更新

app/src/doctor_workstation/ui/report_provenance.py

@@ -5,6 +5,7 @@ from __future__ import annotations
 import re
 from collections.abc import Mapping
 from typing import Any
+from urllib.parse import unquote
 
 _REPORT_SOURCE_COUNTS = (
     ("diagnosis_count", "诊单"),
@@ -25,6 +26,12 @@ _ANCIENT_BOOKS_COMMIT = "db0155dc7c42b9c6b3736896661f317c7110038f"
 _ANCIENT_BOOKS_URL = "https://github.com/xiaopangxia/TCM-Ancient-Books"
 
 
+def _source_text(value: Any, limit: int) -> str:
+    """Keep displayed source metadata brief and free of repository links."""
+    text = re.sub(r"https://github\.com(?:/[^\s，。；、」）》\]]*)?", "", str(value or ""))
+    return " ".join(text.split())[:limit]
+
+
 def report_provenance_lines(raw: Any) -> tuple[str, ...]:
     """Describe recorded inputs; never turn model prose into a claimed citation."""
     report = raw if isinstance(raw, Mapping) else {}
@@ -53,16 +60,23 @@ def report_provenance_lines(raw: Any) -> tuple[str, ...]:
 
     knowledge = report.get("knowledge_source")
     url = knowledge.get("url") if isinstance(knowledge, Mapping) else None
-    source_match = re.fullmatch(
+    legacy_source_match = re.fullmatch(
         re.escape(_NIHAIXIA_SOURCE_URL) + r"/blob/([0-9a-f]{40})/SKILL\.md",
         url,
     ) if isinstance(url, str) else None
-    if isinstance(knowledge, Mapping) and source_match is not None and source_match.group(1) == _NIHAIXIA_COMMIT:
-        commit = knowledge.get("commit")
-        commit_text = commit[:12] if isinstance(commit, str) and commit == _NIHAIXIA_COMMIT else ""
+    valid_knowledge = isinstance(knowledge, Mapping) and knowledge.get("commit") == _NIHAIXIA_COMMIT and (
+        (url in (None, "") and knowledge.get("path") == "SKILL.md")
+        or (
+            legacy_source_match is not None
+            and legacy_source_match.group(1) == _NIHAIXIA_COMMIT
+            and knowledge.get("path") in (None, "SKILL.md")
+        )
+    )
+    if valid_knowledge:
         lines.append(
-            "知识参考：倪海厦 skill · SKILL.md"
-            + (f"（commit {commit_text}）" if commit_text else "")
+            "知识参考：倪海厦临床参考 Skill；本地资料："
+            "server/knowledge/nihaixia-skill/SKILL.md"
+            + f"（版本 {_NIHAIXIA_COMMIT[:12]}）"
         )
         sections = knowledge.get("sections")
         allowed_sections = {"六经速查表", "六经辨证诊断公式", "七步走辨证思维模式"}
@@ -70,36 +84,55 @@ def report_provenance_lines(raw: Any) -> tuple[str, ...]:
             isinstance(item, str) and item in allowed_sections for item in sections
         ):
             lines.append("参考章节：" + "、".join(sections))
-        lines.append("原文：" + url)
     else:
         lines.append("知识参考：此报告未记录所用知识资料版本")
 
     ancient_sources = report.get("ancient_book_sources")
-    shown_urls: set[str] = set()
+    shown_sources: set[tuple[str, int]] = set()
     if isinstance(ancient_sources, (list, tuple)):
         for source in ancient_sources[:12]:
-            if not isinstance(source, Mapping):
+            if not isinstance(source, Mapping) or source.get("commit") != _ANCIENT_BOOKS_COMMIT:
                 continue
             ancient_url = source.get("url")
-            if not isinstance(ancient_url, str) or ancient_url in shown_urls:
+            filename = source.get("path")
+            line_number = source.get("line")
+            if ancient_url not in (None, ""):
+                legacy_match = re.fullmatch(
+                    re.escape(_ANCIENT_BOOKS_URL)
+                    + r"/blob/"
+                    + _ANCIENT_BOOKS_COMMIT
+                    + r"/([^/#?]+\.txt)#L([1-9]\d{0,8})",
+                    ancient_url,
+                ) if isinstance(ancient_url, str) else None
+                if legacy_match is None:
+                    continue
+                legacy_filename = unquote(legacy_match.group(1))
+                legacy_line = int(legacy_match.group(2))
+                if filename not in (None, legacy_filename) or line_number not in (None, legacy_line):
+                    continue
+                filename, line_number = legacy_filename, legacy_line
+            if not (
+                isinstance(filename, str)
+                and re.fullmatch(r"[^/\\\x00-\x1f]{1,255}\.txt", filename)
+                and type(line_number) is int
+                and 0 < line_number < 1_000_000_000
+            ):
                 continue
-            if re.fullmatch(
-                re.escape(_ANCIENT_BOOKS_URL)
-                + r"/blob/"
-                + _ANCIENT_BOOKS_COMMIT
-                + r"/[^/#?]+\.txt#L[1-9]\d*",
-                ancient_url,
-            ) is None or source.get("commit") != _ANCIENT_BOOKS_COMMIT:
+            source_key = (filename, line_number)
+            if source_key in shown_sources:
                 continue
-            shown_urls.add(ancient_url)
-            title = str(source.get("title") or "古籍原文")[:60]
-            chapter = str(source.get("chapter") or "")[:60]
+            shown_sources.add(source_key)
+            title = _source_text(source.get("title"), 60) or "古籍原文"
+            chapter = _source_text(source.get("chapter"), 60)
+            excerpt = _source_text(source.get("excerpt"), 120)
             lines.append(
                 "古籍参考：《" + title + "》"
                 + ("「" + chapter + "」" if chapter else "")
-                + "；原文：" + ancient_url
+                + f"；本地资料：server/knowledge/tcm-ancient-books/{filename}"
+                + f" 第{line_number}行（版本 {_ANCIENT_BOOKS_COMMIT[:12]}）"
+                + ("；摘录：" + excerpt if excerpt else "")
             )
-    if not shown_urls and "-ancient-" in str(
+    if not shown_sources and "-ancient-" in str(
         report.get("prompt_version") or report.get("context_version") or ""
     ):
         lines.append("古籍参考：本次未记录可核查的古籍片段")