fix(paper): 取消路径(cancel_paper) + 引用元数据清洗(丢弃杜撰 DOI/PMID)

paper_writer_service/app/services/paper/write.py

@@ -10,10 +10,11 @@ from .cite import find_cite_numbers, render_gbt7714
 _WRITE_SYSTEM = (
     "你是植物学论文撰写助手，正在撰写论文的某一章节。"
     "引用文献一律用 [n] 指代，n 必须取自下方提供的全局编号列表；如需引用新文献，编号按当前最大编号依次 +1。"
-    "禁止杜撰 DOI/PMID/作者/期刊——只能使用编号列表或知识库线索中真实给出的元数据。"
+    "你只能引用「知识库线索」中列出的文献；新引用的元数据只能照抄线索里的 title 和 doc_id。"
+    "知识库线索里没有 DOI/PMID/作者/期刊/年份，这些字段一律输出空字符串，禁止编造任何 DOI/PMID/作者/期刊。"
     "只输出一个合法 JSON 对象：{\"markdown\": 章节正文(不含章标题), \"citations_used\": "
-    "[{\"temp_n\": 编号, \"title\": \"...\", \"authors\": \"...\", \"journal\": \"...\", "
-    "\"year\": \"...\", \"doi\": \"...\", \"pmid\": \"...\", \"source_doc_id\": \"...\", \"snippet\": \"...\"}]}。"
+    "[{\"temp_n\": 编号, \"title\": \"...\", \"authors\": \"\", \"journal\": \"\", "
+    "\"year\": \"\", \"doi\": \"\", \"pmid\": \"\", \"source_doc_id\": \"...\", \"snippet\": \"...\"}]}。"
     "citations_used 必须列出本节正文中出现的每一个 [n] 引用（既有编号也列出其已有元数据）。"
 )
 
@@ -76,6 +77,13 @@ def _build_write_user(paper: dict, outline: dict, sec: dict, cit_text: str, styl
     parts.append("")
     parts.append("全局已编号文献列表（[n] 引用只能从这里取值，或从最大编号 +1 起新增）：")
     parts.append(cit_text)
+    brief = loads_json(paper.get("brief_json"), {})
+    kb_hints = brief.get("kb_hints") or []
+    if kb_hints:
+        parts.append("")
+        parts.append("知识库线索（新引用只能从这里选，元数据只照抄 title/doc_id）：")
+        for h in kb_hints:
+            parts.append(f"- title={h.get('title', '')} | doc_id={h.get('doc_id', '')} | snippet={str(h.get('snippet', ''))[:120]}")
     return "\n".join(parts)
 
 
@@ -91,6 +99,16 @@ def _kind_of(meta: dict[str, Any]) -> str:
     return "title"
 
 
+def _sanitize_citation(meta: dict[str, Any]) -> dict[str, Any]:
+    """丢弃 LLM 可能杜撰的 doi/pmid/authors/journal/year（知识库线索无这些字段），
+    只保留 title/source_doc_id/snippet，强制走 title kind 降级格式。"""
+    return {
+        "title": (meta.get("title") or "").strip(),
+        "source_doc_id": meta.get("source_doc_id") or "",
+        "snippet": meta.get("snippet") or "",
+    }
+
+
 async def _persist_new_citations(paper_id: str, section_id: str, used: list[Any], existing_nums: set[int]) -> list[dict[str, Any]]:
     """把 citations_used 中的新引用（temp_n 不在既有集合）落库，返回新增摘要。"""
     new_citations: list[dict[str, Any]] = []
@@ -102,13 +120,14 @@ async def _persist_new_citations(paper_id: str, section_id: str, used: list[Any]
         if not tn or tn in existing_nums or tn in seen_new:
             continue
         seen_new.add(tn)
-        ref_text = render_gbt7714(c)
+        meta = _sanitize_citation(c)
+        ref_text = render_gbt7714(meta)
         await db.execute(
             "INSERT INTO citations (paper_id, section_id, temp_n, kind, ref_text, source_doc_id, snippet, created_at) "
             "VALUES (?, ?, ?, ?, ?, ?, ?, ?)",
             (
-                paper_id, section_id, tn, _kind_of(c), ref_text,
-                c.get("source_doc_id") or "", c.get("snippet") or "", now_iso(),
+                paper_id, section_id, tn, _kind_of(meta), ref_text,
+                meta.get("source_doc_id") or "", meta.get("snippet") or "", now_iso(),
             ),
         )
         new_citations.append({"temp_n": tn, "ref_text": ref_text})

sdk/app/services/paper/agent.py

@@ -139,6 +139,14 @@ async def _outline_message(db, session_id: str, outline: dict[str, Any]) -> list
     return blocks
 
 
+_CANCEL_WORDS = ("暂停", "取消", "停止", "不写了", "终止", "中止", "算了")
+
+
+def _is_cancel_request(text: str) -> bool:
+    s = (text or "").strip()
+    return bool(s) and any(w in s for w in _CANCEL_WORDS)
+
+
 async def _followup(db, session_id: str, text: str, paper_id: str, client: PaperMcpClient) -> list[dict[str, Any]]:
     prog = await client.call_tool("get_paper_progress", {"paper_id": paper_id})
     if not prog.get("ok"):
@@ -147,6 +155,14 @@ async def _followup(db, session_id: str, text: str, paper_id: str, client: Paper
         return blocks
 
     status = prog.get("status")
+    if _is_cancel_request(text):
+        res = await client.call_tool("cancel_paper", {"paper_id": paper_id})
+        if res.get("ok"):
+            blocks = [_text_block("论文已取消。如需重新撰写，请告诉我新主题。")]
+        else:
+            blocks = [_text_block(f"取消失败：{res.get('message', '')}")]
+        await _write_message(db, session_id, blocks=blocks, text=blocks[0]["html"])
+        return blocks
     if status == "complete":
         return await _revise(db, session_id, paper_id, text, client)
     if status == "cancelled":
@@ -216,6 +232,9 @@ async def run_writing_loop(db, paper_id: str, session_id: str) -> None:
             if not prog.get("ok"):
                 await _write_message(db, session_id, blocks=[_text_block("撰写进度查询失败。")], text="撰写失败")
                 return
+            if prog.get("status") == "cancelled":
+                await _write_message(db, session_id, blocks=[_text_block("论文已取消。")], text="已取消")
+                return
             sections = prog.get("sections") or []
             pending = [s for s in sections if s.get("status") != "done"]
             if not pending:

tests/paper/test_paper_tools.py

@@ -82,6 +82,22 @@ def test_write_section_creates_citations(paper_db, fake_llm, fake_search):
     assert "[1]" in sec["markdown"]
 
 
+def test_write_section_drops_fabricated_doi(paper_db, fake_llm, fake_search):
+    """杜撰的 doi/pmid/authors/journal 应被丢弃，强制 title kind 降级格式。"""
+    paper_id = _propose(fake_llm, fake_search)["paper_id"]
+    fake_llm.responses.append({
+        "markdown": "正文[1]。",
+        "citations_used": [{"temp_n": 1, "title": "水稻SPL研究", "authors": "X", "journal": "J",
+                            "year": "2020", "doi": "10.1234/fake", "pmid": "123", "source_doc_id": "d1"}],
+    })
+    r = _run(write.write_paper_section(paper_id, "s1"))
+    assert r["ok"] is True
+    cit = _run(dbmod.db.fetchone("SELECT * FROM citations WHERE temp_n = 1"))
+    assert cit["kind"] == "title"
+    assert "知识库资料" in cit["ref_text"]
+    assert "10.1234" not in cit["ref_text"]
+
+
 def test_write_section_invalid_citation_retries_once(paper_db, fake_llm, fake_search):
     paper_id = _propose(fake_llm, fake_search)["paper_id"]
     fake_llm.calls.clear()  # 重置计数，只统计写节调用（大纲已消费 1 次）