"""Graph engines, MinIO archiver, and the ask bridge — contract tests
that run without any service or heavy dependency installed."""

from __future__ import annotations

import pytest

from thicket.graph_store import GRAPH_ENGINES, GraphUnavailable, create_graph


def test_graph_engine_registry():
    assert sorted(GRAPH_ENGINES) == ["graphify", "lightrag"]
    assert GRAPH_ENGINES["lightrag"]["modules"] == ("lightrag",)
    assert GRAPH_ENGINES["graphify"]["modules"] == ("graphify",)


def test_unknown_graph_engine_names_choices(tmp_path):
    with pytest.raises(GraphUnavailable, match="known:"):
        create_graph("memgraph", working_dir=tmp_path, llm_model="m",
                     embed_model="e")


def test_graphify_engine_stages_documents(tmp_path):
    pytest.importorskip("graphify")
    store = create_graph("graphify", working_dir=tmp_path, llm_model="m",
                         embed_model="e")
    store.ingest_document("Doc One", "alpha text")
    store.ingest_document("Doc Two", "beta text")
    staged = list((tmp_path / ".graphify" / "corpus").glob("*.md"))
    assert len(staged) == 2
    store.ingest_document("Doc One", "different content this time")
    # same title + different content -> a digest sibling, not a clobber
    assert len(list((tmp_path / ".graphify" / "corpus").glob("doc-one*.md"))) == 2


def test_minio_archiver_requires_package():
    from thicket.minio_archive import ArchiveUnavailable, MinioArchiver
    import importlib.util
    if importlib.util.find_spec("minio"):
        pytest.skip("minio installed — guard path covered live")
    with pytest.raises(ArchiveUnavailable, match="pip install"):
        MinioArchiver()


def test_ask_rejects_non_sql_targets():
    from thicket.ask_vanna import AskUnavailable, SQL_TARGETS, target_ddl
    assert SQL_TARGETS == ("pgvector", "mariadb")
    with pytest.raises(AskUnavailable, match="no SQL corpus"):
        target_ddl("chroma")


def test_ask_ddl_documents_payload_fields():
    from thicket.ask_vanna import target_ddl
    for target in ("pgvector", "mariadb"):
        ddl = target_ddl(target, "second_brain")
        assert "second_brain" in ddl
        assert "document_title" in ddl and "chunk_index" in ddl