Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.
Python aicode-complexitycode-healthcode-intelligencecode-qualitydead-codedeveloper-toolsdocumentationgit-analyticsmcpopen-sourcepythonrefactoringstatic-analysistechnical-debt
"""``REPOWISE_EMBEDDING_DIMS`` must not reach any embedder constructor unvalidated.
``_server.py:171`` passed ``int(dims) if dims else 768`` straight through, so
``REPOWISE_EMBEDDING_DIMS=-5`` gave the gemini embedder a negative width with
no warning. The same hole existed in ``app.py`` for the non-MCP code path.
These tests verify that non-positive, non-numeric, and otherwise malformed
values fall back to 768 and reach stderr, matching the pattern
``resolve_embedding_timeout`` already follows for timeout env vars.
"""
from __future__ import annotations
import pytest
from repowise.server.mcp_server._server import _embedder_kwargs
def _clear_dims(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.delenv("REPOWISE_EMBEDDING_DIMS", raising=False)
monkeypatch.delenv("REPOWISE_EMBEDDING_MODEL", raising=False)
def test_valid_dims_are_passed_through(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_dims(monkeypatch)
monkeypatch.setenv("REPOWISE_EMBEDDING_DIMS", "256")
kwargs = _embedder_kwargs("gemini")
assert kwargs["output_dimensionality"] == 256
def test_missing_dims_leave_the_constructor_default(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_dims(monkeypatch)
kwargs = _embedder_kwargs("gemini")
# GeminiEmbedder's own default (768) applies; it is declared only there.
assert "output_dimensionality" not in kwargs
@pytest.mark.parametrize("bad", ["-5", "0", "-1", "abc", "3.14", "inf", "nan", ""])
def test_bad_dims_fall_back_to_the_default(
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture, bad: str
) -> None:
_clear_dims(monkeypatch)
if bad:
monkeypatch.setenv("REPOWISE_EMBEDDING_DIMS", bad)
kwargs = _embedder_kwargs("gemini")
assert "output_dimensionality" not in kwargs
if bad:
assert "not a positive integer" in capsys.readouterr().err
def test_non_gemini_does_not_read_dims(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_dims(monkeypatch)
monkeypatch.setenv("REPOWISE_EMBEDDING_DIMS", "256")
kwargs = _embedder_kwargs("openai")
assert "output_dimensionality" not in kwargs
# -- Direct constructor guards (OllamaEmbedder) --
def test_ollama_rejects_negative_dimensions() -> None:
from repowise.core.providers.embedding.ollama import OllamaEmbedder
with pytest.raises(ValueError, match="positive"):
OllamaEmbedder(dimensions=-5)
def test_ollama_rejects_zero_dimensions() -> None:
from repowise.core.providers.embedding.ollama import OllamaEmbedder
with pytest.raises(ValueError, match="positive"):
OllamaEmbedder(dimensions=0)
def test_ollama_bad_env_dims_fall_back(monkeypatch: pytest.MonkeyPatch) -> None:
from repowise.core.providers.embedding.ollama import OllamaEmbedder
_clear_dims(monkeypatch)
monkeypatch.delenv("OLLAMA_EMBEDDING_DIMS", raising=False)
monkeypatch.setenv("REPOWISE_EMBEDDING_DIMS", "-5")
embedder = OllamaEmbedder()
# Should not crash; dimensions fall back to model inference
assert embedder.dimensions > 0