fix: plug all gaps found in deep review of indexing pipeline

surfsense_backend/tests/unit/conftest.py

@@ -1,3 +0,0 @@
-# No fixtures needed for unit tests yet.
-# Unit tests cover pure functions and value objects with no dependencies.
-# External-boundary mocks (llm, embedding_model) live in tests/integration/conftest.py.

surfsense_backend/tests/unit/indexing_pipeline/conftest.py

@@ -0,0 +1,15 @@
+import pytest
+
+
+@pytest.fixture
+def patched_chunker_instance(mocker):
+    mock = mocker.patch("app.indexing_pipeline.document_chunker.config.chunker_instance")
+    mock.chunk.return_value = [mocker.Mock(text="prose chunk")]
+    return mock
+
+
+@pytest.fixture
+def patched_code_chunker_instance(mocker):
+    mock = mocker.patch("app.indexing_pipeline.document_chunker.config.code_chunker_instance")
+    mock.chunk.return_value = [mocker.Mock(text="code chunk")]
+    return mock

surfsense_backend/tests/unit/indexing_pipeline/test_document_chunker.py

@@ -0,0 +1,19 @@
+import pytest
+
+from app.indexing_pipeline.document_chunker import chunk_text
+
+pytestmark = pytest.mark.unit
+
+
+def test_uses_code_chunker_when_flag_is_true(patched_code_chunker_instance):
+    result = chunk_text("def foo(): pass", use_code_chunker=True)
+
+    patched_code_chunker_instance.chunk.assert_called_once_with("def foo(): pass")
+    assert result == ["code chunk"]
+
+
+def test_uses_default_chunker_when_flag_is_false(patched_chunker_instance):
+    result = chunk_text("Some prose text.", use_code_chunker=False)
+
+    patched_chunker_instance.chunk.assert_called_once_with("Some prose text.")
+    assert result == ["prose chunk"]