feat: 更新后端平台模块、前端组件及构建产物，新增工作计划文档

- 更新 backend 平台 API endpoints 及 platform_store
- 更新前端 ComputeNodesView、DataProcessCreateView、FineTuneCreateView 等组件
- 更新前端 API 模块（compute、fineTune）
- 重构 frontend/dist 构建产物（新 hash）
- 新增 docs/2026-07-24-work-plan.md 工作计划文档

Co-Authored-By: Claude <noreply@anthropic.com>

backend/app/db/platform_store.py

@@ -1474,25 +1474,40 @@ class PlatformStore:
     def prepare_compute_job_payload(self, task_id: str, payload: dict[str, Any] | None = None) -> tuple[dict[str, Any], dict[str, Any]]:
         task = self.task(task_id)
         merged = {**task, **(payload or {}), "id": task_id}
-        node = self.schedule_node(merged)
+        node = self.select_compute_node(merged)
         selected_gpus = merged.get("gpus") or [0]
         return node, self._compute_job_payload_from_task_node(merged, node, selected_gpus)
 
+    def prepare_compute_job_payload_from_payload(self, payload: dict[str, Any]) -> tuple[dict[str, Any], dict[str, Any]]:
+        task_id = str(payload.get("task_id") or payload.get("id") or new_id("ft_preview"))
+        name = str(payload.get("name") or task_id)
+        transient_task = {
+            **payload,
+            "id": task_id,
+            "name": name,
+            "status": str(payload.get("status") or "pending"),
+            "progress": int(payload.get("progress") or 0),
+        }
+        node = self.select_compute_node(transient_task)
+        selected_gpus = transient_task.get("gpus") or [0]
+        return node, self._compute_job_payload_from_task_node(transient_task, node, selected_gpus)
+
     def _compute_job_payload_from_task_node(
         self,
         task: dict[str, Any],
         node: dict[str, Any],
         selected_gpus: list[int] | list[Any] | None = None,
     ) -> dict[str, Any]:
+        base_model_id = task.get("base_model") or task.get("model_id")
+        dataset_id = str(task.get("train_dataset_id") or task.get("dataset_id") or "")
         with self.connect() as conn:
-            model = conn.execute("SELECT * FROM models WHERE id=?", (task.get("base_model"),)).fetchone()
-            dataset = conn.execute("SELECT * FROM datasets WHERE id=?", (task.get("train_dataset_id"),)).fetchone()
+            model = conn.execute("SELECT * FROM models WHERE id=?", (base_model_id,)).fetchone()
+            dataset = conn.execute("SELECT * FROM datasets WHERE id=?", (dataset_id,)).fetchone()
             files = conn.execute(
                 "SELECT id, name, size, active_version_id, create_time FROM dataset_files WHERE dataset_id=? ORDER BY create_time",
-                (task.get("train_dataset_id"),),
+                (dataset_id,),
             ).fetchall()
-        model_path = (model and model.get("path")) or task.get("base_model")
-        dataset_id = str(task.get("train_dataset_id") or "")
+        model_path = (model and model.get("path")) or task.get("model_name_or_path") or base_model_id
         if not files:
             raise RuntimeError(f"dataset has no uploaded file: {dataset_id}")
         dataset_key = str(task.get("dataset_key") or llama_dataset_key(dataset_id))
@@ -1897,6 +1912,10 @@ class PlatformStore:
                 raise RuntimeError("compute scheduler is busy, please retry")
             return self._schedule_node_locked(conn, payload)
 
+    def select_compute_node(self, payload: dict[str, Any]) -> dict[str, Any]:
+        with self.connect() as conn:
+            return self._schedule_node_locked(conn, payload)
+
     def _reserve_gpu_allocations(
         self,
         conn: PgConnection,