Preserve ingestion data across rebuilds

backend/app/embeddings.py

@@ -1,5 +1,6 @@
 # Local Embedding Generation using FastEmbed
 import asyncio
+import os
 from typing import List
 from functools import lru_cache
 
@@ -20,7 +21,11 @@ def _load_model():
             print("Loading embedding model (this may take a few minutes on first run)...")
             
             # Use BAAI/bge-small-en-v1.5 - lightweight (~90MB), works offline
-            _embedding_model = TextEmbedding(model_name="BAAI/bge-small-en-v1.5", cache_dir=".embed_cache")
+            cache_dir = os.getenv("EMBEDDING_CACHE_DIR", ".embed_cache")
+            _embedding_model = TextEmbedding(
+                model_name="BAAI/bge-small-en-v1.5",
+                cache_dir=cache_dir,
+            )
             print("Embedding model loaded successfully.")
         
         return _embedding_model
@@ -178,4 +183,4 @@ if __name__ == "__main__":
     assert embed_texts([]) == [], "Empty list should return empty list"
     print("✓ Empty input handling works")
     
-    print("\n✅ All tests passed!")
+    print("\n✅ All tests passed!")