Make MCP server optional dependency

The fastmcp server code is now an optional dependency that can be
installed with the "mcp" extra. Core vector search functionality is
available without the MCP server dependency.

src/vector_search_mcp/engine/qdrant_engine.py

@@ -16,14 +16,16 @@ from typing import final, override
 from qdrant_client import AsyncQdrantClient, models
 
 from ..config import Settings
-from ..models import Condition, Match, MatchAny, MatchExclude, SearchRow
+from ..models import Chunk, Condition, Match, MatchAny, MatchExclude, SearchRow
 from .base_engine import BaseEngine
 
 __all__ = ["QdrantEngine"]
 
 
 @final
-class QdrantEngine(BaseEngine[list[models.ScoredPoint], models.Filter]):
+class QdrantEngine(
+    BaseEngine[list[models.ScoredPoint], models.Filter, models.PointStruct]
+):
     """Qdrant vector database engine implementation.
 
     This class provides a concrete implementation of the BaseEngine interface
@@ -195,3 +197,30 @@ class QdrantEngine(BaseEngine[list[models.ScoredPoint], models.Filter]):
             with_vectors=False,
             score_threshold=threshold,
         )
+
+    @override
+    async def create_index(self, name: str, size: int) -> bool:
+        return await self.client.create_collection(
+            collection_name=name,
+            vectors_config=models.VectorParams(
+                size=size, distance=models.Distance.COSINE
+            ),
+        )
+
+    @override
+    async def run_upload_chunk(
+        self, index_name: str, chunk: models.PointStruct
+    ) -> bool:
+        result = await self.client.upsert(
+            collection_name=index_name,
+            points=[chunk],
+        )
+        return result.status == models.UpdateStatus.ACKNOWLEDGED
+
+    @override
+    def transform_chunk(self, chunk: Chunk) -> models.PointStruct:
+        return models.PointStruct(
+            id=chunk.id,
+            vector=chunk.vector,
+            payload=chunk.payload.model_dump(),
+        )