backend/tests/test_reference_isolation.py

from __future__ import annotations

from backend.core import pii_scrubber
from backend.core.rag_store import TIER_MASTER, TIER_REFERENCE, Chunk, get_rag_store


def test_reference_ingest_strips_cross_property_identifiers():
    store = get_rag_store()
    raw = (
        "Survey for Mr John Smith at 22 High Street, London SE1 2AB. "
        "Contact john.smith@example.com. Report ref NCS-100147."
    )
    n = store.ingest_document(
        "t-ref",
        "reference:past_report.docx",
        [Chunk(text=raw, tier=TIER_REFERENCE)],
        tier=TIER_REFERENCE,
    )
    assert n == 1
    hits = store.search("t-ref", "survey high street", tier=TIER_REFERENCE, top_k=1)
    assert hits
    text = hits[0].text.lower()
    assert "john.smith@example.com" not in text
    assert "se1 2ab" not in text
    assert "ncs-100147" not in text
    assert hits[0].is_scrubbed is True


def test_past_report_not_visible_to_generation_search():
    store = get_rag_store()
    store.ingest_document(
        "t-gen",
        "ref:old",
        [Chunk(text="Property at 99 Baker Street NW1 6XE had subsidence.", tier=TIER_REFERENCE)],
        tier=TIER_REFERENCE,
    )
    store.ingest_document(
        "t-gen",
        "master:m",
        [Chunk(text="The site was inspected externally.", section_id="A", tier=TIER_MASTER)],
        tier=TIER_MASTER,
    )
    gen_hits = store.search_for_generation("t-gen", "property inspected", top_k=5)
    assert all(h.tier == TIER_MASTER for h in gen_hits)
    combined = " ".join(h.text for h in gen_hits).lower()
    assert "baker street" not in combined
    assert "nw1" not in combined


def test_scrub_reference_drops_unsafe_chunk():
    """Chunk is dropped when PII cannot be fully removed (simulated via detect)."""
    # Heavily redacted content should pass; empty after scrub should drop.
    result = pii_scrubber.scrub_reference_for_ingest("   ")
    assert result is None