feat: add CLIP embedding service sidecar for GPU machine

FastAPI service that loads ViT-L-14 model and exposes /embed,
/similarity, and /batch-similarity endpoints. Runs on RTX 3090
with NVIDIA GPU reservation. Includes 9 cosine similarity tests.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

clip-service/tests/conftest.py

@@ -0,0 +1,65 @@
+"""Test configuration: mock heavy dependencies that aren't needed for unit tests."""
+
+import sys
+from unittest.mock import MagicMock
+
+# Mock modules that are not installed in the test environment.
+# The tests only exercise pure numpy math, not model inference.
+for mod_name in [
+    "torch",
+    "open_clip",
+    "fastapi",
+    "fastapi.responses",
+    "pydantic",
+    "PIL",
+    "PIL.Image",
+    "uvicorn",
+]:
+    if mod_name not in sys.modules:
+        sys.modules[mod_name] = MagicMock()
+
+# Provide a real BaseModel stand-in so Pydantic models can be defined
+# (they are referenced at import time in main.py).
+import types
+
+pydantic_mock = sys.modules["pydantic"]
+
+
+class _FakeBaseModel:
+    def __init_subclass__(cls, **kwargs):
+        pass
+
+    def __init__(self, **kwargs):
+        for k, v in kwargs.items():
+            setattr(self, k, v)
+
+
+pydantic_mock.BaseModel = _FakeBaseModel
+
+# Provide a fake FastAPI and HTTPException so the module loads
+fastapi_mock = sys.modules["fastapi"]
+
+
+class _FakeFastAPI:
+    def __init__(self, **kwargs):
+        pass
+
+    def get(self, *a, **kw):
+        def decorator(fn):
+            return fn
+        return decorator
+
+    def post(self, *a, **kw):
+        def decorator(fn):
+            return fn
+        return decorator
+
+
+class _FakeHTTPException(Exception):
+    def __init__(self, status_code=500, detail=""):
+        self.status_code = status_code
+        self.detail = detail
+
+
+fastapi_mock.FastAPI = _FakeFastAPI
+fastapi_mock.HTTPException = _FakeHTTPException