feat(): embedder implementation

pyproject.toml

@@ -5,11 +5,18 @@ description = "template project"
 version = "1.0.0"
 dependencies = [
     "uuid_utils==0.11.1",
-    "adaptix==3.0.0b11",
     "fastapi==0.119.0",
     "uvicorn==0.37.0",
-    "sqlalchemy==2.0.44",
     "dishka==1.7.2",
+    "pydantic[email]>=2.12.4",
+    "levenshtein>=0.27.3",
+    "markupsafe",
+]
+
+[dependency-groups]
+backend = [
+    "adaptix==3.0.0b11",
+    "sqlalchemy==2.0.44",
     "argon2_cffi==23.1.0",
     "cryptography==46.0.3",
     "httpx==0.28.1",
@@ -18,12 +25,12 @@ dependencies = [
     "aioboto3==15.5.0",
     "prometheus-fastapi-instrumentator>=7.1.0",
     "python-multipart>=0.0.20",
-    "pydantic[email]>=2.12.4",
-    "levenshtein>=0.27.3",
     "pgvector>=0.4.1",
 ]
-
-[dependency-groups]
+ml = [
+    "sentence-transformers>=5.1.2",
+    "torch",
+]
 types = [
     "types-cachetools==6.2.0.20250827",
 ]
@@ -62,6 +69,15 @@ filterwarnings = [
 asyncio_default_test_loop_scope = "session"
 asyncio_default_fixture_loop_scope = "session"
 
+[[tool.uv.index]]
+name = "pytorch"
+url = "https://download.pytorch.org/whl/cpu"
+default = false
+
+[tool.uv.sources]
+torch = { index = "pytorch" }
+markupsafe = { git = "https://github.com/pallets/markupsafe", rev = "3.0.2" }
+
 [tool.mypy]
 strict = true
 strict_bytes = true

src/template_project/adapters/embedding/minilm_embedder.py

@@ -0,0 +1,15 @@
+from typing import cast, override
+
+from sentence_transformers import SentenceTransformer
+
+from template_project.application.common.embedding import Embedder
+
+
+class MiniLMEmbedder(Embedder):
+    def __init__(self, model: SentenceTransformer) -> None:
+        self._model = model
+
+    @override
+    async def encode(self, text: str) -> list[float]:
+        embedding = self._model.encode(text)
+        return cast(list[float], embedding.tolist())

src/template_project/application/common/embedding.py

@@ -0,0 +1,8 @@
+from abc import abstractmethod
+from typing import Protocol
+
+
+class Embedder(Protocol):
+    @abstractmethod
+    async def encode(self, text: str) -> list[float]:
+        raise NotImplementedError