Initial commit: RAG MCP Server with relationship graph

Features:
- Vector search with Pinecone + Vertex AI embeddings
- Document relationships (link, unlink, related, graph)
- Auto-link with LLM analysis
- Intelligent merge with Gemini

Modular structure:
- clients/: Pinecone, Vertex AI
- tools/: core, relations, stats
- utils/: validation, logging

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

clients/pinecone.py

@@ -0,0 +1,27 @@
+"""Pinecone client singleton."""
+from pinecone import Pinecone
+from config import PINECONE_API_KEY, PINECONE_INDEX_NAME
+from utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+# Pinecone singleton
+_pc_client = None
+_index = None
+
+def get_index():
+    """
+    Get Pinecone index instance (singleton pattern).
+
+    Returns:
+        Pinecone index instance
+    """
+    global _pc_client, _index
+
+    if _index is None:
+        logger.info("Initializing Pinecone client")
+        _pc_client = Pinecone(api_key=PINECONE_API_KEY)
+        _index = _pc_client.Index(PINECONE_INDEX_NAME)
+        logger.info(f"Connected to Pinecone index: {PINECONE_INDEX_NAME}")
+
+    return _index