社区项目，与 DeepSeek Harness 兼容，但不是 DeepSeek 官方产品，也不表示 DeepSeek 对本项目提供背书、合作或授权。

package/connector/credentials.mjs

@@ -0,0 +1,42 @@
+function requiredString(value, label) {
+  const normalized = String(value ?? '').trim()
+  if (!normalized) throw new TypeError(`${label} must be a non-empty string`)
+  return normalized
+}
+
+function assertBackend(backend) {
+  if (!backend || typeof backend !== 'object') {
+    throw new TypeError('credential backend is required')
+  }
+  for (const method of ['get', 'set', 'delete']) {
+    if (typeof backend[method] !== 'function') {
+      throw new TypeError(`credential backend must implement ${method}()`)
+    }
+  }
+  return backend
+}
+
+/**
+ * Adapt a platform keyring implementation without providing a plaintext fallback.
+ * The backend owns encryption, locking, and OS-specific Secret Service behavior.
+ */
+export function createCredentialStoreAdapter({ backend, namespace = 'scnet' } = {}) {
+  const store = assertBackend(backend)
+  const prefix = requiredString(namespace, 'credential namespace')
+  const keyOf = (key) => `${prefix}.${requiredString(key, 'credential key')}`
+  return Object.freeze({
+    async get(key) {
+      return store.get(keyOf(key))
+    },
+    async set(key, value) {
+      if (value === undefined) throw new TypeError('credential value must not be undefined')
+      return store.set(keyOf(key), structuredClone(value))
+    },
+    async delete(key) {
+      return store.delete(keyOf(key))
+    },
+    async clear() {
+      if (typeof store.clear === 'function') return store.clear(prefix)
+    },
+  })
+}

package/connector/errors.mjs

@@ -0,0 +1,58 @@
+const ERROR_CODES = [
+  'AUTHENTICATION_REQUIRED',
+  'AUTHENTICATION_FAILED',
+  'SCOPE_INSUFFICIENT',
+  'NETWORK_TIMEOUT',
+  'REGION_UNAVAILABLE',
+  'PLATFORM_UNSUPPORTED',
+  'CALLBACK_REPLAY',
+  'CALLBACK_MISMATCH',
+  'CONFIGURATION_REQUIRED',
+  'INTERNAL_ERROR',
+]
+
+export const CONNECTOR_ERROR_CODES = Object.freeze(ERROR_CODES)
+
+function requiredString(value, label) {
+  const normalized = String(value ?? '').trim()
+  if (!normalized) throw new TypeError(`${label} must be a non-empty string`)
+  return normalized
+}
+
+export class ConnectorError extends Error {
+  constructor(code, message, { operation, retryable = false } = {}) {
+    if (!CONNECTOR_ERROR_CODES.includes(code)) {
+      throw new RangeError(`unsupported connector error code: ${code}`)
+    }
+    super(requiredString(message, 'error message'))
+    this.name = 'ConnectorError'
+    this.code = code
+    this.operation = operation === undefined ? undefined : requiredString(operation, 'operation')
+    this.retryable = Boolean(retryable)
+  }
+
+  toJSON() {
+    return {
+      ok: false,
+      error_code: this.code,
+      error: this.message,
+      ...(this.operation ? { operation: this.operation } : {}),
+      retryable: this.retryable,
+    }
+  }
+}
+
+export function connectorError(code, message, options) {
+  return new ConnectorError(code, message, options)
+}
+
+export function errorEnvelope(error, { operation } = {}) {
+  if (error instanceof ConnectorError) {
+    return error.toJSON()
+  }
+  return new ConnectorError(
+    'INTERNAL_ERROR',
+    error instanceof Error ? error.message : String(error || 'unknown connector error'),
+    { operation },
+  ).toJSON()
+}

package/connector/index.mjs

@@ -8,10 +8,22 @@ export {
   getOperationDefinition,
 } from './contract.mjs'
 
+export {
+  CONNECTOR_ERROR_CODES,
+  ConnectorError,
+  connectorError,
+  errorEnvelope,
+} from './errors.mjs'
+
+export {
+  createCredentialStoreAdapter,
+} from './credentials.mjs'
+
 export {
   MemoryCredentialStore,
   TokenLifecycle,
   createAuthorizationRequest,
+  createAuthorizationSession,
   createCodeVerifier,
   createPkceChallenge,
   createRefreshTokenRequest,

package/connector/oauth2.mjs

@@ -1,4 +1,5 @@
 import { createHash, randomBytes, randomUUID } from 'node:crypto'
+import { ConnectorError } from './errors.mjs'
 
 const LOOPBACK_HOSTS = new Set(['127.0.0.1', '[::1]', 'localhost'])
 
@@ -130,6 +131,33 @@ export function validateAuthorizationCallback(callbackUrl, request) {
   })
 }
 
+export function createAuthorizationSession(options = {}) {
+  const request = createAuthorizationRequest(options)
+  let consumed = false
+  return Object.freeze({
+    request,
+    get consumed() {
+      return consumed
+    },
+    consumeCallback(callbackUrl) {
+      if (consumed) {
+        throw new ConnectorError(
+          'CALLBACK_REPLAY',
+          'OAuth authorization callback has already been consumed',
+          { operation: 'login' },
+        )
+      }
+      consumed = true
+      try {
+        return validateAuthorizationCallback(callbackUrl, request)
+      } catch (error) {
+        if (error instanceof ConnectorError) throw error
+        throw new ConnectorError('CALLBACK_MISMATCH', error.message, { operation: 'login' })
+      }
+    },
+  })
+}
+
 export function createTokenExchangeRequest({
   tokenEndpoint,
   clientId,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-scnet",
-  "version": "0.6.7",
+  "version": "0.6.9",
   "description": "Community DeepSeek Harness bundle for SCNet SSH and OpenAPI jobs, files, Notebook resources, Slurm diagnostics, and Hygon DCU/DTK guidance.",
   "type": "module",
   "main": "./index.mjs",

package/skills/scnet-hpc/SKILL.md

@@ -12,6 +12,8 @@ available through the API. Never silently switch backends.
 ## First use and selection
 
 - Run `scripts/setup.sh new|modify|status|reset` for explicit configuration lifecycle.
+  OpenAPI setup requires Python 3. With no saved preference, default to OpenAPI; credentials
+  can come from environment variables or the supported OS credential store.
 - Advanced OpenAPI discovery: `python3 scripts/scnet.py setup`.
 - Read-only preflight: `python3 scripts/scnet.py doctor`.
 - Backend precedence: `--backend`, `SCNET_HPC_BACKEND`, saved setup choice, profile
@@ -67,4 +69,6 @@ override examples.
 - `scripts/probe-cluster.sh`: initial profile creation.
 - `scripts/run-compute-probe.sh`: bounded accelerator probe.
 
-Use compact normalized output by default. Request raw backend data only for diagnosis.
+Use compact normalized output by default. Use `--details` for full normalized fields and
+`--raw` only for diagnosis. Prefer bounded `wait` over sending every polling response to the
+model. `FINALIZING` means the OpenAPI accounting state is not yet confirmed.

package/skills/scnet-hpc/VERSION

@@ -1 +1 @@
-0.6.7
+0.6.9

package/skills/scnet-hpc/references/backends.md

@@ -45,7 +45,9 @@ backend.
 
 ## Common CLI
 
-On a new computer, start with the dependency-light Bash panel:
+On a new computer, start with the Bash panel. OpenAPI setup requires Python 3; without it,
+only first-time SSH setup is supported. Updating an existing main configuration requires Python
+to preserve unrelated fields. New users default to OpenAPI; existing backend choices are retained:
 
 ```bash
 ./scripts/setup.sh
@@ -175,3 +177,22 @@ OPENAPI_USERNAME=""
 ```
 
 Keep AK, SK, tokens, cookies, and passwords out of profiles.
+
+## Compact output and waiting
+
+Default job output includes ID, state, queue, and available exit code or failure reason.
+Empty fields are omitted; zero and false are retained. Use `--details` for all normalized
+fields and `--raw` for backend data. `--json` retains its envelope with compact formatting;
+`--dry-run` always retains the complete preview. Consumers needing the previous complete
+CLI payload must pass `--details`; SDK/backend return schemas are preserved.
+
+```bash
+python3 scripts/scnet.py --region <id> job <job-id>
+python3 scripts/scnet.py --details --region <id> job <job-id>
+python3 scripts/scnet.py --region <id> wait <job-id> --interval 3 --wait-timeout 180
+```
+
+`wait` polls internally and emits only the final result or timeout summary. OpenAPI live
+`COMPLETED` can also describe a recently cancelled job: until accounting confirms the state,
+output reports `FINALIZING`. Waiting may therefore include accounting update latency.
+Transfers return the target path and byte count; logs return only the requested lines.

package/skills/scnet-hpc/references/openapi.md

@@ -37,8 +37,9 @@ export SCNET_OPENAPI_ACCESS_KEY="<access-key>"
 export SCNET_OPENAPI_SECRET_KEY="<secret-key>"
 ```
 
-The client signs the canonical AK/timestamp/user JSON with HMAC-SHA256 and obtains a fresh
-per-region token for each invocation. It does not persist tokens. The setup panel can store the
+The client signs the canonical AK/timestamp/user JSON with HMAC-SHA256 when requesting
+region tokens. Tokens are reused from a permission-restricted disk cache for up to four hours
+when available; an expired or unavailable cache triggers authentication. The setup panel can store the
 platform credential set in macOS Keychain or Linux Secret Service; otherwise use environment
 variables. Region IDs, region users, schedulers, and home paths are discovered and cached as
 non-secret metadata.
@@ -78,14 +79,19 @@ python3 scripts/scnet.py --backend openapi --region <id> \
   logs --path /absolute/path/std.out.123 --lines 200
 python3 scripts/scnet.py --backend openapi --region <id> \
   files --path /absolute/path
-python3 scripts/scnet.py --backend openapi --region <id> \
-  mkdir /absolute/work/dir --parents
 ```
 
 If a region exposes multiple schedulers, add `--scheduler-id`.
 
 ## Mutations
 
+Create a remote directory:
+
+```bash
+python3 scripts/scnet.py --backend openapi --region <id> \
+  mkdir /absolute/work/dir --parents
+```
+
 OpenAPI submission uses the documented BASIC/BASE command mode. Extra scheduler directives can
 be repeated:
 
@@ -174,3 +180,22 @@ the public documentation does not establish an idempotency key for submission.
   interactive `srun`, performance profiling, or low-level scheduler diagnosis.
 - API success is determined by both HTTP transport and the JSON `code`; HTTP 200 alone is not
   sufficient.
+
+## Compact output and waiting
+
+Default job output includes ID, state, queue, and available exit code or failure reason.
+Empty fields are omitted; zero and false are retained. Use `--details` for all normalized
+fields and `--raw` for backend data. `--json` retains its envelope with compact formatting;
+`--dry-run` always retains the complete preview. Consumers needing the previous complete
+CLI payload must pass `--details`; SDK/backend return schemas are preserved.
+
+```bash
+python3 scripts/scnet.py --region <id> job <job-id>
+python3 scripts/scnet.py --details --region <id> job <job-id>
+python3 scripts/scnet.py --region <id> wait <job-id> --interval 3 --wait-timeout 180
+```
+
+`wait` polls internally and emits only the final result or timeout summary. OpenAPI live
+`COMPLETED` can also describe a recently cancelled job: until accounting confirms the state,
+output reports `FINALIZING`. Waiting may therefore include accounting update latency.
+Transfers return the target path and byte count; logs return only the requested lines.

package/skills/scnet-hpc/references/quickstart-en.md

@@ -54,7 +54,23 @@ The repository is designed to be shared by Codex, Claude Code, and OpenCode. Ins
 paths depend on the local agent layout; `install.sh --link` is useful when developing the
 skill from a shared source directory.
 
-## 3. Configure SSH
+## 3. Configure OpenAPI and use the default backend
+
+Python 3 is required for OpenAPI setup. On first use the panel selects `new` and OpenAPI;
+on later runs it selects `modify` and retains the saved backend choice. Supply platform AK/SK
+through the panel, environment variables, or a supported OS credential store.
+
+```bash
+./scripts/setup.sh new --backend openapi
+python3 scripts/scnet.py clusters
+python3 scripts/scnet.py --region <region-id> queues
+python3 scripts/scnet.py --region <region-id> jobs
+```
+
+For submission, cancellation, logs, and files, see `references/openapi.md`. These operations
+use OpenAPI by default when no backend preference overrides it.
+
+## 4. Configure SSH for shell workflows
 
 Download the private key from the SCNet console, then run:
 
@@ -66,7 +82,7 @@ ssh <cluster>
 The script installs the key under `~/.ssh`, adds an SSH config entry, and tests the
 connection. Do not commit private keys, usernames, internal node names, or personal paths.
 
-## 4. Generate a Slurm job
+## 5. Generate a Slurm job for the SSH workflow
 
 Accelerator job:
 
@@ -104,7 +120,7 @@ ssh <cluster> 'sbatch --test-only ~/scripts/myjob.slurm'
 ssh <cluster> 'sbatch ~/scripts/myjob.slurm'
 ```
 
-## 5. Monitor and debug
+## 6. Monitor and debug over SSH
 
 ```bash
 squeue -u "$USER"
@@ -119,7 +135,7 @@ ends with the real program exit code.
 For short experiments, request an interactive allocation with `srun`. Use the partition,
 GRES type, CPU count, and memory limit from the target profile.
 
-## 6. SCNet environment rules
+## 7. SCNet environment rules
 
 - Login nodes are for source management, dependency preparation, file operations, and
   scheduler commands. Whether package indexes and model sites are reachable is cluster-specific.
@@ -130,7 +146,7 @@ GRES type, CPU count, and memory limit from the target profile.
 - Loading a module on a login node does not prove that the runtime works on a compute node.
 - Hygon DCU software capabilities depend on the exact DTK, driver, framework, and `gfx` target.
 
-## 7. Refresh and probe a cluster
+## 8. Refresh and probe a cluster
 
 ```bash
 ./scripts/refresh-cluster.sh --cluster <cluster> --dry-run

package/skills/scnet-hpc/references/setup.md

@@ -27,8 +27,9 @@ Host <集群> <主机名>
   Port <端口>
   IdentityFile ~/.ssh/id_rsa_<集群>
   IdentitiesOnly yes
-  # 以下两行仅 macOS 支持，Linux 上会报 unknown option
+  # UseKeychain 仅适用于 Apple OpenSSH；Linux 上省略
   UseKeychain yes
+  # 标准 OpenSSH 选项，需要运行 ssh-agent
   AddKeysToAgent yes
   # 长连接
   ControlMaster auto
@@ -66,7 +67,8 @@ client_global_hostkeys_prove_confirm: server gave bad signature for RSA key 0
 
 | 项 | macOS | Linux |
 |---|---|---|
-| `UseKeychain` / `AddKeysToAgent` | 支持 | **不支持**，报 unknown option |
+| `UseKeychain` | Apple OpenSSH 支持 | 不支持，应省略 |
+| `AddKeysToAgent` | 支持 | 支持，需要运行 ssh-agent |
 | 下载文件隔离属性 | 需 `xattr -c` | 无此概念 |
 | `sed -i` | 需 `sed -i ''`（BSD sed） | `sed -i`（GNU sed） |
 

package/skills/scnet-hpc/scripts/scnet.py

@@ -79,6 +79,11 @@ def build_parser() -> argparse.ArgumentParser:
         action="store_true",
         help="request backend-specific raw fields when supported",
     )
+    parser.add_argument(
+        "--details",
+        action="store_true",
+        help="include full normalized fields and formatted JSON; --raw includes backend data",
+    )
     subparsers = parser.add_subparsers(dest="operation", required=True)
 
     subparsers.add_parser("backends", help="list built-in and discovered backends")
@@ -224,20 +229,122 @@ def choose_backend(
     )
 
 
-def compact_output(operation: str, data: Any) -> str:
+def _without_empty(data: Any) -> Any:
+    if isinstance(data, dict):
+        return {
+            key: _without_empty(value) for key, value in data.items()
+            if value is not None and value != ""
+        }
+    if isinstance(data, list):
+        return [_without_empty(value) for value in data]
+    return data
+
+
+def summarize_output(operation: str, data: Any) -> Any:
+    """Project CLI output only; keep backend/SDK responses available to other callers."""
+
+    def pick(item, keys):
+        return _without_empty({key: item[key] for key in keys if key in item})
+
+    if operation in {"job", "wait"} and isinstance(data, dict):
+        if isinstance(data.get("job"), dict):
+            return {
+                "job": summarize_output("job", data["job"]),
+                **pick(data, ("waited_seconds", "timed_out")),
+            }
+        result = pick(
+            data,
+            (
+                "job_id",
+                "state",
+                "queue",
+                "exit_code",
+                "reason",
+                "state_confirmed",
+                "reported_state",
+                "waited_seconds",
+                "timed_out",
+            ),
+        )
+        if not result.get("queue") and data.get("partition"):
+            result["queue"] = data["partition"]
+        if "state" in result:
+            result["state"] = (
+                str(result["state"])
+                .split(" BY ", 1)[0]
+                .split(" by ", 1)[0]
+                .rstrip("+")
+            )
+        return result
+    if operation == "jobs" and isinstance(data, dict):
+        return {
+            **pick(data, ("total", "scope")),
+            "items": [
+                summarize_output("job", item) for item in data.get("items", [])
+            ],
+        }
+    if operation == "queues" and isinstance(data, list):
+        return [
+            pick(
+                item,
+                (
+                    "partition",
+                    "availability",
+                    "nodes",
+                    "free_nodes",
+                    "free_cpus",
+                    "max_walltime",
+                ),
+            )
+            for item in data
+        ]
+    if operation in {"upload", "download"} and isinstance(data, dict):
+        path = data.get("remote_path")
+        if not path and data.get("remote_dir") and data.get("remote_filename"):
+            path = str(data["remote_dir"]).rstrip("/") + "/" + str(data["remote_filename"])
+        result = {"path": path}
+        size = data.get("bytes")
+        if size is None and data.get("local_path"):
+            local = Path(str(data["local_path"]))
+            if local.is_file():
+                size = local.stat().st_size
+        if size is not None:
+            result["bytes"] = size
+        return _without_empty(result)
+    if operation == "submit" and isinstance(data, dict):
+        return pick(data, ("job_id",))
+    if operation == "cancel" and isinstance(data, dict):
+        return pick(data, ("job_id", "cancelled"))
+    if operation == "mkdir" and isinstance(data, dict):
+        return pick(data, ("path", "created"))
     if operation == "logs" and isinstance(data, dict):
+        return pick(data, ("lines",))
+    # Diagnostic and configuration responses retain their schema.
+    return data
+
+
+def compact_output(operation: str, data: Any, *, detailed: bool = False) -> str:
+    if operation == "logs" and isinstance(data, dict) and "lines" in data:
+        lines = "\n".join(data.get("lines") or [])
+        if not detailed:
+            return lines
         header = (
             f"# {data.get('path')} "
             f"({data.get('direction')}, {len(data.get('lines') or [])} lines)"
         )
-        return header + "\n" + "\n".join(data.get("lines") or [])
+        return header + "\n" + lines
     if operation == "exec" and isinstance(data, dict):
         return str(data.get("stdout") or "")
-    return json.dumps(data, ensure_ascii=False, indent=2, sort_keys=True)
+    return json.dumps(
+        data,
+        ensure_ascii=False,
+        sort_keys=True,
+        **({"indent": 2} if detailed else {"separators": (",", ":")}),
+    )
 
 
 def _job_terminal(state: Any) -> bool:
-    value = str(state or "").upper()
+    value = str(state or "").upper().split(" BY ", 1)[0].rstrip("+")
     return value in {
         "COMPLETED",
         "FAILED",
@@ -245,7 +352,6 @@ def _job_terminal(state: Any) -> bool:
         "TIMEOUT",
         "NODE_ERROR",
         "ABNORMAL",
-        "EXITING",
         "EXITED",
         "DONE",
         "OUT_OF_MEMORY",
@@ -1615,6 +1721,8 @@ def main(argv: list[str] | None = None) -> int:
                 data = backend.execute(args.operation, options)
             else:
                 data = backend.execute(args.operation, options)
+        detailed = args.details or args.raw or args.dry_run
+        output_data = data if detailed else summarize_output(args.operation, data)
         if args.json:
             print(
                 json.dumps(
@@ -1623,15 +1731,15 @@ def main(argv: list[str] | None = None) -> int:
                         "version": VERSION,
                         "backend": backend_name,
                         "operation": args.operation,
-                        "data": data,
+                        "data": output_data,
                     },
                     ensure_ascii=False,
-                    indent=2,
                     sort_keys=True,
+                    **({"indent": 2} if detailed else {"separators": (",", ":")}),
                 )
             )
         elif args.operation != "setup":
-            print(compact_output(args.operation, data))
+            print(compact_output(args.operation, output_data, detailed=detailed))
         if args.operation == "doctor" and not data.get("ok", False):
             return 1
         return 0

package/skills/scnet-hpc/scripts/scnet_backends/openapi.py

@@ -65,8 +65,18 @@ def normalize_job(item: Mapping[str, Any]) -> dict[str, Any]:
         "raw_state": raw_state,
         "queue": item.get("queue") or item.get("queueName"),
         "user": item.get("user") or item.get("userName"),
-        "nodes": item.get("nodeNumReq") or item.get("nodeUsed") or item.get("nodeNum"),
-        "cpus": item.get("procNumReq") or item.get("procNumUsed") or item.get("nproc"),
+        "nodes": (
+            item.get("nodeNumReq")
+            or item.get("nodeUsed")
+            or item.get("nodeNum")
+            or item.get("nodect")
+        ),
+        "cpus": (
+            item.get("procNumReq")
+            or item.get("procNumUsed")
+            or item.get("nproc")
+            or item.get("jobProcNum")
+        ),
         "gpus": item.get("gpuNumReq") or item.get("gpuNumUsed"),
         "dcus": item.get("dcuNumReq") or item.get("dcuNumUsed"),
         "elapsed": item.get("jobRunTime") or item.get("jobWalltimeUsed"),
@@ -78,6 +88,7 @@ def normalize_job(item: Mapping[str, Any]) -> dict[str, Any]:
         "started_at": item.get("jobStartTime"),
         "ended_at": item.get("jobEndTime") or item.get("acctTime"),
         "exit_code": item.get("exitCode"),
+        "platform_exit_status": item.get("jobExitStatus"),
         "reason": item.get("reason"),
         "work_dir": item.get("workDir") or item.get("workdir"),
         "stdout": item.get("outputPath"),
@@ -405,6 +416,32 @@ class OpenAPIBackend(Backend):
             if candidate.get("jobId") or candidate.get("job_id"):
                 result = normalize_job(candidate)
                 result["source"] = "realtime"
+                if result["state"] == "COMPLETED":
+                    # The live service also reports statC immediately after cancellation.
+                    # Prefer accounting evidence; until it exists, do not claim completion.
+                    items, _, history_raw = self._history_jobs(
+                        hpc_url, token, scheduler_id, job_id=job_id, days=30, limit=1
+                    )
+                    history = next(
+                        (
+                            item
+                            for item in items
+                            if str(item.get("jobId") or item.get("job_id"))
+                            == job_id
+                        ),
+                        None,
+                    )
+                    if history is not None:
+                        result = normalize_job(history)
+                        result["source"] = "history"
+                        if options.get("raw"):
+                            result["raw"] = {"item": history, "response": history_raw}
+                        return result
+                    result.update(
+                        state="FINALIZING",
+                        reported_state="COMPLETED",
+                        state_confirmed=False,
+                    )
                 if options.get("raw"):
                     result["raw"] = candidate
                 return result

package/skills/scnet-hpc/scripts/scnet_backends/ssh.py

@@ -146,10 +146,15 @@ class SSHBackend(Backend):
             "cpus",
             "reason_or_nodes",
         ]
-        output = self._ssh(
-            "LC_ALL=C squeue -h "
-            f"-j {shlex.quote(job_id)} -o '%i|%T|%P|%M|%l|%D|%C|%R'"
-        ).strip()
+        try:
+            output = self._ssh(
+                "LC_ALL=C squeue -h "
+                f"-j {shlex.quote(job_id)} -o '%i|%T|%P|%M|%l|%D|%C|%R'"
+            ).strip()
+        except BackendError as exc:
+            if "invalid job id specified" not in str(exc).lower():
+                raise
+            output = ""
         if output:
             return _split_pipe(output.splitlines()[0], fields)
         output = self._ssh(

package/skills/scnet-hpc/scripts/setup.sh

@@ -92,8 +92,11 @@ choose() {
     shift
     local choices=("$@")
     local index choice
+    local default_index="${SCNET_SETUP_CHOICE_DEFAULT:-1}"
+    [[ "$default_index" =~ ^[0-9]+$ ]] || die "默认选项必须是数字"
+    [ "$default_index" -ge 1 ] && [ "$default_index" -le "${#choices[@]}" ] || die "默认选项超出范围"
     if [ -t 0 ] && [ -t 2 ] && [ "${TERM:-dumb}" != "dumb" ]; then
-        local selected=0
+        local selected=$((default_index - 1))
         local number=""
         local key sequence value
         local line_count=$(( ${#choices[@]} + 1 ))
@@ -165,11 +168,11 @@ choose() {
     for index in "${!choices[@]}"; do
         printf '  %d. %s\n' "$((index + 1))" "${choices[$index]}" >&2
     done
-    printf '请输入 1-%d 的数字并按 Enter；直接按 Enter 使用默认项 1。\n' \
-        "${#choices[@]}" >&2
+    printf '请输入 1-%d 的数字并按 Enter；直接按 Enter 使用默认项 %d。\n' \
+        "${#choices[@]}" "$default_index" >&2
     printf '当前终端不支持交互导航，已使用数字输入模式。\n' >&2
     local selected
-    selected=$(ask "选择" "1")
+    selected=$(ask "选择" "$default_index")
     [[ "$selected" =~ ^[0-9]+$ ]] || die "选择必须是数字"
     [ "$selected" -ge 1 ] && [ "$selected" -le "${#choices[@]}" ] \
         || die "选择超出范围"
@@ -195,6 +198,18 @@ json_string() {
 write_config() {
     local config_root="${XDG_CONFIG_HOME:-$HOME/.config}/scnet-hpc"
     local config_path="$config_root/config.json"
+    if [ -n "${PYTHON_BIN:-}" ]; then
+        "$PYTHON_BIN" - "$SCRIPT_DIR" "$CLUSTER" "$DEFAULT_BACKEND" <<'PYCONFIG'
+import sys
+sys.path.insert(0, sys.argv[1])
+from scnet_config import load_user_config, save_user_config
+config = load_user_config()
+config.update(version=1, cluster=sys.argv[2], default_backend=sys.argv[3])
+print(save_user_config(config))
+PYCONFIG
+        return
+    fi
+    [ ! -f "$config_path" ] || die "更新已有主配置需要 Python 3，以保留其他配置字段"
     mkdir -p "$config_root"
     chmod 700 "$config_root"
 
@@ -353,10 +368,25 @@ case "$(uname -s)" in
 esac
 printf '\n' >&2
 
-available=$(list_clusters)
-[ -n "$available" ] || die "clusters/ 下没有 profile"
+config_root="${XDG_CONFIG_HOME:-$HOME/.config}/scnet-hpc"
+PYTHON_BIN="$(resolve_python)" || PYTHON_BIN=""
+saved_backend=$(json_string_value "$config_root/config.json" default_backend)
+action_default=2
+if [ -n "$PYTHON_BIN" ]; then
+    action_default=$("$PYTHON_BIN" - "$SCRIPT_DIR" <<'PYDEFAULT'
+import sys
+sys.path.insert(0, sys.argv[1])
+from scnet_config import load_user_config
+from scnet import _has_configuration
+print(3 if _has_configuration(load_user_config()) else 2)
+PYDEFAULT
+    )
+elif [ -f "$config_root/config.json" ] || [ -f "$config_root/openapi.json" ] \
+    || compgen -G "$config_root/ssh/*.json" >/dev/null; then
+    action_default=3
+fi
 if [ -z "$ACTION" ]; then
-    action_label=$(choose "选择配置操作：" \
+    action_label=$(SCNET_SETUP_CHOICE_DEFAULT="$action_default" choose "选择配置操作：" \
         "查看当前状态（status）" \
         "新增连接（new）" \
         "修改现有配置（modify）" \
@@ -381,7 +411,9 @@ if [ "$ACTION" = status ] || [ "$ACTION" = reset ]; then
 fi
 
 if [ -z "$BACKEND" ]; then
-    BACKEND=$(choose "选择要管理的 backend：" \
+    backend_default=1
+    [ "$saved_backend" != ssh ] || backend_default=2
+    BACKEND=$(SCNET_SETUP_CHOICE_DEFAULT="$backend_default" choose "选择要管理的 backend：" \
         "openapi（推荐：平台级 AK/SK 和多区域）" \
         "ssh（区域级用户名和私钥）" \
         "both（同时管理）")
@@ -396,6 +428,17 @@ case "$BACKEND" in
     *) die "backend 必须是 ssh、openapi 或 both" ;;
 esac
 
+# Check requirements before installing keys or changing local configuration.
+if [ "$BACKEND" = openapi ] || [ "$BACKEND" = both ]; then
+    [ -n "$PYTHON_BIN" ] || die "OpenAPI 配置需要 Python 3；请设置 DSH_SCNET_PYTHON 或 SCNET_PYTHON，或安装 python3"
+fi
+if [ "$BACKEND" = ssh ] || [ "$BACKEND" = both ]; then
+    [ -n "$PYTHON_BIN" ] || [ ! -f "$config_root/config.json" ] \
+        || die "更新已有主配置需要 Python 3，以保留其他配置字段"
+    available=$(list_clusters)
+    [ -n "$available" ] || die "clusters/ 下没有 profile"
+fi
+
 if { [ "$BACKEND" = ssh ] || [ "$BACKEND" = both ]; } \
     && [ -z "$CLUSTER" ]; then
     cluster_choices=()
@@ -429,7 +472,9 @@ fi
 
 DEFAULT_BACKEND="$BACKEND"
 if [ "$BACKEND" = both ]; then
-    DEFAULT_BACKEND=$(choose "选择默认 backend：" "openapi" "ssh")
+    backend_default=1
+    [ "$saved_backend" != ssh ] || backend_default=2
+    DEFAULT_BACKEND=$(SCNET_SETUP_CHOICE_DEFAULT="$backend_default" choose "选择默认 backend：" "openapi" "ssh")
 fi
 
 if [ "$BACKEND" = ssh ] || [ "$BACKEND" = both ]; then
@@ -458,8 +503,8 @@ if [ "$BACKEND" = openapi ] || [ "$BACKEND" = both ]; then
         "$PYTHON_BIN" "$SCRIPT_DIR/scnet.py" "${openapi_args[@]}"
     else
         printf '%s\n' \
-            "OpenAPI backend 需要 Python 3；SSH 配置不受影响。" \
-            "请设置 DSH_SCNET_PYTHON 或 SCNET_PYTHON 指向 Python 3，或安装 python3。" >&2
+            "OpenAPI backend 需要 Python 3。" >&2
+        exit 1
     fi
 fi