app/tests/test_embeddings.py

"""Unit tests for the LangChain embedding factory.

LangChain experiment branch: the factory now returns a
``langchain_core.embeddings.Embeddings`` object instead of the custom
``EmbeddingClient`` interface.
"""

import pytest
from langchain_core.embeddings import Embeddings


def test_factory_returns_embeddings_object(monkeypatch: pytest.MonkeyPatch) -> None:
    """get_embedding_client() must return a LangChain Embeddings object."""
    import app.embeddings.factory as _factory

    monkeypatch.setattr(_factory, "_instance", None)
    client = _factory.get_embedding_client()
    assert isinstance(client, Embeddings)


def test_factory_singleton(monkeypatch: pytest.MonkeyPatch) -> None:
    """get_embedding_client() must return the same instance on repeated calls."""
    import app.embeddings.factory as _factory

    monkeypatch.setattr(_factory, "_instance", None)
    c1 = _factory.get_embedding_client()
    c2 = _factory.get_embedding_client()
    assert c1 is c2, "Factory returned a different instance on second call"


def test_factory_openai_when_key_set(monkeypatch: pytest.MonkeyPatch) -> None:
    """Factory should use OpenAIEmbeddings when openai_api_key is non-empty."""
    from langchain_openai import OpenAIEmbeddings

    import app.embeddings.factory as _factory

    monkeypatch.setattr(_factory, "_instance", None)
    fake_settings = type("S", (), {"openai_api_key": "sk-test", "embedding_model": "text-embedding-3-small"})()
    monkeypatch.setattr(_factory, "settings", fake_settings)

    client = _factory.get_embedding_client()
    assert isinstance(client, OpenAIEmbeddings)


def test_factory_huggingface_without_key(monkeypatch: pytest.MonkeyPatch) -> None:
    """Factory should fall back to HuggingFaceEmbeddings when no OpenAI key is set."""
    from langchain_huggingface import HuggingFaceEmbeddings

    import app.embeddings.factory as _factory

    monkeypatch.setattr(_factory, "_instance", None)
    fake_settings = type("S", (), {"openai_api_key": "", "local_embedding_model": "all-MiniLM-L6-v2"})()
    monkeypatch.setattr(_factory, "settings", fake_settings)

    client = _factory.get_embedding_client()
    assert isinstance(client, HuggingFaceEmbeddings)


def test_embeddings_embed_query_returns_vector(monkeypatch: pytest.MonkeyPatch) -> None:
    """embed_query() should return a non-empty list of floats."""
    import app.embeddings.factory as _factory

    monkeypatch.setattr(_factory, "_instance", None)
    client = _factory.get_embedding_client()
    vec = client.embed_query("The property is a Victorian terrace.")
    assert isinstance(vec, list)
    assert len(vec) > 0
    assert all(isinstance(v, float) for v in vec)


def test_embeddings_embed_documents_batch(monkeypatch: pytest.MonkeyPatch) -> None:
    """embed_documents() should return one vector per input text."""
    import app.embeddings.factory as _factory

    monkeypatch.setattr(_factory, "_instance", None)
    client = _factory.get_embedding_client()
    texts = ["first sentence", "second sentence", "third sentence"]
    vecs = client.embed_documents(texts)
    assert len(vecs) == 3
    assert all(len(v) > 0 for v in vecs)