feat(server): 参考 BidEngine 补五项：原因码、截断重跑、项目信息逐项表态、占位符、提示词规则对应校验

- 待确认记原因码（extract.FAILURES，存在 ai_values.failures），解析结果按原因计数
  （verify_failures）；前端待确认条目显示具体原因
- 相邻分块重叠约 800 字；模型输出被截断或不是合法 JSON 时对半重跑（最多 3 层），
  拆到底仍不完整记 output_incomplete。修复：以前截断时整块条目被当成「没有结果」悄悄丢掉
- 项目基本信息提示词 v2：12 个字段逐项表态 found / ambiguous / not_found，跨分块合并为
  found / absent / needs_review（info_fields）
- 「____元」「××」「（ ）」「待填」等占位符不算取值
- 提示词 substantive v4，qualification / scoring / schedule / tech_param v3：每条禁止事项
  对应一个原因码，对照表在 prompts/README.md
- 顾问（Fable）深读参考项目的结论补进 ref-bid-engine-analysis.md §5；架构 §13 SK12
- 待确认：国动 4 → 2，S02 13 → 3，S06 57 → 55

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

apps/web/src/components/ChecklistItemRow.vue

@@ -7,7 +7,7 @@
 import { computed } from "vue";
 import CitationBlock from "./CitationBlock.vue";
 import { groupCitations, type CitationGroup } from "@/lib/citations";
-import { category, IMPORTANCE_MARK, IMPORTANCE_NAME, pendingReason, specFacts } from "@/lib/labels";
+import { category, failureText, IMPORTANCE_MARK, IMPORTANCE_NAME, pendingReason, specFacts } from "@/lib/labels";
 import type { ChecklistItem } from "@/lib/queries";
 
 const props = defineProps<{ item: ChecklistItem; pending?: boolean; selected?: boolean; activeCitation?: string | null }>();
@@ -36,7 +36,7 @@ const facts = computed(() => specFacts(props.item.spec));
         <span v-else class="tag">{{ item.origin_label || (item.origin === "human" ? "人工添加" : "规则") }}</span>
         <span v-for="c in item.categories" :key="c" class="tag">{{ category(c) }}</span>
         <span class="tag">{{ item.lot_no ? `采购包 ${item.lot_no}` : "通用" }}</span>
-        <span v-if="pending && item.pending_reason" class="flag">{{ pendingReason(item.pending_reason) }}</span>
+        <span v-if="pending && item.pending_reason" class="flag">{{ [pendingReason(item.pending_reason), failureText(item.ai_values)].filter(Boolean).join(" · ") }}</span>
         <span v-if="item.no_citation_ack" class="flag">人工确认、无出处</span>
         <span v-if="item.edited_fields.length" class="flag">已修改</span>
       </div>

apps/web/src/lib/labels.ts

@@ -53,6 +53,32 @@ export const PENDING_REASON: Dict<Schemas["PendingReason"]> = {
 };
 export const pendingReason = (s: string | null | undefined) => pick(PENDING_REASON, s);
 
+/** 没通过校验的具体原因（服务端 checklist/extract.py 的 FAILURES，记在 ai_values.failures） */
+export const FAILURE: Record<string, string> = {
+  evidence_missing: "没有给出处",
+  sid_format: "句子编号格式不对",
+  sid_unknown: "句子编号对不上",
+  quote_too_short: "引文太短",
+  quote_not_original: "引文和原文不一致",
+  basis_unmatched: "原文里找不到实质性条款的依据",
+  value_unverified: "取值在原文里找不到",
+  points_unverified: "分值在原文里找不到",
+  date_unverified: "日期时间在原文里找不到",
+  amount_unverified: "金额在原文里找不到",
+  placeholder_value: "原文这里是待填写的空白",
+  marker_uncovered: "",
+};
+/** 待确认条目的具体原因，最多列两条；和 pending_reason 重复的不再写 */
+export function failureText(aiValues: unknown): string {
+  const fs = (aiValues as { failures?: unknown } | null)?.failures;
+  if (!Array.isArray(fs)) return "";
+  return fs
+    .map((f) => FAILURE[String(f)] ?? "")
+    .filter(Boolean)
+    .slice(0, 2)
+    .join("、");
+}
+
 /* ---------- 修改记录（审计事件） ---------- */
 export const HISTORY_ACTION: Record<string, string> = {
   "checklist_item.extract": "AI 提取",