Merge branch 'dev' into 'main'

Quellen wieder zugreifbar

See merge request kantz/tutor_react!10

math-tutor/backend/app/deterministic_services/referenz_decoder.py

 from __future__ import annotations
 
-from typing import Dict, Iterable, Tuple
+from typing import Dict, Iterable, Protocol, Set, Tuple
 import re
 from urllib.parse import quote
 
-from app.deterministic_services import Source, SourceID
-
 
 _BRACKET_RE = re.compile(r"\[(?P<ref>[^\[\]]+)\]")
 
 
-def _source_id_key(source_id: SourceID) -> str:
+class _SourceIDLike(Protocol):
+    chapter_title: str | None
+    section_title: str | None
+    subsection_title: str | None
+    title: str | None
+    doc_type: str | None
+
+    def to_string(self) -> str: ...
+
+
+class _SourceLike(Protocol):
+    source_id: _SourceIDLike
+
+
+def _source_id_key(source_id: _SourceIDLike) -> str:
     parts = [
         source_id.chapter_title or "",
         source_id.section_title or "",
@@ -30,9 +42,11 @@ def _normalize_ref_token(token: str) -> str:
     return t.casefold()
 
 
-def _build_source_index(sources: Iterable[Source]) -> Dict[str, str]:
+def _build_source_index(sources: Iterable[_SourceLike]) -> Tuple[Dict[str, str], Set[str]]:
     index: Dict[str, str] = {}
+    doc_types: Set[str] = set()
     for source in sources:
+        doc_types.add(_normalize_ref_token(source.source_id.doc_type or ""))
         key = _source_id_key(source.source_id)
         normalized_key = _normalize_ref_token(key)
         index.setdefault(normalized_key, key)
@@ -46,22 +60,46 @@ def _build_source_index(sources: Iterable[Source]) -> Dict[str, str]:
             source.source_id.subsection_title,
             source.source_id.title,
         ]
-        filtered = [p for p in parts if p and p.lower() != "none"]
-        if filtered:
-            minimal = "|".join(filtered + [source.source_id.doc_type])
-            index.setdefault(_normalize_ref_token(minimal), key)
+        for start in range(len(parts)):
+            filtered = [
+                p for p in parts[start:]
+                if p and p.lower() != "none"
+            ]
+            if filtered:
+                minimal = "|".join(filtered + [source.source_id.doc_type])
+                index.setdefault(_normalize_ref_token(minimal), key)
 
         if source.source_id.title:
             compact = f"{source.source_id.title}|{source.source_id.doc_type}"
             index.setdefault(_normalize_ref_token(compact), key)
-    return index
+    return index, {dtype for dtype in doc_types if dtype}
+
+
+def _candidate_tokens(token: str, doc_types: Set[str]) -> Iterable[str]:
+    normalized = _normalize_ref_token(token)
+    yield normalized
+
+    parts = [part for part in normalized.split("|") if part]
+    if not parts:
+        return
+
+    for idx, part in enumerate(parts):
+        if part in doc_types:
+            yield "|".join(parts[: idx + 1])
+
+    if len(parts) >= 2:
+        yield "|".join(parts[-2:])
+    if len(parts) >= 3:
+        yield "|".join(parts[-3:])
+    if len(parts) >= 4:
+        yield "|".join(parts[-4:])
 
 
-def decode_references(text: str, sources: Iterable[Source]) -> Tuple[str, int]:
+def decode_references(text: str, sources: Iterable[_SourceLike]) -> Tuple[str, int]:
     if not text:
         return text, 0
 
-    index = _build_source_index(sources)
+    index, doc_types = _build_source_index(sources)
     if not index:
         return text, 0
 
@@ -75,8 +113,11 @@ def decode_references(text: str, sources: Iterable[Source]) -> Tuple[str, int]:
         if end < len(text) and text[end:end + 1] == "(":
             return match.group(0)
 
-        normalized = _normalize_ref_token(ref)
-        key = index.get(normalized)
+        key = None
+        for candidate in _candidate_tokens(ref, doc_types):
+            key = index.get(candidate)
+            if key:
+                break
         if not key:
             return match.group(0)
 

math-tutor/backend/test/referenz_decoder_unit_test.py

+from __future__ import annotations
+
+from dataclasses import dataclass
+from pathlib import Path
+import importlib.util
+import unittest
+
+
+_MODULE_PATH = (
+    Path(__file__).resolve().parents[1]
+    / "app"
+    / "deterministic_services"
+    / "referenz_decoder.py"
+)
+_SPEC = importlib.util.spec_from_file_location("referenz_decoder", _MODULE_PATH)
+assert _SPEC is not None and _SPEC.loader is not None
+_MODULE = importlib.util.module_from_spec(_SPEC)
+_SPEC.loader.exec_module(_MODULE)
+decode_references = _MODULE.decode_references
+
+
+@dataclass
+class SourceIDStub:
+    chapter_title: str | None
+    section_title: str | None
+    subsection_title: str | None
+    title: str
+    doc_type: str
+
+    def to_string(self) -> str:
+        parts = [self.title]
+        if self.subsection_title:
+            parts.insert(0, self.subsection_title)
+        if self.section_title:
+            parts.insert(0, self.section_title)
+        if self.chapter_title:
+            parts.insert(0, self.chapter_title)
+        return f"[{'|'.join(parts)}|{self.doc_type}]"
+
+
+@dataclass
+class SourceStub:
+    source_id: SourceIDStub
+
+
+def _mk_source() -> SourceStub:
+    return SourceStub(
+        source_id=SourceIDStub(
+            chapter_title="Kapitel 1",
+            section_title="Lineare Funktionen",
+            subsection_title="Steigung",
+            title="Definition",
+            doc_type="subsection",
+        )
+    )
+
+
+class ReferenzDecoderUnitTest(unittest.TestCase):
+    def test_decodes_reference_without_chapter(self) -> None:
+        source = _mk_source()
+        text = "Siehe [Lineare Funktionen|Steigung|Definition|subsection]."
+
+        decoded, count = decode_references(text, [source])
+
+        self.assertEqual(count, 1)
+        self.assertIn(
+            "(doc://Kapitel%201%7CLineare%20Funktionen%7CSteigung%7CDefinition%7Csubsection)",
+            decoded,
+        )
+
+    def test_decodes_reference_with_extra_tail_segment(self) -> None:
+        source = _mk_source()
+        text = "Siehe [Lineare Funktionen/Steigung/Definition | subsection | ...]."
+
+        decoded, count = decode_references(text, [source])
+
+        self.assertEqual(count, 1)
+        self.assertIn(
+            "(doc://Kapitel%201%7CLineare%20Funktionen%7CSteigung%7CDefinition%7Csubsection)",
+            decoded,
+        )
+
+    def test_skips_existing_markdown_links(self) -> None:
+        source = _mk_source()
+        text = "Bereits verlinkt: [Definition](doc://foo)."
+
+        decoded, count = decode_references(text, [source])
+
+        self.assertEqual(count, 0)
+        self.assertEqual(decoded, text)
+
+
+if __name__ == "__main__":
+    unittest.main()

math-tutor/frontend/src/components/Chat/MessageBubble.tsx

 import { useEffect, useRef } from "react";
-import ReactMarkdown from "react-markdown";
+import ReactMarkdown, { defaultUrlTransform } from "react-markdown";
 import type { RetrievedDoc } from "../Retrieval/DocPanel";
 import { t } from "../../i18n";
 
@@ -54,6 +54,12 @@ export default function MessageBubble({
       </div>
       <div className="message-text">
         <ReactMarkdown
+          urlTransform={(url) => {
+            if (url.startsWith("doc://")) {
+              return url;
+            }
+            return defaultUrlTransform(url);
+          }}
           components={{
             a: ({ href, children, ...props }) => {
               const target = href || "";