diff --git a/fennel/CHANGELOG.md b/fennel/CHANGELOG.md
index 579d5db8..d0f39ef6 100644
--- a/fennel/CHANGELOG.md
+++ b/fennel/CHANGELOG.md
@@ -1,6 +1,8 @@
 # Changelog
+## [1.5.51] - 2024-11-09
+- Fix bug in mock client for session based dedup operator
 
-## [1.5.49] - 2024-11-09
+## [1.5.50] - 2024-11-09
 - Add json_extract and split string expressions
 
 ## [1.5.48] - 2024-11-06
diff --git a/fennel/client_tests/test_dedup_session.py b/fennel/client_tests/test_dedup_session.py
new file mode 100644
index 00000000..7d9bb844
--- /dev/null
+++ b/fennel/client_tests/test_dedup_session.py
@@ -0,0 +1,86 @@
+import time
+from datetime import datetime, timedelta, timezone
+import pandas as pd
+
+from fennel.datasets import (
+    dataset,
+    field,
+    pipeline,
+    Sum,
+    Count,
+)
+from fennel.dtypes import Continuous, Session
+from fennel.lib import inputs
+from fennel.testing import mock, log
+
+__owner__ = "eng@fennel.ai"
+
+
+@mock
+def test_dedup_session(client):
+    @dataset
+    class DatasetA:
+        a: int
+        user_id: str
+        time: datetime
+
+    @dataset(index=True)
+    class AggDataset:
+        user_id: str = field(key=True)
+        count_a: int
+        sum_a: int
+        time: datetime
+
+        @pipeline
+        @inputs(DatasetA)
+        def pipeline(cls, data_a):
+            data_b = data_a.dedup(by=["user_id", "a"], window=Session(gap="1s"))
+            return data_b.groupby("user_id").aggregate(
+                count_a=Count(window=Continuous("forever")),
+                sum_a=Sum(of="a", window=Continuous("forever")),
+            )
+
+    client.commit(
+        message="initial commit",
+        datasets=[DatasetA, AggDataset],
+    )
+
+    day = timedelta(days=1)
+    min = timedelta(minutes=1)
+    now = datetime.now(timezone.utc)
+
+    time.sleep(2)
+    now2 = datetime.now(timezone.utc)
+
+    columns = ["a", "user_id", "time"]
+    data = [
+        [1, "u1", now],
+        [1, "u1", now],
+        [1, "u2", now],
+    ]
+    df = pd.DataFrame(data, columns=columns)
+    log(DatasetA, df)
+
+    data = [
+        [1, "u1", now2 + day],
+        [2, "u1", now2 + day],
+        [0, "u2", now2 + day],
+        [5, "u1", now2 + 2 * day + min],
+    ]
+    df = pd.DataFrame(data, columns=columns)
+    log(DatasetA, df)
+
+    data, found = client.lookup(
+        "AggDataset",
+        timestamps=pd.Series(
+            [now + min, now2 + day + min, now2 + 2 * day + 2 * min]
+        ),
+        keys=pd.DataFrame({"user_id": ["u1", "u1", "u1"]}),
+    )
+    assert data.shape == (3, 4)
+    assert data.iloc[0]["count_a"] == 1
+    assert data.iloc[0]["sum_a"] == 1
+    assert data.iloc[1]["count_a"] == 3
+    assert data.iloc[1]["sum_a"] == 4
+    assert data.iloc[2]["count_a"] == 4
+    assert data.iloc[2]["sum_a"] == 9
diff --git a/fennel/testing/executor.py b/fennel/testing/executor.py
index 6b7c1f73..bfd6e69d 100644
--- a/fennel/testing/executor.py
+++ b/fennel/testing/executor.py
@@ -755,12 +755,19 @@ def visitDedup(self, obj):
             )
             df = df.drop(columns=["__@@__tumbling_id", "__@@__index"])
         elif isinstance(obj.window, Session):
-            df["__@@__index"] = df.index
+            df["__@@__index"] = range(df.shape[0])
             # Using a stable sort algorithm to preserve the order of rows that have the same timestamp within the same session
             df = df.sort_values(input_ret.timestamp_field, kind="mergesort")
-            df["__@@__session_id"] = generate_session_id(
-                df[input_ret.timestamp_field], obj.window
-            )
+            index_to_session_id = {}
+            for _, group in df.groupby(obj.by):
+                session_ids = generate_session_id(
+                    group[input_ret.timestamp_field], obj.window
+                )
+                for i, session_id in enumerate(session_ids):
+                    index_to_session_id[group.iloc[i]["__@@__index"]] = (
+                        session_id
+                    )
+            df["__@@__session_id"] = df["__@@__index"].map(index_to_session_id)
             df = (
                 df.groupby(obj.by + ["__@@__session_id"])
                 .last()
@@ -768,6 +775,7 @@ def visitDedup(self, obj):
                 .sort_values("__@@__index")
             )
             df = df.drop(columns=["__@@__session_id", "__@@__index"])
+            df = df.reset_index(drop=True)
         else:
             raise Exception(f"Unsupported window type {type(obj.window)}")
 
diff --git a/pyproject.toml b/pyproject.toml
index 67a1ebb6..43e56275 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "fennel-ai"
-version = "1.5.50"
+version = "1.5.51"
 description = "The modern realtime feature engineering platform"
 authors = ["Fennel AI <developers@fennel.ai>"]
 packages = [{ include = "fennel" }]