fix: address chunker review feedback

src/companion/rag/chunker.py

@@ -45,7 +45,7 @@ def sliding_window_chunks(text: str, chunk_size: int, chunk_overlap: int) -> Lis
 
 def section_based_chunks(
     text: str, section_tags: List[str], chunk_size: int, chunk_overlap: int
-) -> List[tuple]:
+) -> List[tuple[str, str]]:
     lines = text.split("\n")
     sections = []
     current_tag = None
@@ -100,7 +100,9 @@ def match_chunking_rule(
             continue
         if fnmatch.fnmatch(relative_path, pattern):
             return rule
-    return rules.get("default")
+    return rules.get("default") or ChunkingRule(
+        strategy="sliding_window", chunk_size=500, chunk_overlap=100
+    )
 
 
 def chunk_file(
@@ -126,9 +128,9 @@ def chunk_file(
         )
         chunks = []
         for i, (chunk_text, section) in enumerate(raw_chunks):
-            section_hashtags = re.findall(r"#\w+", chunk_text)
+            section_hashtags = re.findall(r"#[\w\-]+", chunk_text)
             section_wikilinks = re.findall(r"\[\[(.*?)\]\]", chunk_text)
-            section_tags = list(set(section_hashtags + section_wikilinks))
+            section_tags = list(dict.fromkeys(section_hashtags + section_wikilinks))
 
             chunk = Chunk(
                 text=chunk_text,
@@ -147,9 +149,9 @@ def chunk_file(
         raw_chunks = sliding_window_chunks(text, rule.chunk_size, rule.chunk_overlap)
         chunks = []
         for i, chunk_text in enumerate(raw_chunks):
-            chunk_hashtags = re.findall(r"#\w+", chunk_text)
+            chunk_hashtags = re.findall(r"#[\w\-]+", chunk_text)
             chunk_wikilinks = re.findall(r"\[\[(.*?)\]\]", chunk_text)
-            chunk_tags = list(set(chunk_hashtags + chunk_wikilinks))
+            chunk_tags = list(dict.fromkeys(chunk_hashtags + chunk_wikilinks))
 
             chunk = Chunk(
                 text=chunk_text,

tests/test_chunker.py

@@ -1,4 +1,12 @@
-from companion.rag.chunker import sliding_window_chunks
+import tempfile
+from pathlib import Path
+
+from companion.rag.chunker import (
+    sliding_window_chunks,
+    section_based_chunks,
+    chunk_file,
+    ChunkingRule,
+)
 
 
 def test_sliding_window_basic():
@@ -10,14 +18,13 @@ def test_sliding_window_basic():
     last_five = chunks[0].split()[-5:]
     first_chunk1 = chunks[1].split()[:5]
     assert last_five == first_chunk1
-from companion.rag.chunker import section_based_chunks, chunk_file, ChunkingRule
-import tempfile
-from pathlib import Path
 
 
 def test_section_based_chunks_splits_on_tags():
     text = "#DayInShort: good day\n#mentalhealth: stressed\n#work: busy"
-    chunks = section_based_chunks(text, ["#DayInShort", "#mentalhealth", "#work"], chunk_size=10, chunk_overlap=2)
+    chunks = section_based_chunks(
+        text, ["#DayInShort", "#mentalhealth", "#work"], chunk_size=10, chunk_overlap=2
+    )
     assert len(chunks) == 3
     assert chunks[0][1] == "#DayInShort"
     assert chunks[1][1] == "#mentalhealth"
@@ -29,13 +36,72 @@ def test_chunk_file_extracts_metadata():
         vault = Path(tmp)
         journal = vault / "Journal" / "2026" / "04" / "2026-04-12.md"
         journal.parent.mkdir(parents=True)
-        journal.write_text("#DayInShort: good day\n#Relations: [[Person/Vinay]] visited.", encoding="utf-8")
+        journal.write_text(
+            "#DayInShort: good day\n#Relations: [[Person/Vinay]] visited.",
+            encoding="utf-8",
+        )
         rules = {
-            "default": ChunkingRule(strategy="sliding_window", chunk_size=500, chunk_overlap=100),
+            "default": ChunkingRule(
-            "Journal/**": ChunkingRule(strategy="section", chunk_size=300, chunk_overlap=50, section_tags=["#DayInShort", "#Relations"]),
+                strategy="sliding_window", chunk_size=500, chunk_overlap=100
+            ),
+            "Journal/**": ChunkingRule(
+                strategy="section",
+                chunk_size=300,
+                chunk_overlap=50,
+                section_tags=["#DayInShort", "#Relations"],
+            ),
         }
         chunks = chunk_file(journal, vault, rules, modified_at=1234567890.0)
         assert len(chunks) == 2
         assert chunks[0].source_directory == "Journal"
         assert chunks[0].date == "2026-04-12"
         assert "Person/Vinay" in (chunks[1].tags or [])
+
+
+def test_sliding_window_empty_input():
+    chunks = sliding_window_chunks("", chunk_size=20, chunk_overlap=5)
+    assert chunks == []
+
+
+def test_chunk_file_root_level_file():
+    with tempfile.TemporaryDirectory() as tmp:
+        vault = Path(tmp)
+        note = vault / "note.md"
+        note.write_text("hello world", encoding="utf-8")
+        rules = {
+            "default": ChunkingRule(
+                strategy="sliding_window", chunk_size=500, chunk_overlap=100
+            )
+        }
+        chunks = chunk_file(note, vault, rules, modified_at=1.0)
+        assert len(chunks) == 1
+        assert chunks[0].source_directory == ""
+        assert chunks[0].source_file == "note.md"
+
+
+def test_chunk_file_preserves_tag_order():
+    with tempfile.TemporaryDirectory() as tmp:
+        vault = Path(tmp)
+        note = vault / "tags.md"
+        note.write_text("#first #second #first #third", encoding="utf-8")
+        rules = {
+            "default": ChunkingRule(
+                strategy="sliding_window", chunk_size=500, chunk_overlap=100
+            )
+        }
+        chunks = chunk_file(note, vault, rules, modified_at=1.0)
+        assert chunks[0].tags == ["#first", "#second", "#third"]
+
+
+def test_chunk_file_hyphenated_hashtags():
+    with tempfile.TemporaryDirectory() as tmp:
+        vault = Path(tmp)
+        note = vault / "hyphen.md"
+        note.write_text("#mental-health check-in", encoding="utf-8")
+        rules = {
+            "default": ChunkingRule(
+                strategy="sliding_window", chunk_size=500, chunk_overlap=100
+            )
+        }
+        chunks = chunk_file(note, vault, rules, modified_at=1.0)
+        assert "#mental-health" in chunks[0].tags