breakpilot-core/rag-service/tests/test_document_upload.py

"""Tests for document upload payload building — structural metadata (D2)."""

# Mirror the constant from api/documents.py to avoid heavy import chain
# (api → jose, qdrant_client, minio, etc.)
_STRUCT_FIELDS = ("section", "section_title", "paragraph", "paragraph_num", "page")


def _build_payload(
    chunk: str,
    index: int,
    chunks_meta: list[dict],
    extra_metadata: "dict | None" = None,
) -> dict:
    """Replicate the payload-building logic from documents.py for unit testing."""
    payload = {
        "document_id": "test-doc-id",
        "object_name": "test/path.pdf",
        "filename": "path.pdf",
        "chunk_index": index,
        "chunk_text": chunk,
        "data_type": "law",
        "bundesland": "bund",
        "use_case": "compliance",
        "year": "2026",
        **(extra_metadata or {}),
    }
    if index < len(chunks_meta):
        meta = chunks_meta[index]
        for field in _STRUCT_FIELDS:
            value = meta.get(field)
            if value is not None and value != "":
                payload[field] = value
    return payload


class TestPayloadStructuralMetadata:
    """Tests for structural metadata merging into Qdrant payloads."""

    def test_payload_contains_structural_metadata(self):
        """Metadata fields from chunks_with_metadata land in the payload."""
        meta = [
            {
                "text": "chunk text",
                "section": "§ 312k",
                "section_title": "Kuendigungsbutton",
                "paragraph": "Abs. 1",
                "paragraph_num": 1,
                "page": 847,
                "index": 0,
            }
        ]

        payload = _build_payload("chunk text", 0, meta)

        assert payload["section"] == "§ 312k"
        assert payload["section_title"] == "Kuendigungsbutton"
        assert payload["paragraph"] == "Abs. 1"
        assert payload["paragraph_num"] == 1
        assert payload["page"] == 847

    def test_payload_without_metadata_backwards_compat(self):
        """Empty metadata list → payload has no structural fields."""
        payload = _build_payload("chunk text", 0, [])

        for field in _STRUCT_FIELDS:
            assert field not in payload

    def test_payload_skips_empty_values(self):
        """Empty string and None values are NOT added to payload."""
        meta = [
            {
                "text": "chunk text",
                "section": "",
                "section_title": "",
                "paragraph": "",
                "paragraph_num": None,
                "page": None,
                "index": 0,
            }
        ]

        payload = _build_payload("chunk text", 0, meta)

        for field in _STRUCT_FIELDS:
            assert field not in payload

    def test_metadata_overrides_extra_metadata(self):
        """Auto-extracted metadata takes precedence over manual extra_metadata."""
        meta = [
            {
                "text": "chunk text",
                "section": "§ 25",
                "section_title": "",
                "paragraph": "",
                "paragraph_num": None,
                "page": None,
                "index": 0,
            }
        ]
        extra = {"section": "manual-value"}

        payload = _build_payload("chunk text", 0, meta, extra_metadata=extra)

        assert payload["section"] == "§ 25"

    def test_partial_metadata_alignment(self):
        """3 chunks but only 2 metadata entries → third payload has no structural fields."""
        meta = [
            {
                "text": "c1",
                "section": "§ 1",
                "section_title": "",
                "paragraph": "",
                "paragraph_num": None,
                "page": None,
                "index": 0,
            },
            {
                "text": "c2",
                "section": "§ 2",
                "section_title": "",
                "paragraph": "",
                "paragraph_num": None,
                "page": None,
                "index": 1,
            },
        ]

        p0 = _build_payload("c1", 0, meta)
        p1 = _build_payload("c2", 1, meta)
        p2 = _build_payload("c3", 2, meta)

        assert p0["section"] == "§ 1"
        assert p1["section"] == "§ 2"
        assert "section" not in p2

    def test_zero_paragraph_num_is_kept(self):
        """paragraph_num=0 is a valid value and should be stored."""
        meta = [
            {
                "text": "chunk",
                "section": "",
                "section_title": "",
                "paragraph": "",
                "paragraph_num": 0,
                "page": None,
                "index": 0,
            }
        ]

        payload = _build_payload("chunk", 0, meta)

        # 0 is not None and not "" → should be stored
        assert payload["paragraph_num"] == 0

    def test_page_zero_is_kept(self):
        """page=0 is a valid value (first page) and should be stored."""
        meta = [
            {
                "text": "chunk",
                "section": "",
                "section_title": "",
                "paragraph": "",
                "paragraph_num": None,
                "page": 0,
                "index": 0,
            }
        ]

        payload = _build_payload("chunk", 0, meta)

        assert payload["page"] == 0