diff --git a/neso_solar_consumer/format_forecast.py b/neso_solar_consumer/format_forecast.py
new file mode 100644
index 0000000..f1348a9
--- /dev/null
+++ b/neso_solar_consumer/format_forecast.py
@@ -0,0 +1,54 @@
+import logging
+from datetime import datetime, timezone
+import pandas as pd
+from nowcasting_datamodel.models import ForecastSQL, ForecastValue
+from nowcasting_datamodel.read.read import get_latest_input_data_last_updated, get_location
+from nowcasting_datamodel.read.read_models import get_model
+
+# Configure logging (set to INFO for production; use DEBUG during debugging)
+logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")
+logger = logging.getLogger(__name__)
+
+
+def format_to_forecast_sql(data: pd.DataFrame, model_tag: str, model_version: str, session) -> list:
+    logger.info("Starting format_to_forecast_sql process...")
+
+    # Step 1: Retrieve model metadata
+    model = get_model(name=model_tag, version=model_version, session=session)
+    input_data_last_updated = get_latest_input_data_last_updated(session=session)
+
+    # Step 2: Fetch or create the location
+    location = get_location(session=session, gsp_id=0)  # National forecast
+
+    # Step 3: Process all rows into ForecastValue objects
+    forecast_values = []
+    for _, row in data.iterrows():
+        if pd.isnull(row["start_utc"]) or pd.isnull(row["solar_forecast_kw"]):
+            logger.warning(f"Skipping row due to missing data: {row}")
+            continue
+
+        try:
+            target_time = datetime.fromisoformat(row["start_utc"]).replace(tzinfo=timezone.utc)
+        except ValueError:
+            logger.warning(f"Invalid datetime format: {row['start_utc']}. Skipping row.")
+            continue
+
+        forecast_value = ForecastValue(
+            target_time=target_time,
+            expected_power_generation_megawatts=row["solar_forecast_kw"] / 1000,  # Convert to MW
+        ).to_orm()
+        forecast_values.append(forecast_value)
+
+    # Step 4: Create a single ForecastSQL object
+    forecast = ForecastSQL(
+        model=model,
+        forecast_creation_time=datetime.now(tz=timezone.utc),
+        location=location,
+        input_data_last_updated=input_data_last_updated,
+        forecast_values=forecast_values,
+        historic=False,
+    )
+    logger.info(f"Created ForecastSQL object with {len(forecast_values)} values.")
+
+    # Return a single ForecastSQL object in a list
+    return [forecast]
diff --git a/tests/conftest.py b/tests/conftest.py
new file mode 100644
index 0000000..1ba90d1
--- /dev/null
+++ b/tests/conftest.py
@@ -0,0 +1,84 @@
+import pytest
+from typing import Generator
+from sqlalchemy import create_engine
+from sqlalchemy.orm import sessionmaker
+from nowcasting_datamodel.models.base import Base_Forecast
+from nowcasting_datamodel.models import MLModelSQL
+from testcontainers.postgres import PostgresContainer
+
+# Shared Test Configuration Constants
+RESOURCE_ID = "db6c038f-98af-4570-ab60-24d71ebd0ae5"
+LIMIT = 5
+COLUMNS = ["DATE_GMT", "TIME_GMT", "EMBEDDED_SOLAR_FORECAST"]
+RENAME_COLUMNS = {
+    "DATE_GMT": "start_utc",
+    "TIME_GMT": "end_utc",
+    "EMBEDDED_SOLAR_FORECAST": "solar_forecast_kw",
+}
+MODEL_NAME = "real_data_model"
+MODEL_VERSION = "1.0"
+
+
+@pytest.fixture(scope="session")
+def postgres_container():
+    """
+    Fixture to spin up a PostgreSQL container for the entire test session.
+    This fixture uses `testcontainers` to start a fresh PostgreSQL container and provides
+    the connection URL dynamically for use in other fixtures.
+    """
+    with PostgresContainer("postgres:15.5") as postgres:
+        postgres.start()
+        yield postgres.get_connection_url()
+
+
+@pytest.fixture(scope="function")
+def db_session(postgres_container) -> Generator:
+    """
+    Fixture to set up and tear down a PostgreSQL database session for testing.
+    This fixture:
+    - Connects to the PostgreSQL container provided by `postgres_container`.
+    - Creates a fresh database schema before each test.
+    - Adds a dummy ML model for test purposes.
+    - Tears down the database session and cleans up resources after each test.
+    Args:
+        postgres_container (str): The dynamic connection URL provided by PostgresContainer.
+    Returns:
+        Generator: A SQLAlchemy session object.
+    """
+    # Use the dynamic connection URL
+    engine = create_engine(postgres_container)
+    Base_Forecast.metadata.drop_all(engine)  # Drop all tables to ensure a clean slate
+    Base_Forecast.metadata.create_all(engine)  # Recreate the tables
+
+    # Establish session
+    Session = sessionmaker(bind=engine)
+    session = Session()
+
+    # Insert a dummy model for testing
+    session.query(MLModelSQL).delete()  # Clean up any pre-existing data
+    model = MLModelSQL(name=MODEL_NAME, version=MODEL_VERSION)
+    session.add(model)
+    session.commit()
+
+    yield session  # Provide the session to the test
+
+    # Cleanup: close session and dispose of engine
+    session.close()
+    engine.dispose()
+
+
+@pytest.fixture(scope="session")
+def test_config():
+    """
+    Fixture to provide shared test configuration constants.
+    Returns:
+        dict: A dictionary of test configuration values.
+    """
+    return {
+        "resource_id": RESOURCE_ID,
+        "limit": LIMIT,
+        "columns": COLUMNS,
+        "rename_columns": RENAME_COLUMNS,
+        "model_name": MODEL_NAME,
+        "model_version": MODEL_VERSION,
+    }
\ No newline at end of file
diff --git a/tests/test_format_forecast.py b/tests/test_format_forecast.py
new file mode 100644
index 0000000..1a55b99
--- /dev/null
+++ b/tests/test_format_forecast.py
@@ -0,0 +1,36 @@
+from neso_solar_consumer.fetch_data import fetch_data
+from neso_solar_consumer.format_forecast import format_to_forecast_sql
+from nowcasting_datamodel.models import ForecastSQL, ForecastValue
+
+
+def test_format_to_forecast_sql_real(db_session, test_config):
+    """
+    Test `format_to_forecast_sql` with real data fetched from the NESO API.
+    """
+    # Step 1: Fetch data from the API
+    data = fetch_data(
+        test_config["resource_id"],
+        test_config["limit"],
+        test_config["columns"],
+        test_config["rename_columns"],
+    )
+    assert not data.empty, "fetch_data returned an empty DataFrame!"
+
+    # Step 2: Format the data into a ForecastSQL object
+    forecasts = format_to_forecast_sql(
+        data, test_config["model_name"], test_config["model_version"], db_session
+    )
+    assert len(forecasts) == 1, f"Expected 1 ForecastSQL object, got {len(forecasts)}"
+
+    # Step 3: Validate the ForecastSQL content
+    forecast = forecasts[0]
+
+    # Filter out invalid rows from the DataFrame (like your function would)
+    valid_data = data.drop_duplicates(subset=["start_utc", "end_utc", "solar_forecast_kw"])
+    valid_data = valid_data[
+        valid_data["start_utc"].apply(lambda x: isinstance(x, str) and len(x) > 0)  # Valid datetime
+    ]
+
+    assert len(forecast.forecast_values) == len(valid_data), (
+        f"Mismatch in ForecastValue entries! Expected {len(valid_data)} but got {len(forecast.forecast_values)}."
+    )