made local rag needs optional (prevents having to have torch / sentence-transformers if you never use local embedding)

src/rag_engine.py

@@ -16,6 +16,7 @@ from src.file_cache import ASTParser
 _SENTENCE_TRANSFORMERS = None
 _GOOGLE_GENAI          = None
 _CHROMADB              = None
+LOCAL_RAG_INSTALL_HINT = "Local RAG embeddings require sentence-transformers. Install with manual_slop[local-rag] to use local embeddings."
 
 
 def _get_sentence_transformers():
@@ -24,6 +25,10 @@ def _get_sentence_transformers():
   try:
    from sentence_transformers import SentenceTransformer
    _SENTENCE_TRANSFORMERS = SentenceTransformer
+  except ModuleNotFoundError as e:
+   if e.name == "sentence_transformers":
+    raise ImportError(LOCAL_RAG_INSTALL_HINT) from e
+   raise
   except Exception as e:
    sys.stderr.write(f"FAILED to import sentence_transformers: {e}\n")
    sys.stderr.flush()
@@ -52,22 +57,12 @@ class BaseEmbeddingProvider:
 
 class LocalEmbeddingProvider(BaseEmbeddingProvider):
  def __init__(self, model_name: str = 'all-MiniLM-L6-v2'):
-  self.model = None
-  try:
-   ST = _get_sentence_transformers()
-   if ST:
-    self.model = ST(model_name)
-  except Exception as e:
-   sys.stderr.write(f"LocalEmbeddingProvider failed to load model {model_name}: {e}. Using dummy embeddings.\n")
-   sys.stderr.flush()
+  ST = _get_sentence_transformers()
+  self.model = ST(model_name)
 
  def embed(self, texts: List[str]) -> List[List[float]]:
-  if self.model:
-   embeddings = self.model.encode(texts)
-   return embeddings.tolist()
-  else:
-   # Dummy embeddings (384 dims for all-MiniLM-L6-v2)
-   return [[0.0] * 384 for _ in texts]
+  embeddings = self.model.encode(texts)
+  return embeddings.tolist()
 
 class GeminiEmbeddingProvider(BaseEmbeddingProvider):
  def __init__(self, model_name: str = 'gemini-embedding-001'):