Pablan, as it stands

Self-hosted knowledge management for SMEs: a split-screen Markdown editor
whose sections an LLM refines while you write, and RAG question answering
over the documents that result. FastAPI + Postgres/pgvector on the back,
SvelteKit on the front, everything OpenAI-compatible and self-hostable.

Squashed into a single commit; the development history stays local.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CA43ZJda8Rbp2hKXNy8f6b

backend/tests/test_excerpt.py

@@ -0,0 +1,82 @@
+"""The citation popover shows prose, not Markdown source.
+
+The popover is a small hover surface showing a fragment, so the excerpt is
+reduced to prose rather than rendered: a cited table would otherwise become
+a real table squeezed into ~320px, and a cited heading would render at h2
+size. Clicking the badge opens the document in the side panel, which does
+render Markdown properly.
+
+These cases are all real shapes from the fixture corpus.
+"""
+
+from app.modes.query import EXCERPT_CHARS
+from app.modes.query import excerpt as make_excerpt
+
+
+def test_a_cited_table_reads_as_prose() -> None:
+    """The worst case, and the one that sent this back for a second pass:
+    the divider row is pure punctuation and survives naive stripping."""
+    excerpt = make_excerpt(
+        "| Code | Bedeutung | Sofortmaßnahme |\n"
+        "|-------|------------|----------------|\n"
+        "| E-101 | Not-Aus-Kreis unterbrochen | Alle Not-Aus-Taster prüfen |"
+    )
+    assert "|" not in excerpt
+    assert "---" not in excerpt
+    assert excerpt.startswith("Code · Bedeutung · Sofortmaßnahme · E-101")
+
+
+def test_inline_emphasis_loses_its_markers_not_its_words() -> None:
+    for source, expected in [
+        ("**Eingang erfassen**: im ERP anlegen.", "Eingang erfassen: im ERP anlegen."),
+        ("die Presse muss *drucklos* sein", "die Presse muss drucklos sein"),
+        ("__Achtung__ beim Anfahren", "Achtung beim Anfahren"),
+        ("setze `max_turns` hoch", "setze max_turns hoch"),
+        ("**fett _und_ kursiv**", "fett und kursiv"),
+    ]:
+        assert make_excerpt(source) == expected
+
+
+def test_links_keep_their_text_and_drop_their_target() -> None:
+    excerpt = make_excerpt(
+        "Siehe [das Handbuch](https://intranet.example/doc.pdf) dazu."
+    )
+    assert excerpt == "Siehe das Handbuch dazu."
+    # A URL in a hover preview is noise, and it is also the part a reader
+    # cannot click here anyway.
+    assert "http" not in excerpt
+
+
+def test_block_markers_and_fences_are_dropped() -> None:
+    excerpt = make_excerpt(
+        "## Wartung\n\n"
+        "- **Getriebeöl GX-220**: monatlich\n"
+        "1. Bettbahnöl wöchentlich\n"
+        "> Wichtig: erst entlüften\n"
+        "```yaml\n"
+        "interval: 30\n"
+        "```"
+    )
+    assert excerpt.startswith("Wartung Getriebeöl GX-220: monatlich")
+    for marker in ("##", "```", "> ", "**"):
+        assert marker not in excerpt
+
+
+def test_plain_prose_is_left_alone() -> None:
+    source = "Die Kaffeemaschine wird freitags entkalkt."
+    assert make_excerpt(source) == source
+
+
+def test_underscores_inside_identifiers_survive() -> None:
+    """`result_document_id` is not italics. Emphasis needs a non-space
+    character on both sides, which is what keeps snake_case intact."""
+    assert make_excerpt("das Feld result_document_id bleibt leer") == (
+        "das Feld result_document_id bleibt leer"
+    )
+
+
+def test_long_content_is_cut_on_a_word_boundary() -> None:
+    excerpt = make_excerpt("Wartung " * 200)
+    assert len(excerpt) <= EXCERPT_CHARS + 1  # the ellipsis
+    assert excerpt.endswith("…")
+    assert not excerpt.rstrip("…").endswith(" ")