fix(settings): avoid side effects and create data dir at startup

Intrinsical-AI · Intrinsical-AI · commit 650ecef108f7 · 2026-02-15T06:04:50.000+01:00
diff --git a/src/local_rag_backend/app/main.py b/src/local_rag_backend/app/main.py
@@ -27,7 +27,8 @@
     from collections.abc import AsyncGenerator
     from importlib.resources.abc import Traversable
 
-logging.basicConfig(level=logging.INFO, stream=sys.stdout)
+_LOG_LEVEL = getattr(logging, settings.log_level, logging.INFO)
+logging.basicConfig(level=_LOG_LEVEL, stream=sys.stdout)
 logger = logging.getLogger(__name__)
 
 # Frontend directory in the source tree (works for editable installs / running from repo)
@@ -38,6 +39,8 @@
 async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     """Manage application startup and shutdown events."""
     logger.info("Initializing RAG service...")
+    # Ensure the data directory exists (SQLite cannot create parent directories).
+    settings.data_dir.mkdir(parents=True, exist_ok=True)
     # Use the module reference so tests can monkeypatch `db_base.engine` / `db_base.SessionLocal`.
     db_base.Base.metadata.create_all(bind=db_base.engine)
     # Best-effort preload: don't prevent the API from starting just because an LLM
diff --git a/src/local_rag_backend/bootstrap.py b/src/local_rag_backend/bootstrap.py
@@ -5,6 +5,7 @@
 from typing import TYPE_CHECKING
 
 from local_rag_backend.app.factory import build_rag_service
+from local_rag_backend.settings import settings
 
 if TYPE_CHECKING:
     from local_rag_backend.core.services.rag import RagService
@@ -17,6 +18,7 @@ def bootstrap_rag_service() -> RagService:
     Note: this is a non-cached builder. FastAPI uses an internal cached singleton via DI.
     """
 
+    settings.data_dir.mkdir(parents=True, exist_ok=True)
     return build_rag_service()
 
 
diff --git a/src/local_rag_backend/scripts/bootstrap.py b/src/local_rag_backend/scripts/bootstrap.py
@@ -41,6 +41,8 @@ def main(csv_path: str | Path | None = None, **kwargs: Any) -> None:
     if "settings" not in kwargs:
         kwargs["settings"] = default_settings
     settings = kwargs["settings"]
+    # Ensure data dir exists before touching SQLite.
+    settings.data_dir.mkdir(parents=True, exist_ok=True)
     # 1) Create engine and session based on the updated URL
     engine = create_engine(settings.sqlite_url, connect_args={"check_same_thread": False})
     session_local = sessionmaker(bind=engine, autocommit=False, autoflush=False)
diff --git a/src/local_rag_backend/scripts/build_index.py b/src/local_rag_backend/scripts/build_index.py
@@ -44,6 +44,8 @@ def main() -> None:
     Initializes DB and optionally FAISS index using functionality from data_loader.
     """
     logger.info("Starting build_index script...")
+    # Ensure data dir exists before touching SQLite.
+    settings.data_dir.mkdir(parents=True, exist_ok=True)
 
     # 1. Create a dedicated engine and SessionLocal for this script
     #    This avoids conflicts with the main app engine.
diff --git a/src/local_rag_backend/settings.py b/src/local_rag_backend/settings.py
@@ -105,12 +105,19 @@ def __init__(self, **kwargs: Any) -> None: ...
         env_file=".env", env_file_encoding="utf-8", case_sensitive=False, extra="ignore"
     )
 
+    @field_validator("log_level", mode="before")
+    @classmethod
+    def _normalize_log_level(cls, v: Any) -> Any:
+        # Make env/config more forgiving while keeping a strict Literal type.
+        if isinstance(v, str):
+            return v.upper()
+        return v
+
     @field_validator("data_dir", mode="before")
     @classmethod
-    def _ensure_data_dir_exists(cls, v: Any) -> Path:
-        path = Path(v)
-        path.mkdir(parents=True, exist_ok=True)
-        return path
+    def _normalize_data_dir(cls, v: Any) -> Path:
+        # Keep Settings side-effect free; callers are responsible for creating directories.
+        return Path(v)
 
     @field_validator("sqlite_url")
     @classmethod
diff --git a/tests/unit/settings/test_settings_validators.py b/tests/unit/settings/test_settings_validators.py
@@ -18,3 +18,16 @@ def test_chunk_overlap_lt_chars_validator():
     """Ensure overlap must be strictly less than chunk size."""
     with pytest.raises(ValueError):
         Settings(ingest_chunk_chars=100, ingest_chunk_overlap=100)
+
+
+def test_log_level_is_normalized_to_uppercase():
+    s = Settings(log_level="debug")
+    assert s.log_level == "DEBUG"
+
+
+def test_data_dir_is_not_created_as_a_side_effect(tmp_path):
+    d = tmp_path / "new-data-dir"
+    assert not d.exists()
+    s = Settings(data_dir=d)
+    assert s.data_dir == d
+    assert not d.exists()
