diff --git a/tools/validation/README.md b/tools/validation/README.md
new file mode 100644
index 00000000000..9ae5493d649
--- /dev/null
+++ b/tools/validation/README.md
@@ -0,0 +1,3 @@
+The idea of this solution is to run the same query in both online and offline modes in OpenMLDB and compare their results to ensure they produce the same outcome. 
+
+Run the test_consistency.sh script in /tmp/ and ensure that the paths for OPENMLDB_BIN_PATH, zk_cluster, and zk_root_path are set correctly.
\ No newline at end of file
diff --git a/tools/validation/data.parquet b/tools/validation/data.parquet
new file mode 100644
index 00000000000..36fd1dfc5ef
Binary files /dev/null and b/tools/validation/data.parquet differ
diff --git a/tools/validation/offline.sql b/tools/validation/offline.sql
new file mode 100644
index 00000000000..3a67e567a05
--- /dev/null
+++ b/tools/validation/offline.sql
@@ -0,0 +1,5 @@
+SET @@execute_mode='offline';
+SET @@sync_job = "true";
+USE demo_db;
+SELECT c1, c2, sum(c3) OVER w1 AS w1_c3_sum FROM demo_table1 WINDOW w1 AS (PARTITION BY demo_table1.c1 ORDER BY demo_table1.c6 ROWS BETWEEN 2 PRECEDING AND CURRENT ROW) INTO outfile "file:///tmp/offline.csv" OPTIONS (header = false, mode = 'overwrite');
+
diff --git a/tools/validation/online.sql b/tools/validation/online.sql
new file mode 100644
index 00000000000..45253df11d6
--- /dev/null
+++ b/tools/validation/online.sql
@@ -0,0 +1,2 @@
+USE demo_db;
+SELECT c1, c2, sum(c3) OVER w1 AS w1_c3_sum FROM demo_table1 WINDOW w1 AS (PARTITION BY demo_table1.c1 ORDER BY demo_table1.c6 ROWS BETWEEN 2 PRECEDING AND CURRENT ROW) CONFIG (execute_mode = 'request', values = ("aaa", 11, 22, 1.2, 1.3, 1635247427000, "2021-05-20"));
diff --git a/tools/validation/prepare_data.sql b/tools/validation/prepare_data.sql
new file mode 100644
index 00000000000..60f93f696ed
--- /dev/null
+++ b/tools/validation/prepare_data.sql
@@ -0,0 +1,12 @@
+CREATE DATABASE demo_db;
+USE demo_db;
+CREATE TABLE demo_table1(c1 string, c2 int, c3 bigint, c4 float, c5 double, c6 timestamp, c7 date);
+SET @@execute_mode='offline';
+SET @@sync_job = "true";
+LOAD DATA INFILE 'file:///tmp/data.parquet' INTO TABLE demo_table1 options(format='parquet', mode='append');
+SET @@execute_mode='offline';
+SELECT c1, c2, sum(c3) OVER w1 AS w1_c3_sum FROM demo_table1 WINDOW w1 AS (PARTITION BY demo_table1.c1 ORDER BY demo_table1.c6 ROWS BETWEEN 2 PRECEDING AND CURRENT ROW) INTO OUTFILE '/tmp/feature_data' OPTIONS(mode='overwrite');
+SET @@execute_mode='online';
+DEPLOY demo_data_service SELECT c1, c2, sum(c3) OVER w1 AS w1_c3_sum FROM demo_table1 WINDOW w1 AS (PARTITION BY demo_table1.c1 ORDER BY demo_table1.c6 ROWS BETWEEN 2 PRECEDING AND CURRENT ROW);
+SET @@execute_mode='online';
+LOAD DATA INFILE 'file:///tmp/data.parquet' INTO TABLE demo_table1 options(format='parquet', header=true, mode='append');
diff --git a/tools/validation/test.py b/tools/validation/test.py
new file mode 100644
index 00000000000..f48b7ce7e94
--- /dev/null
+++ b/tools/validation/test.py
@@ -0,0 +1,59 @@
+import subprocess
+import os
+import csv
+import sys
+
+openmldb_binary_path = sys.argv[1]
+# Execute the offline SQL command
+subprocess.run(f"{openmldb_binary_path} --zk_cluster=127.0.0.1:2181 --zk_root_path=/openmldb --role=sql_client < offline.sql", 
+               shell=True, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
+
+# Define the directory containing the CSV files
+csv_dir = "/tmp/offline.csv/"
+
+# List all files in the directory
+files = os.listdir(csv_dir)
+
+# Filter out non-CSV files (and avoid reading .crc files)
+csv_files = [file for file in files if file.endswith('.csv')]
+
+# Initialize an empty list to store the combined data
+combined_data = []
+
+# Read and concatenate all CSV files
+for file in csv_files:
+    with open(os.path.join(csv_dir, file), newline='') as csvfile:
+        reader = csv.reader(csvfile)
+        for row in reader:
+            combined_data.append(row)
+
+# Define the command to be executed
+command = f"{openmldb_binary_path} --zk_cluster=127.0.0.1:2181 --zk_root_path=/openmldb --role=sql_client < online.sql"
+
+# Execute the command
+try:
+    result = subprocess.run(command, shell=True, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
+    output = result.stdout
+
+    # Extract the relevant line containing the data
+    lines = output.splitlines()
+    for i, line in enumerate(lines):
+        if "c1" in line and "c2" in line and "w1_c3_sum" in line:
+            data_line = lines[i + 2]  # The line containing the data is two lines below the header
+            break
+
+    # Split the line into an array
+    data_array = data_line.split()
+
+    # Check if the specific row exists
+    row_exists = any(row == data_array for row in combined_data)
+    
+    if row_exists:
+        print("Online and offline data consistent")
+    else:
+        print("Online and offline data not consistent")
+
+    # Print the resulting array
+except subprocess.CalledProcessError as e:
+    print("An error occurred while executing the command:", e)
+    print("Error Output:\n", e.stderr)
diff --git a/tools/validation/test_consistency.sh b/tools/validation/test_consistency.sh
new file mode 100644
index 00000000000..632acfaba50
--- /dev/null
+++ b/tools/validation/test_consistency.sh
@@ -0,0 +1,3 @@
+OPENMLDB_BIN_PATH=/work/openmldb/bin/openmldb
+$OPENMLDB_BIN_PATH --zk_cluster=127.0.0.1:2181 --zk_root_path=/openmldb --role=sql_client  < /tmp/prepare_data.sql
+python3 test.py $OPENMLDB_BIN_PATH
\ No newline at end of file