diff --git a/src/quartz_api/internal/backends/quartzdb/test_csv_download.py b/src/quartz_api/internal/backends/quartzdb/test_csv_download.py
new file mode 100644
index 0000000..3790ffc
--- /dev/null
+++ b/src/quartz_api/internal/backends/quartzdb/test_csv_download.py
@@ -0,0 +1,147 @@
+"""Test CSV download endpoint."""
+
+import datetime as dt
+import io
+import pathlib
+
+import pandas as pd
+import pytest
+from fastapi.testclient import TestClient
+from pvsite_datamodel.read.model import get_or_create_model
+from pvsite_datamodel.sqlmodels import ForecastSQL, ForecastValueSQL, LocationSQL
+from pyhocon import ConfigFactory
+from sqlalchemy.engine import Engine
+from sqlalchemy.orm import Session
+
+from quartz_api.cmd.main import _create_server
+from quartz_api.internal import models
+
+from .client import Client
+
+
+@pytest.fixture()
+def client(engine: Engine, db_session: Session) -> Client:
+    """Hooks Client into pytest db_session fixture."""
+    client = Client(database_url=str(engine.url))
+    client.session = db_session
+
+    return client
+
+
+@pytest.fixture()
+def test_client(client: Client) -> TestClient:
+    """Create a FastAPI test client."""
+    # Load config and override routers to include 'regions'
+    conf_path = pathlib.Path(__file__).parent.parent.parent.parent / "cmd" / "server.conf"
+    conf = ConfigFactory.parse_file(conf_path.as_posix())
+
+    # Override config to include regions router
+    override_conf = ConfigFactory.from_dict({"api": {"routers": "regions"}})
+    conf = override_conf.with_fallback(conf)
+
+    # Create the FastAPI app
+    app = _create_server(conf=conf)
+
+    # Override the database dependency with our test client
+    app.dependency_overrides[models.get_db_client] = lambda: client
+
+    return TestClient(app)
+
+
+class TestCsvDownload:
+    """Test CSV download functionality."""
+
+    @pytest.mark.usefixtures("forecast_values_wind")
+    def test_csv_download_success(
+        self,
+        test_client: TestClient,
+    ) -> None:
+        """Test successful CSV download from the API endpoint."""
+        # Make request to CSV download endpoint
+        response = test_client.get(
+            "/wind/testID/forecast/csv",
+            params={"forecast_horizon": "latest"},
+        )
+
+        # Assert successful response
+        assert response.status_code == 200
+        assert response.headers["content-type"] == "text/csv; charset=utf-8"
+        assert "Content-Disposition" in response.headers
+        assert "attachment" in response.headers["Content-Disposition"]
+        assert ".csv" in response.headers["Content-Disposition"]
+
+        # Parse CSV content
+        csv_content = response.text
+        df = pd.read_csv(io.StringIO(csv_content))
+
+        # Validate CSV structure
+        assert "Date [IST]" in df.columns
+        assert "Time" in df.columns
+        assert "PowerMW" in df.columns
+        assert len(df) > 0
+
+    @pytest.mark.usefixtures("forecast_values_wind")
+    def test_csv_download_day_ahead(
+        self,
+        test_client: TestClient,
+        client: Client,
+    ) -> None:
+        """Test CSV download with day_ahead forecast horizon."""
+
+        # 1. Setup: Create 'Day Ahead' forecast data (for tomorrow)
+        # Get the wind site
+        site = client.session.query(LocationSQL).filter(LocationSQL.asset_type == "wind").first()
+
+        # Create a forecast timestamped now
+        now = dt.datetime.now(tz=dt.UTC)
+
+        # Create a model entry
+        ml_model = get_or_create_model(client.session, "windnet_india_adjust")
+
+        forecast = ForecastSQL(
+            location_uuid=site.location_uuid,
+            forecast_version="0.0.0",
+            timestamp_utc=now,
+        )
+        client.session.add(forecast)
+        client.session.commit()
+
+        # Add forecast values for "Tomorrow"
+        # We need data that falls into "tomorrow" in IST.
+        # IST is UTC+5:30.
+        # Let's just generate 24 hours of data starting from now + 24h to be safe
+        start_future = now + dt.timedelta(hours=24)
+
+        forecast_values = []
+        for i in range(48):  # 12 hours of data
+            start_utc = start_future + dt.timedelta(minutes=15 * i)
+            end_utc = start_utc + dt.timedelta(minutes=15)
+
+            fv = ForecastValueSQL(
+                forecast_power_kw=123.45 + i,
+                forecast_uuid=forecast.forecast_uuid,
+                start_utc=start_utc,
+                end_utc=end_utc,
+                horizon_minutes=(start_utc - now).total_seconds() / 60,
+            )
+            fv.ml_model = ml_model
+            forecast_values.append(fv)
+
+        client.session.add_all(forecast_values)
+        client.session.commit()
+
+        response = test_client.get(
+            "/wind/testID/forecast/csv",
+            params={"forecast_horizon": "day_ahead"},
+        )
+
+        # The response should be successful even if there's no data for tomorrow
+        assert response.status_code == 200
+        assert response.headers["content-type"] == "text/csv; charset=utf-8"
+
+        # Parse CSV - it may be empty if no forecast data for tomorrow
+        csv_content = response.text
+        if csv_content.strip():  # Only validate if there's content
+            df = pd.read_csv(io.StringIO(csv_content))
+            assert len(df.columns) == 3
+            assert list(df.columns) == ["Date [IST]", "Time", "PowerMW"]