feat: release 1.5.1 with human review and batch filters

apps/admin_console/routers/automation.py

@@ -13,7 +13,7 @@ from apps.admin_console.services import automation_service as service
 from apps.admin_console.services import batch_archive as archive_service
 from apps.admin_console.services import report_export, task_import
 from artemis.config.settings import settings
-from artemis.runtime.case_result import read_case_result
+from apps.admin_console.services.case_review import read_reviewed_result
 
 router = APIRouter(prefix="/api/automation", tags=["automation"])
 
@@ -24,6 +24,12 @@ class ReportPreferences(BaseModel):
     )
 
 
+class CaseReviewRequest(BaseModel):
+    model_config = {"extra": "forbid"}
+    status: Literal["passed", "failed"]
+    reason: str = Field(default="", max_length=2000)
+
+
 class BatchRequest(ReportPreferences):
     request_id: str | None = Field(default=None, pattern=r"^[A-Za-z0-9_-]{1,100}$")
     name: str = Field(min_length=1, max_length=200)
@@ -125,8 +131,8 @@ def batch_by_request(request_id: str):
 def batch_detail(batch_id: str):
     jobs = store.jobs(batch_id)
     for job in jobs:
-        job["test_result"] = read_case_result(
-            job["session_id"], job["status"], job.get("error") or ""
+        job["test_result"] = read_reviewed_result(
+            job["session_id"], job["status"], job.get("error") or "", storage=store
         )
     return {**require("batch", batch_id), "jobs": jobs}
 
@@ -207,6 +213,42 @@ async def download_report(
         raise HTTPException(404, str(exc)) from exc
 
 
+@router.put("/reports/{session_id}/review")
+def review_case(session_id: str, value: CaseReviewRequest):
+    report(session_id)
+    job = store.job(session_id) or {}
+    batch_id = job.get("batch_id")
+    if (
+        batch_id
+        and archive_service.status(batch_id, require("batch", batch_id))["status"] == "preparing"
+    ):
+        raise HTTPException(409, "批次报告正在打包，请完成后再保存人工复验")
+    with report_export.EXPORT_LOCK:
+        current = report(session_id)
+        if current["status"] not in {
+            "completed",
+            "success",
+            "failed",
+            "cancelled",
+            "interrupted",
+            "skipped",
+        }:
+            raise HTTPException(409, "请在任务结束后进行人工复验")
+        try:
+            store.save_case_review(session_id, value.status, value.reason)
+        except ValueError as exc:
+            raise HTTPException(409, str(exc)) from exc
+        if batch_id:
+            archive_service.forget(batch_id)
+            # A finishing worker can publish report_ready just after export;
+            # removing only the generated ZIP also prevents that stale cache hit.
+            try:
+                archive_service.archive_path(batch_id).unlink(missing_ok=True)
+            except OSError:
+                pass  # An open download may hold the file; report_ready is already invalidated.
+        return report(session_id)
+
+
 @router.get("/history")
 def history(
     q: str = Query(default="", max_length=200),
@@ -215,6 +257,7 @@ def history(
     end: float | None = None,
     page: int = Query(default=1, ge=1),
     size: int = Query(default=30, ge=1, le=100),
+    batch_id: str = Query(default="", max_length=100),
 ):
     # SQL filtering/pagination searches the entire persisted history, not a UI cache.
     query = """WITH history AS (
@@ -226,7 +269,7 @@ def history(
         COALESCE(json_extract(j.payload,'$.case.category'),'') category,
         COALESCE(json_extract(j.payload,'$.case.subcategory'),'') subcategory,
         COALESCE(json_extract(j.payload,'$.case.preconditions'),'') preconditions,
-        COALESCE(json_extract(j.payload,'$.error'),'') error
+        COALESCE(json_extract(j.payload,'$.error'),'') error, j.batch_id
       FROM sessions s LEFT JOIN automation_jobs j ON s.session_id=j.session_id
       UNION ALL
       SELECT j.session_id,json_extract(j.payload,'$.case.goal'),
@@ -237,11 +280,22 @@ def history(
         COALESCE(json_extract(j.payload,'$.case.category'),''),
         COALESCE(json_extract(j.payload,'$.case.subcategory'),''),
         COALESCE(json_extract(j.payload,'$.case.preconditions'),''),
-        COALESCE(json_extract(j.payload,'$.error'),'')
+        COALESCE(json_extract(j.payload,'$.error'),''), j.batch_id
       FROM automation_jobs j WHERE NOT EXISTS (SELECT 1 FROM sessions s WHERE s.session_id=j.session_id)
     ) """
     where = " WHERE status IN ('completed','success','failed','cancelled','interrupted','skipped')"
     args = []
+    if batch_id == "__batch__":
+        where += " AND batch_id IS NOT NULL"
+    elif batch_id == "__single__":
+        where += " AND batch_id IS NULL"
+    elif batch_id:
+        try:
+            identifier(batch_id)
+        except ValueError as exc:
+            raise HTTPException(400, str(exc)) from exc
+        where += " AND batch_id = ?"
+        args.append(batch_id)
     if q:
         where += " AND (name LIKE ? ESCAPE '\\' OR initial_goal LIKE ? ESCAPE '\\' OR session_id LIKE ? ESCAPE '\\' OR tags LIKE ? ESCAPE '\\' OR case_id LIKE ? ESCAPE '\\' OR category LIKE ? ESCAPE '\\' OR subcategory LIKE ? ESCAPE '\\' OR preconditions LIKE ? ESCAPE '\\')"
         escaped = "%" + q.replace("\\", "\\\\").replace("%", "\\%").replace("_", "\\_") + "%"
@@ -266,8 +320,8 @@ def history(
     items = [dict(row) for row in rows]
     jobs = store.jobs_for_sessions([item["session_id"] for item in items])
     for item in items:
-        item["test_result"] = read_case_result(
-            item["session_id"], item["status"], item.get("error") or ""
+        item["test_result"] = read_reviewed_result(
+            item["session_id"], item["status"], item.get("error") or "", storage=store
         )
         if job := jobs.get(item["session_id"]):
             item["name"] = task_import.execution_title(job)

apps/admin_console/routers/sessions.py

@@ -20,7 +20,7 @@ from fastapi import APIRouter, HTTPException
 
 from artemis.config import DB_PATH, TRACES_PATH
 from artemis.runtime import trace_store
-from artemis.runtime.case_result import read_case_result
+from apps.admin_console.services.case_review import read_reviewed_result
 
 try:
     from admin_console.core.state import state
@@ -83,6 +83,7 @@ def _list_sessions_sync():
         job = jobs_by_id.get(s_id)
         if job:
             case = job.get("case") or {}
+            row_dict["batch_id"] = job.get("batch_id")
             row_dict["case_id"] = case.get("case_id") or ""
             row_dict["display_name"] = execution_title(job) or row_dict.get("initial_goal")
             row_dict["iteration"] = job.get("iteration", 1)
@@ -131,8 +132,12 @@ def _list_sessions_sync():
         else:
             row_dict["video_url"] = None
         row_dict["recording_status"] = recording_status
-        row_dict["test_result"] = read_case_result(
-            s_id, row_dict.get("status") or "", row_dict.get("error") or "", root=TRACES_PATH
+        row_dict["test_result"] = read_reviewed_result(
+            s_id,
+            row_dict.get("status") or "",
+            row_dict.get("error") or "",
+            root=TRACES_PATH,
+            storage=store,
         )
 
         agent_names = agent_names_by_session.get(s_id, [])
@@ -194,8 +199,10 @@ async def get_session_details(session_id: str):
     if not row:
         raise HTTPException(status_code=404, detail=f"Session {session_id} not found")
     result = dict(row)
-    result["test_result"] = read_case_result(
-        session_id, result.get("status") or "", root=TRACES_PATH
+    job = store.job(session_id) or {}
+    result["batch_id"] = job.get("batch_id")
+    result["test_result"] = read_reviewed_result(
+        session_id, result.get("status") or "", root=TRACES_PATH, storage=store
     )
     return result
 

apps/admin_console/routers/tasks.py

@@ -20,6 +20,8 @@ from fastapi import APIRouter, HTTPException, Request
 from fastapi.responses import StreamingResponse
 from pydantic import BaseModel, Field
 from artemis.core.diagnostics import readiness_engine
+from apps.admin_console.services.automation_store import store
+from apps.admin_console.services.task_import import execution_title
 from artemis.runtime import DeviceExecutionLock, device_pool
 from artemis.runtime.task_control import (
     add_instruction,
@@ -423,6 +425,15 @@ async def get_status():
             seen_ids.add(sid)
             queue_data.append(g_item)
 
+    batch_jobs = store.jobs_for_sessions(
+        [str(item["session_id"]) for item in [*queue_data, *active_tasks] if item.get("session_id")]
+    )
+    queue_data = [{**item} for item in queue_data]
+    for item in [*queue_data, *active_tasks]:
+        if job := batch_jobs.get(str(item.get("session_id"))):
+            item["batch_id"] = job.get("batch_id")
+            item["display_name"] = execution_title(job)
+
     if is_running:
         is_paused = state.is_paused
         return {

apps/admin_console/services/automation_store.py

@@ -149,6 +149,52 @@ class AutomationStore:
     def preferences(self) -> dict:
         return self.get("settings", "reports") or {"formats": ["docx", "xlsx"]}
 
+    def save_case_review(self, session_id: str, status: str, reason: str) -> dict:
+        """Append reviews atomically; invalidate the batch's generated report."""
+        identifier(session_id)
+        if status not in {"passed", "failed"}:
+            raise ValueError("人工复验结果只能为通过或不通过")
+        review = {
+            "status": status,
+            "reason": reason.strip()
+            or ("人工复验确认通过。" if status == "passed" else "人工复验确认不通过。"),
+            "reviewed_at": time.time(),
+        }
+        with self.connection() as conn:
+            conn.execute("BEGIN IMMEDIATE")
+            row = conn.execute(
+                "SELECT payload FROM automation_records WHERE kind='case_review' AND id=?",
+                (session_id,),
+            ).fetchone()
+            record = json.loads(row[0]) if row else {"reviews": []}
+            record["reviews"].append(review)
+            conn.execute(
+                "INSERT INTO automation_records VALUES (?,?,?,?) ON CONFLICT(kind,id) DO UPDATE SET payload=excluded.payload,updated_at=excluded.updated_at",
+                (
+                    "case_review",
+                    session_id,
+                    json.dumps(record, ensure_ascii=False),
+                    review["reviewed_at"],
+                ),
+            )
+            job = conn.execute(
+                "SELECT batch_id FROM automation_jobs WHERE session_id=?", (session_id,)
+            ).fetchone()
+            if job and job[0]:
+                batch = conn.execute(
+                    "SELECT payload FROM automation_records WHERE kind='batch' AND id=?", (job[0],)
+                ).fetchone()
+                if not batch or json.loads(batch[0]).get("status") == "deleting":
+                    raise ValueError("批次正在删除，不能保存复验")
+                payload = json.loads(batch[0])
+                payload["report_ready"] = False
+                conn.execute(
+                    "UPDATE automation_records SET payload=?,updated_at=? WHERE kind='batch' AND id=?",
+                    (json.dumps(payload, ensure_ascii=False), review["reviewed_at"], job[0]),
+                )
+            conn.commit()
+        return review
+
     def create_batch(self, batch: dict, jobs: list[dict], payload: dict) -> dict:
         """Commit a batch and all its jobs atomically; repeated request IDs never duplicate work."""
         request_id = payload.get("request_id")
@@ -247,6 +293,11 @@ class AutomationStore:
             batch = json.loads(row[0])
             if batch.get("status") != "deleting":
                 raise ValueError("批次尚未准备删除")
+            conn.execute(
+                "DELETE FROM automation_records WHERE kind='case_review' AND id IN "
+                "(SELECT session_id FROM automation_jobs WHERE batch_id=?)",
+                (batch_id,),
+            )
             conn.execute("DELETE FROM automation_jobs WHERE batch_id=?", (batch_id,))
             conn.execute(
                 "DELETE FROM automation_records WHERE kind='batch' AND id=?",
@@ -260,9 +311,14 @@ class AutomationStore:
         with self.connection() as conn:
             if session_id:
                 conn.execute("DELETE FROM automation_jobs WHERE session_id=?", (session_id,))
+                conn.execute(
+                    "DELETE FROM automation_records WHERE kind='case_review' AND id=?",
+                    (session_id,),
+                )
             else:
                 conn.execute("DELETE FROM automation_jobs")
                 conn.execute("DELETE FROM automation_records WHERE kind='batch'")
+                conn.execute("DELETE FROM automation_records WHERE kind='case_review'")
             conn.commit()
 
 

apps/admin_console/services/case_review.py

@@ -0,0 +1,40 @@
+"""Human review overlays the AI verdict without changing execution evidence."""
+
+from pathlib import Path
+
+from apps.admin_console.services.automation_store import AutomationStore
+from artemis.runtime.case_result import read_case_result
+
+
+def reviewed_result(
+    session_id: str, ai_result: dict | None, storage: AutomationStore
+) -> dict | None:
+    original = {**ai_result, "source": "ai"} if ai_result else None
+    record = storage.get("case_review", session_id)
+    reviews = (record or {}).get("reviews") or []
+    if not reviews:
+        return original
+    latest = reviews[-1]
+    return {
+        "status": latest["status"],
+        "reason": latest["reason"],
+        "evidence": [],
+        "source": "human",
+        "reviewed_at": latest["reviewed_at"],
+        "ai_result": original,
+    }
+
+
+def read_reviewed_result(
+    session_id: str,
+    execution_status: str,
+    error: str = "",
+    *,
+    storage: AutomationStore,
+    root: Path | None = None,
+) -> dict | None:
+    return reviewed_result(
+        session_id,
+        read_case_result(session_id, execution_status, error, root=root),
+        storage,
+    )

apps/admin_console/services/report_export.py

@@ -24,6 +24,7 @@ from apps.admin_console.database.repositories.step_repository import StepReposit
 from apps.admin_console.services.automation_store import store, identifier
 from apps.admin_console.services.task_import import execution_title
 from artemis.runtime.case_result import RESULT_LABELS, resolve_case_result
+from apps.admin_console.services.case_review import reviewed_result
 
 EXPORT_ROOT = WORKSPACE_ROOT / "artifacts" / "reports"
 EXPORT_LOCK = threading.RLock()
@@ -65,7 +66,10 @@ def text(value) -> str:
 def verdict(outcome: dict | None) -> str:
     result = (outcome or {}).get("test_result")
     if isinstance(result, dict) and result.get("status") in RESULT_LABELS:
-        return RESULT_LABELS[result["status"]]
+        label = RESULT_LABELS[result["status"]]
+        if result.get("source"):
+            label += "（人工确认）" if result["source"] == "human" else "（AI确认）"
+        return label
     tests = (outcome or {}).get("tests") or {}
     if tests.get("failed", 0):
         return "验证失败"
@@ -183,9 +187,10 @@ def session_report(session_id: str, db_path=None) -> dict:
     case = meta.get("case") or {}
     status = meta.get("terminal_status") or meta.get("status") or session.get("status", "unknown")
     error = failure_reason(directory, status, meta.get("error"))
-    test_result = resolve_case_result(outcome, status, error)
+    ai_result = resolve_case_result(outcome, status, error)
+    test_result = reviewed_result(session_id, ai_result.model_dump() if ai_result else None, store)
     if test_result:
-        outcome = {**outcome, "test_result": test_result.model_dump()}
+        outcome = {**outcome, "test_result": test_result}
     verification = verdict(outcome)
     if status in {"failed", "interrupted"} and verification.startswith("未验证"):
         verification = "未验证（任务异常终止）"
@@ -204,7 +209,7 @@ def session_report(session_id: str, db_path=None) -> dict:
         "expected": case.get("expected", ""),
         "status": status,
         "verdict": verification,
-        "test_result": test_result.model_dump() if test_result else None,
+        "test_result": test_result,
         "outcome": outcome,
         "content": content,
         "steps": steps,
@@ -241,6 +246,13 @@ def metadata(report: dict) -> list[tuple]:
         ("预期结果", report["expected"] or "未指定"),
         ("执行状态", STATUS.get(report["status"], report["status"])),
         ("验证结论", report["verdict"]),
+        ("人工复验时间", time_text((report.get("test_result") or {}).get("reviewed_at"))),
+        (
+            "原 AI 结论",
+            verdict({"test_result": (report.get("test_result") or {}).get("ai_result")})
+            if (report.get("test_result") or {}).get("source") == "human"
+            else "—",
+        ),
         ("判定依据", (report.get("test_result") or {}).get("reason") or "未记录"),
         (
             "验收证据",