fix: collection field extraction and add regression tests for sort-key and nested layout

src/buddhagpt/corpus.py

@@ -23,9 +23,11 @@ def load_bilara_suttas(root: Path) -> Iterator[dict]:
         segments = json.loads(f.read_text())
         uid = f.name.split("_")[0]
         title = next(iter(segments.values()), uid)
+        # Derive collection from first directory under sutta/ (e.g., an, dn, kn, mn, sn)
+        collection = f.relative_to(base).parts[0]
         yield {
             "uid": uid,
             "title": title.strip(),
             "text": segments_to_text(segments),
-            "collection": f.parent.name,
+            "collection": collection,
         }

tests/test_corpus.py

@@ -6,7 +6,22 @@ def test_segments_to_text_joins_in_key_order():
     segs = {"mn21:1.2": "second.", "mn21:1.1": "First."}
     assert segments_to_text(segs) == "First. second."
 
+def test_segments_to_text_handles_range_keys():
+    """Regression test: range keys like 'an1.142-149:1.0' must not crash.
+    This would fail with the brief's original int/str-mixing sort key."""
+    segs = {
+        "an1.142-149:1.0": "Range intro.",
+        "an1.142-149:1.1": "First sentence.",
+        "an1.140:1.0": "Earlier part.",
+    }
+    result = segments_to_text(segs)
+    # Should not raise TypeError; should contain all texts
+    assert "Range intro." in result
+    assert "First sentence." in result
+    assert "Earlier part." in result
+
 def test_load_bilara_suttas(tmp_path):
+    """Original test: flat structure."""
     d = tmp_path / "translation/en/sujato/sutta/mn"
     d.mkdir(parents=True)
     (d / "mn21_translation-en-sujato.json").write_text(
@@ -15,3 +30,17 @@ def test_load_bilara_suttas(tmp_path):
     suttas = list(load_bilara_suttas(tmp_path))
     assert suttas[0]["uid"] == "mn21"
     assert "So I have heard." in suttas[0]["text"]
+    assert suttas[0]["collection"] == "mn"
+
+def test_load_bilara_suttas_nested_layout(tmp_path):
+    """Regression test: nested structure (an/an1/, sn/sn35/, etc.)
+    verifies collection is extracted from top-level nikaya, not immediate parent."""
+    d = tmp_path / "translation/en/sujato/sutta/an/an1"
+    d.mkdir(parents=True)
+    (d / "an1.1-10_translation-en-sujato.json").write_text(
+        json.dumps({"an1.1-10:0.1": "AN 1.1-10", "an1.1-10:1.1": "Content."})
+    )
+    suttas = list(load_bilara_suttas(tmp_path))
+    assert suttas[0]["uid"] == "an1.1-10"
+    assert suttas[0]["collection"] == "an", f"Expected 'an', got '{suttas[0]['collection']}'"
+    assert "Content." in suttas[0]["text"]