test: verify real kbqa workflow

backend/tests/unit/test_file_validation.py

@@ -0,0 +1,13 @@
+from pathlib import Path
+
+import pytest
+
+from kbqa.api.errors import AppError
+from kbqa.indexing.loaders import load_document
+
+
+def test_text_loader_rejects_non_utf8(tmp_path: Path) -> None:
+    path = tmp_path / "bad.txt"
+    path.write_bytes(b"\xff\xfe")
+    with pytest.raises(AppError, match="UTF-8"):
+        load_document(path, "txt")

backend/tests/unit/test_splitter.py

@@ -0,0 +1,12 @@
+from kbqa.indexing.loaders import LoadedPage
+from kbqa.indexing.splitter import split_pages
+
+
+def test_split_pages_is_deterministic() -> None:
+    document_id = "12345678-1234-5678-1234-567812345678"
+    pages = [LoadedPage(content="知识库" * 300, page_number=1)]
+    first = split_pages(document_id, pages, 100, 10)
+    second = split_pages(document_id, pages, 100, 10)
+    assert [chunk.id for chunk in first] == [chunk.id for chunk in second]
+    assert first[0].page_number == 1
+    assert [chunk.order_index for chunk in first] == list(range(len(first)))

backend/tests/unit/test_sse.py

@@ -0,0 +1,8 @@
+from kbqa.chat.sse import encode_event, encode_ping
+
+
+def test_sse_encodes_single_line_utf8_json() -> None:
+    assert encode_event("token", {"content": "知识"}) == (
+        'event: token\ndata: {"content":"知识"}\n\n'.encode()
+    )
+    assert encode_ping() == b": ping\n\n"