Orchestrator/bahn/teamlandkarte-mcp/tests/test_search_cache.py

from __future__ import annotations

from teamlandkarte_mcp.cache.search_cache import SearchCache


def test_filter_id_increments_under_same_search_id() -> None:
    cache = SearchCache(ttl_minutes=60, max_size=10)
    search_id = cache.store_search(
        task_id=None, requirements={}, results={"by_category": {}}
    )

    filter_id_1 = cache.add_filter(
        search_id, filtered_results={"by_category": {}}, filter_meta={}
    )
    filter_id_2 = cache.add_filter(
        search_id, filtered_results={"by_category": {}}, filter_meta={}
    )

    assert filter_id_1 == "filter-1"
    assert filter_id_2 == "filter-2"

    entry = cache.get(search_id)
    assert entry is not None
    assert filter_id_1 in entry.filters
    assert filter_id_2 in entry.filters


# ---------------------------------------------------------------------------
# Team-search coverage (Profile_Type="team")
# ---------------------------------------------------------------------------
#
# The SearchCache itself is search-type-agnostic - it stores whatever
# ``results`` payload it receives. The tests below exercise the same
# cache primitives (``store_search``, ``add_filter``, ``get``) for
# ``team_search`` payloads in both ``matching_method`` flavours
# (``score`` and ``llm_fulltext``) so the entire cache surface is
# covered for both ``Profile_Type`` values.
#
# _Requirements: 8.1, 8.2, 8.3, 12.7_


def _team_payload(matching_method: str) -> dict:
    """Return a minimal team_search payload mimicking the shapes that
    ``find_matching_teams`` persists in either ``matching_method`` mode.
    """
    base_team = {
        "team_id": "t1",
        "ouid": "ou-1",
        "team_name": "Team Alpha",
        "focus_name": "Backend Developer",
        "about_us": "",
        "offerings": "",
        "interests": "",
        "competences": [{"name": "python", "top_competency": True}],
        "references": [],
        "category": "Top",
    }
    if matching_method == "score":
        scored_team = {
            **base_team,
            "competence_score": 1.0,
            "role_score": 1.0,
            "overall_score": 1.0,
        }
        return {
            "search_type": "team_search",
            "matching_method": "score",
            "reference": {
                "availability_date_start": None,
                "availability_date_end": None,
            },
            "summary": {"Top": 1, "Good": 0, "Partial": 0, "Low": 0,
                        "Irrelevant": 0},
            "by_category": {
                "Top": [scored_team],
                "Good": [],
                "Partial": [],
                "Low": [],
                "Irrelevant": [],
            },
        }
    # llm_fulltext
    llm_team = {**base_team, "rationale": "Backend match."}
    return {
        "search_type": "team_search",
        "matching_method": "llm_fulltext",
        "reference": {
            "availability_date_start": None,
            "availability_date_end": None,
        },
        "summary": {"Top": 1, "Good": 0, "Partial": 0, "Low": 0,
                    "Irrelevant": 0},
        "by_category": {
            "Top": [llm_team],
            "Good": [],
            "Partial": [],
            "Low": [],
            "Irrelevant": [],
        },
        "errors": [],
    }


def test_team_search_cache_roundtrip_score_mode() -> None:
    """``team_search`` payloads round-trip through the cache and keep
    their ``search_type``/``matching_method`` markers (Anforderungen
    8.1, 8.2, 8.3)."""
    cache = SearchCache(ttl_minutes=60, max_size=10)
    payload = _team_payload("score")

    search_id = cache.store_search(
        task_id=None,
        requirements={"role_name": "Backend Developer",
                      "competences": ["python"]},
        results=payload,
    )

    entry = cache.get(search_id)
    assert entry is not None
    assert entry.results["search_type"] == "team_search"
    assert entry.results["matching_method"] == "score"
    # The cached items keep the team-specific fields used by the
    # team-search column layout (Anforderung 8.4).
    top = entry.results["by_category"]["Top"]
    assert top and top[0]["team_id"] == "t1"
    assert "competence_score" in top[0]
    assert "role_score" in top[0]
    assert "overall_score" in top[0]


def test_team_search_cache_roundtrip_llm_fulltext_mode() -> None:
    """``team_search`` LLM-mode payloads survive a cache round trip
    including the ``rationale`` field used by ``Begründung``."""
    cache = SearchCache(ttl_minutes=60, max_size=10)
    payload = _team_payload("llm_fulltext")

    search_id = cache.store_search(
        task_id=None,
        requirements={"role_name": "Backend Developer",
                      "competences": ["python"]},
        results=payload,
    )

    entry = cache.get(search_id)
    assert entry is not None
    assert entry.results["search_type"] == "team_search"
    assert entry.results["matching_method"] == "llm_fulltext"
    # ``errors`` list is preserved for the LLM-fulltext path
    # (Anforderung 7.6).
    assert entry.results.get("errors") == []
    top = entry.results["by_category"]["Top"]
    assert top and top[0]["team_id"] == "t1"
    assert top[0]["category"] == "Top"
    assert top[0]["rationale"] == "Backend match."


def test_team_search_filter_ids_increment_independently() -> None:
    """``add_filter`` produces incrementing ids for ``team_search``
    entries just like for capacity searches.
    """
    cache = SearchCache(ttl_minutes=60, max_size=10)
    search_id = cache.store_search(
        task_id=None,
        requirements={"role_name": "Backend Developer",
                      "competences": ["python"]},
        results=_team_payload("score"),
    )

    fid1 = cache.add_filter(
        search_id,
        filtered_results={"by_category": {}, "search_type": "team_search",
                          "matching_method": "score"},
        filter_meta={"role_filter": "Backend"},
    )
    fid2 = cache.add_filter(
        search_id,
        filtered_results={"by_category": {}, "search_type": "team_search",
                          "matching_method": "score"},
        filter_meta={"role_filter": "Frontend"},
    )

    assert fid1 == "filter-1"
    assert fid2 == "filter-2"

    entry = cache.get(search_id)
    assert entry is not None
    # The filter payloads keep the team_search marker independent of
    # the parent entry, so downstream consumers can distinguish them.
    for fid in (fid1, fid2):
        fdata = entry.filters[fid]
        assert fdata["results"]["search_type"] == "team_search"
        assert fdata["results"]["matching_method"] == "score"