GraphRAG Llama Index MCP Server

#!/bin/bash set -e echo "Starting GraphRAG Query Environment..." >&2 # Check if S3_BUCKET_NAME is provided for remote DB download if [ -n "$S3_BUCKET_NAME" ]; then echo "S3_BUCKET_NAME is set ($S3_BUCKET_NAME). Syncing database from S3..." >&2 DB_NAME=${GRAPHRAG_DATABASE:-default} TARGET_DIR="/root/.graphrag/index-vault/${DB_NAME}" S3_PATH="s3://${S3_BUCKET_NAME}/backups/index-vault/${DB_NAME}" echo "Targeting local path: ${TARGET_DIR}" >&2 # Use AWS CLI to sync the database folder from S3 mkdir -p "$TARGET_DIR" aws s3 sync "$S3_PATH" "$TARGET_DIR" --delete if [ $? -eq 0 ]; then echo "✅ Database synced from S3 successfully." >&2 # Register the database file (assumes standard naming convention) DB_FILE="${TARGET_DIR}/${DB_NAME}.duckdb" if [ -f "$DB_FILE" ]; then python graphrag_cli.py register "$DB_NAME" --db-path "$DB_FILE" >&2 fi else echo "❌ Failed to sync from S3. Check IAM permissions or bucket name." >&2 fi else echo "ℹ️ S3_BUCKET_NAME not set. Running with local/mounted database." >&2 fi echo "Starting MCP server..." >&2 # Use exec to replace the bash process with the node process (for signal handling) exec npx tsx mcp_server.ts