refactor: centralize embedding configuration in Settings

- Add openai_embedding_model and local_embedding_model to Settings
- Update LlamaIndexRAGService to use settings.openai_embedding_model
- Update EmbeddingService to use settings.local_embedding_model
- Both services now read from centralized config instead of hardcoding
- Constraint (OpenAI-only for cloud embeddings) is now explicit in config

src/services/embeddings.py

@@ -11,18 +11,24 @@ import chromadb
 import structlog
 from sentence_transformers import SentenceTransformer
 
+from src.utils.config import settings
 from src.utils.models import Evidence
 
 
 class EmbeddingService:
-    """Handles text embedding and vector storage.
+    """Handles text embedding and vector storage using local sentence-transformers.
 
     All embedding operations run in a thread pool to avoid blocking
     the async event loop.
+
+    Note:
+        Uses local sentence-transformers models (no API key required).
+        Model is configured via settings.local_embedding_model.
     """
 
-    def __init__(self, model_name: str = "all-MiniLM-L6-v2"):
-        self._model = SentenceTransformer(model_name)
+    def __init__(self, model_name: str | None = None):
+        self._model_name = model_name or settings.local_embedding_model
+        self._model = SentenceTransformer(self._model_name)
         self._client = chromadb.Client()  # In-memory for hackathon
         self._collection = self._client.create_collection(
             name="evidence", metadata={"hnsw:space": "cosine"}

src/services/llamaindex_rag.py

@@ -27,7 +27,7 @@ class LlamaIndexRAGService:
         self,
         collection_name: str = "deepcritical_evidence",
         persist_dir: str | None = None,
-        embedding_model: str = "text-embedding-3-small",
+        embedding_model: str | None = None,
         similarity_top_k: int = 5,
     ) -> None:
         """
@@ -36,7 +36,7 @@ class LlamaIndexRAGService:
         Args:
             collection_name: Name of the ChromaDB collection
             persist_dir: Directory to persist ChromaDB data
-            embedding_model: OpenAI embedding model to use
+            embedding_model: OpenAI embedding model (defaults to settings.openai_embedding_model)
             similarity_top_k: Number of top results to retrieve
         """
         # Lazy import - only when instantiated
@@ -64,6 +64,7 @@ class LlamaIndexRAGService:
         self.collection_name = collection_name
         self.persist_dir = persist_dir or settings.chroma_db_path
         self.similarity_top_k = similarity_top_k
+        self.embedding_model = embedding_model or settings.openai_embedding_model
 
         # Validate API key before use
         if not settings.openai_api_key:
@@ -75,7 +76,7 @@ class LlamaIndexRAGService:
             api_key=settings.openai_api_key,
         )
         self._Settings.embed_model = OpenAIEmbedding(
-            model=embedding_model,
+            model=self.embedding_model,
             api_key=settings.openai_api_key,
         )
 

src/utils/config.py

@@ -29,6 +29,17 @@ class Settings(BaseSettings):
     openai_model: str = Field(default="gpt-4o", description="OpenAI model name")
     anthropic_model: str = Field(default="claude-sonnet-4-20250514", description="Anthropic model")
 
+    # Embedding Configuration
+    # Note: OpenAI embeddings require OPENAI_API_KEY (Anthropic has no embeddings API)
+    openai_embedding_model: str = Field(
+        default="text-embedding-3-small",
+        description="OpenAI embedding model (used by LlamaIndex RAG)",
+    )
+    local_embedding_model: str = Field(
+        default="all-MiniLM-L6-v2",
+        description="Local sentence-transformers model (used by EmbeddingService)",
+    )
+
     # PubMed Configuration
     ncbi_api_key: str | None = Field(
         default=None, description="NCBI API key for higher rate limits"