Fix long ingestion and Qdrant compatibility

backend/app/ingest.py

@@ -154,7 +154,8 @@ async def ingest_library(library_id: str, name: str, description: Optional[str]
         # Embed chunks and prepare for storage
         print(f"    Chunked into {len(chunks)} pieces (approx. {num_tokens} tokens)")
         
-        embeddings = embed_texts(chunks)
+        # FastEmbed is synchronous and CPU-heavy; keep it off the API event loop.
+        embeddings = await asyncio.to_thread(embed_texts, chunks)
         
         # Build chunk dicts
         chunk_dicts = []

backend/requirements.txt

@@ -6,7 +6,7 @@ python-dotenv==1.0.0
 python-multipart==0.0.6
 
 # Qdrant Vector Store Client
-qdrant-client==1.7.0
+qdrant-client==1.18.0
 
 # Text Processing for token estimation
 tiktoken==0.7.0