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>

server/src/bid_doctor/ai/providers/base.py

@@ -12,6 +12,8 @@ class Completion:
     tokens_in: int | None = None
     tokens_out: int | None = None
     output_mode: str = "json_mode"
+    # 输出不完整：被截断（finish_reason=length）或不是合法 JSON。调用方要拆小输入重跑，不能当成「没有结果」
+    incomplete: bool = False
 
 
 class TransientError(Exception):

server/src/bid_doctor/ai/providers/fake.py

@@ -135,15 +135,25 @@ def _tech_param(lines: list[tuple[str, str, str, str]]) -> dict[str, Any]:
     return {"items": items}
 
 
+INFO_KEYS = (
+    "name", "code", "purchaser", "agency", "procurement_method", "legal_regime", "budget_total", "ceiling_total",
+    "eval_method", "accept_consortium", "accept_imported", "sme_only",
+)  # fmt: skip
+
+
 def _project_info(lines: list[tuple[str, str, str, str]]) -> dict[str, Any]:
+    """v2 的格式：每个字段都表态 found / not_found。"""
     fields: dict[str, Any] = {}
     for sid, _, p, _ in lines:
         for key, label in (("name", "项目名称"), ("code", "项目编号"), ("purchaser", "采购人")):
             if key not in fields and (m := re.search(label + r"[：:]\s*([^，。；]+)", p)):
-                fields[key] = {"value": m.group(1).strip(), "sid": sid, "quote": p[:30]}
+                fields[key] = {"status": "found", "value": m.group(1).strip(), "sid": sid, "quote": p[:30]}
         if "budget_total" not in fields and (m := re.search(r"预算[^0-9]{0,8}(\d+(?:\.\d+)?)\s*(万元|元)", p)):
-            fields["budget_total"] = {"value": m.group(1), "unit": m.group(2), "sid": sid, "quote": p[:30]}
-    return {"fields": fields, "lots": []}
+            fields["budget_total"] = {
+                "status": "found", "value": m.group(1), "unit": m.group(2), "sid": sid, "quote": p[:30],
+            }  # fmt: skip
+    none = {"status": "not_found", "value": None, "unit": None, "sid": None, "quote": None}
+    return {"fields": {k: fields.get(k, none) for k in INFO_KEYS}, "lots": []}
 
 
 TASKS: dict[str, Callable[[list[tuple[str, str, str, str]]], dict[str, Any]]] = {

server/src/bid_doctor/ai/providers/openai_compat.py

@@ -41,16 +41,20 @@ class OpenAICompatProvider:
             raise TransientError("server", f"模型接口 {r.status_code}")
         r.raise_for_status()
         payload = r.json()
-        raw = payload["choices"][0]["message"]["content"]
+        choice = payload["choices"][0]
+        raw = choice["message"]["content"] or ""
         usage = payload.get("usage") or {}
+        incomplete = choice.get("finish_reason") == "length"
         try:
             data = json.loads(raw)
         except json.JSONDecodeError:
-            data = {"items": []}  # 内容质量问题：不算任务失败，相关条目由规则兜底进待确认区（架构 §3.6）
+            # 以前当成「没有结果」：输出被截断时，这一整块的条目会悄悄丢掉（2026-11-03 审计发现，输出正好 8192 token）
+            data, incomplete = {}, True
         return Completion(
             data=data,
             raw=raw,
             model=payload.get("model", model),
             tokens_in=usage.get("prompt_tokens"),
             tokens_out=usage.get("completion_tokens"),
+            incomplete=incomplete,
         )