Buckets:

download
raw
4.53 kB
#!/usr/bin/env python3
"""
Nancy Brain MCP Server Launcher
This script provides an easy way to launch the Nancy Brain MCP server
with the correct configuration and paths.
Prerequisites:
conda run -n roman-slack-bot pip install -e .
Usage:
python run_mcp_server.py
"""
import os
# Fix OpenMP issue before importing any ML libraries
os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"
import sys
import asyncio
from pathlib import Path
# Add the current directory to path so we can import our modules
sys.path.insert(0, str(Path(__file__).parent))
async def main():
"""Launch the Nancy Brain MCP server with default configuration."""
port = int(os.environ.get("MCP_PORT", "8000"))
# Default paths (adjust these to match your setup)
base_path = Path(__file__).parent
config_path = base_path / "config" / "repositories.yml"
embeddings_path = base_path / "knowledge_base" / "embeddings"
weights_path = base_path / "config" / "index_weights.yaml"
# Check if paths exist
missing_paths = []
if not config_path.exists():
missing_paths.append(f"Config file: {config_path}")
if not embeddings_path.exists():
missing_paths.append(f"Embeddings directory: {embeddings_path}")
if weights_path and not weights_path.exists():
print(f"❌ Weights file not found: {weights_path}")
sys.exit(1)
if missing_paths:
print("❌ Missing required files:")
for path in missing_paths:
print(f" {path}")
print("\nPlease ensure you have:")
print("1. Built the knowledge base with embeddings")
print("2. Created the repositories.yml config file")
sys.exit(1)
# Validate that the weights file is NOT a model weights file (should not contain per-document weights)
import yaml
try:
with open(weights_path, "r") as f:
data = yaml.safe_load(f) or {}
forbidden_keys = {"model_weights", "doc_weights", "documents"}
if any(k in data for k in forbidden_keys):
print(
f"❌ ERROR: The weights file '{weights_path}' appears to be a model "
f"weights file (contains {forbidden_keys}). Please provide an "
"index_weights.yaml file for extension/path weights only."
)
sys.exit(1)
except Exception as e:
print(f"❌ Failed to validate weights file: {e}")
sys.exit(1)
print("🚀 Starting Nancy Brain MCP Server...")
print(f"📂 Config: {config_path}")
print(f"🔍 Embeddings: {embeddings_path}")
if weights_path:
print(f"⚖️ Weights: {weights_path}")
print("\n🔗 MCP Server ready for connections via stdio and/or HTTP")
print("💡 Connect this server to Claude Desktop, VS Code, or other MCP clients")
print("\n📋 Available tools:")
print(" • search_knowledge_base - Search Nancy's knowledge base")
print(" • retrieve_document_passage - Get specific document sections")
print(" • retrieve_multiple_passages - Batch retrieve documents")
print(" • explore_document_tree - Browse the document structure")
print(" • set_retrieval_weights - Adjust search priorities")
print(" • get_system_status - Check server health and version")
print("\n" + "=" * 60)
# Import here to avoid early crashes with txtai/torch
import subprocess
import sys
# Always use connectors/mcp_server/server.py as the real entrypoint for HTTP/stdio
server_script = str(Path(__file__).parent / "connectors/mcp_server/server.py")
transport = os.environ.get("MCP_TRANSPORT", "http-and-stdio").strip().lower()
args = [
sys.executable,
server_script,
str(config_path),
str(embeddings_path),
"--weights",
str(weights_path),
"--port",
str(port),
]
# stdio is the default mode for server.py, so we only need to pass flags
# for non-default transports.
if transport == "http":
args.append("--http")
elif transport == "http-and-stdio":
args.append("--http-and-stdio")
# For "stdio" or any other value, rely on the server's default stdio mode.
try:
proc = subprocess.Popen(args)
proc.wait()
except KeyboardInterrupt:
print("\n👋 Nancy Brain MCP Server shutting down...")
except Exception as e:
print(f"❌ Server error: {e}")
sys.exit(1)
if __name__ == "__main__":
asyncio.run(main())

Xet Storage Details

Size:
4.53 kB
·
Xet hash:
0d576aaf44b3891fd69f673ab17ba1ec8f835a93de9e3ebb26652acb8265fc34

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.