File size: 13,377 Bytes
21ff762 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 | """
test_embedding_backend.py -- Tests for the Phase 2 embedding abstraction.
Network and model downloads are avoided: both backends are stubbed at the
import boundary so tests run in any CI environment.
"""
from __future__ import annotations
import sys
from dataclasses import fields
from pathlib import Path
from typing import Any, Sequence
from unittest.mock import MagicMock
import numpy as np
import pytest
SRC_DIR = Path(__file__).resolve().parents[1]
if str(SRC_DIR) not in sys.path:
sys.path.insert(0, str(SRC_DIR))
import embedding_backend as eb # noqa: E402
# ────────────────────────────────────────────────────────────────────
# Factory
# ────────────────────────────────────────────────────────────────────
def test_factory_defaults_to_sentence_transformers() -> None:
e = eb.get_embedder()
assert isinstance(e, eb.SentenceTransformerEmbedder)
assert e.name.startswith("sentence-transformers:")
@pytest.mark.parametrize("alias", ["sentence-transformers", "st", "sbert", "SBERT", ""])
def test_factory_accepts_st_aliases(alias: str) -> None:
assert isinstance(eb.get_embedder(alias), eb.SentenceTransformerEmbedder)
@pytest.mark.parametrize("alias", ["ollama", "Ollama", "OL"])
def test_factory_accepts_ollama_aliases(alias: str) -> None:
assert isinstance(eb.get_embedder(alias), eb.OllamaEmbedder)
def test_factory_rejects_unknown_backend() -> None:
with pytest.raises(ValueError, match="unknown embedding backend"):
eb.get_embedder("cohere")
def test_factory_applies_custom_model_on_ollama() -> None:
e = eb.get_embedder(
"ollama", model="mxbai-embed-large", base_url="http://localhost:11434"
)
assert isinstance(e, eb.OllamaEmbedder)
assert e.model_name == "mxbai-embed-large"
def test_factory_ollama_honours_env_url_when_local(
monkeypatch: pytest.MonkeyPatch,
) -> None:
monkeypatch.setenv("OLLAMA_URL", "http://127.0.0.1:9999")
e = eb.get_embedder("ollama")
assert isinstance(e, eb.OllamaEmbedder)
assert e.base_url == "http://127.0.0.1:9999"
def test_factory_ollama_env_url_non_local_is_rejected(
monkeypatch: pytest.MonkeyPatch,
) -> None:
monkeypatch.setenv("OLLAMA_URL", "http://169.254.169.254")
with pytest.raises(ValueError, match="not local"):
eb.get_embedder("ollama")
def test_factory_ollama_allow_remote_opt_in(
monkeypatch: pytest.MonkeyPatch,
) -> None:
monkeypatch.setenv("OLLAMA_URL", "http://my-remote-host:11434")
e = eb.get_embedder("ollama", allow_remote=True)
assert isinstance(e, eb.OllamaEmbedder)
assert e.base_url == "http://my-remote-host:11434"
# ────────────────────────────────────────────────────────────────────
# _l2_normalize
# ────────────────────────────────────────────────────────────────────
def test_l2_normalize_produces_unit_rows() -> None:
m = np.array([[3.0, 4.0], [1.0, 0.0], [0.0, 0.0]], dtype=np.float32)
out = eb._l2_normalize(m)
assert out.dtype == np.float32
np.testing.assert_allclose(np.linalg.norm(out[0]), 1.0, atol=1e-6)
np.testing.assert_allclose(np.linalg.norm(out[1]), 1.0, atol=1e-6)
np.testing.assert_array_equal(out[2], np.zeros(2, dtype=np.float32))
# ────────────────────────────────────────────────────────────────────
# SentenceTransformerEmbedder
# ────────────────────────────────────────────────────────────────────
class _FakeSTModel:
"""Minimal stand-in for sentence_transformers.SentenceTransformer."""
def __init__(self, dim: int = 8) -> None:
self._dim = dim
def get_sentence_embedding_dimension(self) -> int:
return self._dim
def encode(self, texts: Sequence[str], **_: Any) -> np.ndarray:
rows = []
for t in texts:
h = abs(hash(t))
rng = np.random.default_rng(h % (2**32))
rows.append(rng.normal(size=self._dim))
return np.asarray(rows, dtype=np.float32)
def _install_fake_st(monkeypatch: pytest.MonkeyPatch) -> None:
fake_module = MagicMock()
fake_module.SentenceTransformer = lambda model_name: _FakeSTModel()
monkeypatch.setitem(sys.modules, "sentence_transformers", fake_module)
def test_st_embedder_empty_input_returns_empty_matrix(
monkeypatch: pytest.MonkeyPatch,
) -> None:
_install_fake_st(monkeypatch)
e = eb.SentenceTransformerEmbedder()
out = e.embed([])
assert out.shape == (0, 0)
assert out.dtype == np.float32
def test_st_embedder_returns_normalised_matrix(
monkeypatch: pytest.MonkeyPatch,
) -> None:
_install_fake_st(monkeypatch)
e = eb.SentenceTransformerEmbedder()
out = e.embed(["hello", "world"])
assert out.shape == (2, 8)
norms = np.linalg.norm(out, axis=1)
np.testing.assert_allclose(norms, np.ones(2), atol=1e-6)
def test_st_embedder_sets_encode_batch_size(monkeypatch: pytest.MonkeyPatch) -> None:
class _CapturingSTModel(_FakeSTModel):
last_kwargs: dict[str, Any] = {}
def encode(self, texts: Sequence[str], **kwargs: Any) -> np.ndarray:
type(self).last_kwargs = dict(kwargs)
return super().encode(texts, **kwargs)
fake_module = MagicMock()
fake_module.SentenceTransformer = lambda model_name: _CapturingSTModel()
monkeypatch.setitem(sys.modules, "sentence_transformers", fake_module)
e = eb.SentenceTransformerEmbedder()
e.embed([f"text {i}" for i in range(600)])
assert _CapturingSTModel.last_kwargs["batch_size"] == eb._ST_ENCODE_BATCH_SIZE
def test_st_embedder_dim_is_minus_one_before_load(
monkeypatch: pytest.MonkeyPatch,
) -> None:
_install_fake_st(monkeypatch)
e = eb.SentenceTransformerEmbedder()
assert e.dim == -1
e.embed(["warmup"])
assert e.dim == 8
def test_st_embedder_missing_package_raises(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setitem(sys.modules, "sentence_transformers", None)
e = eb.SentenceTransformerEmbedder()
with pytest.raises(RuntimeError, match="sentence-transformers is not installed"):
e.embed(["hi"])
def test_st_embedder_model_field_is_not_in_init() -> None:
# ``_model`` must not be injectable via the constructor — prevents
# callers from slipping in a fake implementation.
init_fields = {f.name for f in fields(eb.SentenceTransformerEmbedder) if f.init}
assert "_model" not in init_fields
# The field still exists on instances (default None).
assert eb.SentenceTransformerEmbedder()._model is None
# ────────────────────────────────────────────────────────────────────
# OllamaEmbedder — SSRF guard
# ────────────────────────────────────────────────────────────────────
@pytest.mark.parametrize(
"bad_url",
[
"http://169.254.169.254", # AWS IMDS
"http://metadata.google.internal",
"http://10.0.0.5",
"http://internal-service.corp",
],
)
def test_ollama_rejects_non_local_host_by_default(bad_url: str) -> None:
with pytest.raises(ValueError, match="not local"):
eb.OllamaEmbedder(base_url=bad_url)
@pytest.mark.parametrize(
"bad_url",
[
"file:///etc/passwd",
"ftp://localhost:11434",
"gopher://localhost",
"",
"not-a-url",
],
)
def test_ollama_rejects_bad_scheme_or_missing_host(bad_url: str) -> None:
with pytest.raises(ValueError):
eb.OllamaEmbedder(base_url=bad_url)
@pytest.mark.parametrize(
"good_url",
[
"http://localhost:11434",
"http://127.0.0.1",
"https://localhost",
"http://[::1]:11434",
],
)
def test_ollama_accepts_local_hosts(good_url: str) -> None:
e = eb.OllamaEmbedder(base_url=good_url)
assert e.base_url == good_url
def test_ollama_allow_remote_opt_in() -> None:
e = eb.OllamaEmbedder(base_url="http://my-gpu-box:11434", allow_remote=True)
assert e.allow_remote is True
# ────────────────────────────────────────────────────────────────────
# OllamaEmbedder — embed()
# ────────────────────────────────────────────────────────────────────
class _FakeResponse:
def __init__(self, payload: dict[str, Any], status: int = 200) -> None:
self._payload = payload
self.status_code = status
def raise_for_status(self) -> None:
if self.status_code >= 400:
raise RuntimeError(f"HTTP {self.status_code}")
def json(self) -> dict[str, Any]:
return self._payload
def _install_fake_requests(
monkeypatch: pytest.MonkeyPatch,
vectors: list[list[float]],
*,
missing_key: bool = False,
) -> list[dict[str, Any]]:
calls: list[dict[str, Any]] = []
queue = list(vectors)
def fake_post(url: str, *, json: dict[str, Any], timeout: float) -> _FakeResponse:
calls.append({"url": url, "json": json, "timeout": timeout})
if missing_key:
return _FakeResponse({"not_embedding": queue.pop(0)})
return _FakeResponse({"embedding": queue.pop(0)})
fake_module = MagicMock()
fake_module.post = fake_post
monkeypatch.setitem(sys.modules, "requests", fake_module)
return calls
def test_ollama_embedder_empty_input_returns_empty() -> None:
out = eb.OllamaEmbedder().embed([])
assert out.shape == (0, 0)
def test_ollama_embedder_posts_per_text_and_normalises(
monkeypatch: pytest.MonkeyPatch,
) -> None:
calls = _install_fake_requests(
monkeypatch,
vectors=[[3.0, 4.0], [1.0, 0.0]],
)
e = eb.OllamaEmbedder(base_url="http://localhost:11434")
out = e.embed(["a", "b"])
assert len(calls) == 2
assert calls[0]["url"] == "http://localhost:11434/api/embeddings"
assert calls[0]["json"] == {"model": eb.DEFAULT_OLLAMA_MODEL, "prompt": "a"}
np.testing.assert_allclose(np.linalg.norm(out, axis=1), np.ones(2), atol=1e-6)
assert out.dtype == np.float32
def test_ollama_embedder_missing_key_raises_with_index(
monkeypatch: pytest.MonkeyPatch,
) -> None:
_install_fake_requests(monkeypatch, vectors=[[1.0]], missing_key=True)
with pytest.raises(eb.OllamaEmbedderError, match="text #0") as exc_info:
eb.OllamaEmbedder().embed(["x"])
assert exc_info.value.index == 0
assert "missing 'embedding' key" in str(exc_info.value)
def test_ollama_embedder_partial_failure_reports_index(
monkeypatch: pytest.MonkeyPatch,
) -> None:
# First two succeed, third raises — error must carry index=2.
fake_module = MagicMock()
call_count = {"n": 0}
def fake_post(url: str, *, json: dict[str, Any], timeout: float) -> _FakeResponse:
call_count["n"] += 1
if call_count["n"] <= 2:
return _FakeResponse({"embedding": [float(call_count["n"])]})
raise RuntimeError("boom")
fake_module.post = fake_post
monkeypatch.setitem(sys.modules, "requests", fake_module)
with pytest.raises(eb.OllamaEmbedderError, match="text #2") as exc_info:
eb.OllamaEmbedder().embed(["a", "b", "c"])
assert exc_info.value.index == 2
def test_ollama_embedder_missing_requests_raises(
monkeypatch: pytest.MonkeyPatch,
) -> None:
monkeypatch.setitem(sys.modules, "requests", None)
with pytest.raises(RuntimeError, match="requests is required"):
eb.OllamaEmbedder().embed(["x"])
def test_ollama_embedder_reports_known_dim_for_nomic() -> None:
assert eb.OllamaEmbedder().dim == eb._NOMIC_EMBED_TEXT_DIM
assert eb.OllamaEmbedder(model_name="other").dim == -1
# ────────────────────────────────────────────────────────────────────
# Protocol conformance
# ────────────────────────────────────────────────────────────────────
def test_both_backends_conform_to_protocol() -> None:
assert isinstance(eb.SentenceTransformerEmbedder(), eb.Embedder)
assert isinstance(eb.OllamaEmbedder(), eb.Embedder)
|