From 02314db5174a3405cad75afb717e532b2fc8cfbf Mon Sep 17 00:00:00 2001
From: "D. Lowl" <d.lowl@d-lowl.space>
Date: Wed, 16 Aug 2023 18:00:29 +0400
Subject: [PATCH 1/8] Update dvc files

---
 data/mind_data_raw.csv.dvc | 4 ++--
 data/nhs_data_raw.csv.dvc  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/data/mind_data_raw.csv.dvc b/data/mind_data_raw.csv.dvc
index 8f79853..3756535 100644
--- a/data/mind_data_raw.csv.dvc
+++ b/data/mind_data_raw.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: dad43927d0712f97a8c261d2ab8dba81
-  size: 10548104
+- md5: 3bdc23a1738ddf268a78cbe8f6a026e7
+  size: 10562463
   hash: md5
   path: mind_data_raw.csv
diff --git a/data/nhs_data_raw.csv.dvc b/data/nhs_data_raw.csv.dvc
index 480b135..c8f978c 100644
--- a/data/nhs_data_raw.csv.dvc
+++ b/data/nhs_data_raw.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: 7e645f219661c3dd173669d284383bbe
-  size: 2688031
+- md5: 046f4080f9d8c2e2abcfa563c41d95b6
+  size: 2687654
   hash: md5
   path: nhs_data_raw.csv

From 75d622a96d846a3aba016ccd63851f4774abe7f6 Mon Sep 17 00:00:00 2001
From: "D. Lowl" <d.lowl@d-lowl.space>
Date: Wed, 16 Aug 2023 18:00:44 +0400
Subject: [PATCH 2/8] Update dvc files

---
 data/mind_data_validated.csv.dvc | 2 +-
 data/nhs_data_validated.csv.dvc  | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/data/mind_data_validated.csv.dvc b/data/mind_data_validated.csv.dvc
index 1d047a1..ed731bd 100644
--- a/data/mind_data_validated.csv.dvc
+++ b/data/mind_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: 027332cb6e506fd47d3962471c1246d2
+- md5: de6fed15842b8661c864c01a253d04a2
   size: 1859742
   hash: md5
   path: mind_data_validated.csv
diff --git a/data/nhs_data_validated.csv.dvc b/data/nhs_data_validated.csv.dvc
index ad9a96e..af7623e 100644
--- a/data/nhs_data_validated.csv.dvc
+++ b/data/nhs_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: 1534192fdce6f9c7bde650cf0374629c
+- md5: 1d5ebb8903a5470374ae9271ddbd9f59
   size: 797289
   hash: md5
   path: nhs_data_validated.csv

From 40cc786a9752f8928b86073f512a04ca0b5fe5c0 Mon Sep 17 00:00:00 2001
From: "D. Lowl" <d.lowl@d-lowl.space>
Date: Thu, 17 Aug 2023 17:52:53 +0400
Subject: [PATCH 3/8] Update dvc files

---
 data/mind_data_validated.csv.dvc | 4 ++--
 data/nhs_data_validated.csv.dvc  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/data/mind_data_validated.csv.dvc b/data/mind_data_validated.csv.dvc
index ed731bd..94f6330 100644
--- a/data/mind_data_validated.csv.dvc
+++ b/data/mind_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: de6fed15842b8661c864c01a253d04a2
-  size: 1859742
+- md5: c76568f0dd939286c633b2ed3a557630
+  size: 2336306
   hash: md5
   path: mind_data_validated.csv
diff --git a/data/nhs_data_validated.csv.dvc b/data/nhs_data_validated.csv.dvc
index af7623e..5a52378 100644
--- a/data/nhs_data_validated.csv.dvc
+++ b/data/nhs_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: 1d5ebb8903a5470374ae9271ddbd9f59
-  size: 797289
+- md5: c4b3c09b6499a4faeec99fd7dbc27d4f
+  size: 811904
   hash: md5
   path: nhs_data_validated.csv

From 023a2cbe37fb2f1afce420823ea153fe3d4459b9 Mon Sep 17 00:00:00 2001
From: "D. Lowl" <d.lowl@d-lowl.space>
Date: Thu, 17 Aug 2023 17:55:08 +0400
Subject: [PATCH 4/8] Update dvc files

---
 data/mind_data_validated.csv.dvc | 4 ++--
 data/nhs_data_validated.csv.dvc  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/data/mind_data_validated.csv.dvc b/data/mind_data_validated.csv.dvc
index 94f6330..c51ea00 100644
--- a/data/mind_data_validated.csv.dvc
+++ b/data/mind_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: c76568f0dd939286c633b2ed3a557630
-  size: 2336306
+- md5: 45a2629d51c001b2556f2586966ef80b
+  size: 6778
   hash: md5
   path: mind_data_validated.csv
diff --git a/data/nhs_data_validated.csv.dvc b/data/nhs_data_validated.csv.dvc
index 5a52378..ebf7bff 100644
--- a/data/nhs_data_validated.csv.dvc
+++ b/data/nhs_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: c4b3c09b6499a4faeec99fd7dbc27d4f
-  size: 811904
+- md5: ed6e0add96a44532f0a0ac73448d1382
+  size: 148389
   hash: md5
   path: nhs_data_validated.csv

From 2e2626a44e699071a6910e6f6e9729276e6760eb Mon Sep 17 00:00:00 2001
From: "D. Lowl" <d.lowl@d-lowl.space>
Date: Thu, 17 Aug 2023 18:08:03 +0400
Subject: [PATCH 5/8] Update dvc files

---
 data/mind_data_validated.csv.dvc | 4 ++--
 data/nhs_data_validated.csv.dvc  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/data/mind_data_validated.csv.dvc b/data/mind_data_validated.csv.dvc
index c51ea00..94f6330 100644
--- a/data/mind_data_validated.csv.dvc
+++ b/data/mind_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: 45a2629d51c001b2556f2586966ef80b
-  size: 6778
+- md5: c76568f0dd939286c633b2ed3a557630
+  size: 2336306
   hash: md5
   path: mind_data_validated.csv
diff --git a/data/nhs_data_validated.csv.dvc b/data/nhs_data_validated.csv.dvc
index ebf7bff..5a52378 100644
--- a/data/nhs_data_validated.csv.dvc
+++ b/data/nhs_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: ed6e0add96a44532f0a0ac73448d1382
-  size: 148389
+- md5: c4b3c09b6499a4faeec99fd7dbc27d4f
+  size: 811904
   hash: md5
   path: nhs_data_validated.csv

From 198ffc0a5348912d6bcca84b1e81a3da3891997f Mon Sep 17 00:00:00 2001
From: "D. Lowl" <d.lowl@d-lowl.space>
Date: Thu, 17 Aug 2023 18:38:28 +0400
Subject: [PATCH 6/8] Update dvc files

---
 data/mind_data_validated.csv.dvc | 4 ++--
 data/nhs_data_validated.csv.dvc  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/data/mind_data_validated.csv.dvc b/data/mind_data_validated.csv.dvc
index 94f6330..f7cfda0 100644
--- a/data/mind_data_validated.csv.dvc
+++ b/data/mind_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: c76568f0dd939286c633b2ed3a557630
-  size: 2336306
+- md5: 3a7b9eb1525ae9021318306d874adfbe
+  size: 2541363
   hash: md5
   path: mind_data_validated.csv
diff --git a/data/nhs_data_validated.csv.dvc b/data/nhs_data_validated.csv.dvc
index 5a52378..ae53e4f 100644
--- a/data/nhs_data_validated.csv.dvc
+++ b/data/nhs_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: c4b3c09b6499a4faeec99fd7dbc27d4f
-  size: 811904
+- md5: 89494c5ba3ee949b09244f50f121fda1
+  size: 933121
   hash: md5
   path: nhs_data_validated.csv

From 9592b0966abda00b265319db45a0984495d70813 Mon Sep 17 00:00:00 2001
From: "D. Lowl" <d.lowl@d-lowl.space>
Date: Thu, 17 Aug 2023 18:40:00 +0400
Subject: [PATCH 7/8] Update dvc files

---
 data/mind_data_validated.csv.dvc | 4 ++--
 data/nhs_data_validated.csv.dvc  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/data/mind_data_validated.csv.dvc b/data/mind_data_validated.csv.dvc
index f7cfda0..94f6330 100644
--- a/data/mind_data_validated.csv.dvc
+++ b/data/mind_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: 3a7b9eb1525ae9021318306d874adfbe
-  size: 2541363
+- md5: c76568f0dd939286c633b2ed3a557630
+  size: 2336306
   hash: md5
   path: mind_data_validated.csv
diff --git a/data/nhs_data_validated.csv.dvc b/data/nhs_data_validated.csv.dvc
index ae53e4f..5a52378 100644
--- a/data/nhs_data_validated.csv.dvc
+++ b/data/nhs_data_validated.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: 89494c5ba3ee949b09244f50f121fda1
-  size: 933121
+- md5: c4b3c09b6499a4faeec99fd7dbc27d4f
+  size: 811904
   hash: md5
   path: nhs_data_validated.csv

From f68a52d1aa34f284735e1840c996e710cc642e10 Mon Sep 17 00:00:00 2001
From: "D. Lowl" <d.lowl@d-lowl.space>
Date: Thu, 17 Aug 2023 20:15:48 +0400
Subject: [PATCH 8/8] Add intelligent chunking experiment

---
 app/app.py                                    |  18 +-
 data/mind_data_raw.csv.dvc                    |   4 +-
 data/nhs_data_raw.csv.dvc                     |   4 +-
 .../config_data_preparation_pipeline.yaml     |   8 +-
 .../data_preparation_pipeline.py              |  16 +-
 .../compute_embedding_drift_step.py           |  10 +-
 .../clean_data_step/clean_data_step.py        |   2 +-
 .../split_nhs_pages_step.py                   | 110 ++++++++++++
 .../test_split_nhs_pages_step.py              | 161 ++++++++++++++++++
 9 files changed, 306 insertions(+), 27 deletions(-)
 create mode 100644 steps/data_preparation_steps/split_nhs_pages_step/split_nhs_pages_step.py
 create mode 100644 tests/test_steps/test_data_preparation_steps/test_split_nhs_pages_step.py

diff --git a/app/app.py b/app/app.py
index 5afb025..d7e7375 100644
--- a/app/app.py
+++ b/app/app.py
@@ -27,7 +27,7 @@
 
 
 # Setup for chroma vector store
-CHROMA_SERVER_HOST_NAME = "chroma-service.default"
+CHROMA_SERVER_HOST_NAME = "localhost"
 CHROMA_SERVER_PORT = 8000
 DEFAULT_EMBED_MODEL = "base"  # ["base", "large", "xl"]
 N_CLOSEST_MATCHES = 3
@@ -99,7 +99,7 @@ def _get_prediction_endpoint() -> Optional[str]:
     Returns:
         Optional[str]: the url endpoint if it exists and is valid, None otherwise.
     """
-    return f"http://{SELDON_SERVICE_NAME}.{SELDON_NAMESPACE}:{SELDON_PORT}/v2/models/transformer/infer"
+    return f"http://localhost:9000/v2/models/transformer/infer"
 
 
 @st.cache_data(show_spinner=False)
@@ -307,7 +307,7 @@ def show_disclaimer() -> None:
 def show_settings() -> None:
     """Show inference settings on the sidebar."""
     st.title("Settings")
-    st.session_state.temperature = st.slider("Temperature", min_value=0.0, max_value=2.0, value=0.8)
+    st.session_state.temperature = st.slider("Temperature", min_value=0.0, max_value=2.0, value=0.0)
     st.session_state.max_length = st.slider("Max response length", min_value=50, max_value=500, value=300, step=1)
     st.session_state.prompt_template = st.select_slider("Prompt template", options=["simple", "complex", "advanced"], value="simple")
 
@@ -383,6 +383,8 @@ def main() -> None:
                             embedding_function=embed_function,
                         )
 
+                        print(context)
+
                         # Create a dict of prompt and context
                         message = {"prompt_query": prompt, "context": context}
 
@@ -396,11 +398,11 @@ def main() -> None:
 
                         full_response += f"{source}: {assistant_response}  \n"
 
-                        if metric_service_endpoint:
-                            result = post_response_to_metric_service(
-                                metric_service_endpoint, assistant_response
-                            )
-                            logging.info(result.text)
+                        # if metric_service_endpoint:
+                        #     result = post_response_to_metric_service(
+                        #         metric_service_endpoint, assistant_response
+                        #     )
+                        #     logging.info(result.text)
 
                     message_placeholder.markdown(full_response)
 
diff --git a/data/mind_data_raw.csv.dvc b/data/mind_data_raw.csv.dvc
index 3756535..8f79853 100644
--- a/data/mind_data_raw.csv.dvc
+++ b/data/mind_data_raw.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: 3bdc23a1738ddf268a78cbe8f6a026e7
-  size: 10562463
+- md5: dad43927d0712f97a8c261d2ab8dba81
+  size: 10548104
   hash: md5
   path: mind_data_raw.csv
diff --git a/data/nhs_data_raw.csv.dvc b/data/nhs_data_raw.csv.dvc
index c8f978c..480b135 100644
--- a/data/nhs_data_raw.csv.dvc
+++ b/data/nhs_data_raw.csv.dvc
@@ -1,5 +1,5 @@
 outs:
-- md5: 046f4080f9d8c2e2abcfa563c41d95b6
-  size: 2687654
+- md5: 7e645f219661c3dd173669d284383bbe
+  size: 2688031
   hash: md5
   path: nhs_data_raw.csv
diff --git a/pipelines/data_preparation_pipeline/config_data_preparation_pipeline.yaml b/pipelines/data_preparation_pipeline/config_data_preparation_pipeline.yaml
index ecf18b1..056ff89 100644
--- a/pipelines/data_preparation_pipeline/config_data_preparation_pipeline.yaml
+++ b/pipelines/data_preparation_pipeline/config_data_preparation_pipeline.yaml
@@ -5,9 +5,13 @@ steps:
       data_version: "data/second_version"
       data_postfix: "raw"
       reference_data_version: "data/second_version"
-  clean_data:
+  clean_data_mind:
     enable_cache: False
-  validate_data:
+  clean_data_nhs:
+    enable_cache: False
+  validate_data_mind:
+    enable_cache: False
+  validate_data_nhs:
     enable_cache: False
   version_data:
     enable_cache: False
diff --git a/pipelines/data_preparation_pipeline/data_preparation_pipeline.py b/pipelines/data_preparation_pipeline/data_preparation_pipeline.py
index acd0a1d..942a9a7 100644
--- a/pipelines/data_preparation_pipeline/data_preparation_pipeline.py
+++ b/pipelines/data_preparation_pipeline/data_preparation_pipeline.py
@@ -1,5 +1,6 @@
 """Data preparation pipeline."""
 from steps.data_preparation_steps import clean_data, validate_data
+from steps.data_preparation_steps.split_nhs_pages_step.split_nhs_pages_step import split_pages
 from steps.data_versioning_steps import version_data
 from steps.generic_steps import load_data
 from zenml import pipeline
@@ -20,12 +21,13 @@ def data_preparation_pipeline() -> None:
     """
     _, _, mind_df, nhs_df = load_data()
 
-    mind_df = clean_data(mind_df)
-    nhs_df = clean_data(nhs_df)
+    mind_df = split_pages(mind_df, "mind", after=["load_data"], id="split_pages_mind")
+    nhs_df = split_pages(nhs_df, "nhs", after=["load_data"], id="split_pages_nhs")
 
-    mind_df = validate_data(mind_df, "mind")
-    nhs_df = validate_data(nhs_df, "nhs")
+    mind_df = clean_data(mind_df, after=["split_pages_mind"], id="clean_data_mind")
+    nhs_df = clean_data(nhs_df, after=["split_pages_nhs"], id="clean_data_nhs")
 
-    version_data(
-        after=["load_data", "clean_data", "validate_data"], data_postfix="validated"
-    )
+    mind_df = validate_data(mind_df, "mind", after=["clean_data_mind"], id="validate_data_mind")
+    nhs_df = validate_data(nhs_df, "nhs", after=["clean_data_nhs"], id="validate_data_nhs")
+
+    version_data(after=["validate_data_mind", "validate_data_nhs"], data_postfix="validated")
diff --git a/steps/data_embedding_steps/compute_embedding_drift_step/compute_embedding_drift_step.py b/steps/data_embedding_steps/compute_embedding_drift_step/compute_embedding_drift_step.py
index 059f028..6ea93c9 100644
--- a/steps/data_embedding_steps/compute_embedding_drift_step/compute_embedding_drift_step.py
+++ b/steps/data_embedding_steps/compute_embedding_drift_step/compute_embedding_drift_step.py
@@ -149,11 +149,11 @@ def compute_embedding_drift(
     payload = build_embedding_drift_payload(
         reference_data_version, current_data_version, distance
     )
-    response = requests.post(
-        f"http://{MONITORING_METRICS_HOST_NAME}:{MONITORING_METRICS_PORT}/embedding_drift",
-        json=payload,
-    )
+    # response = requests.post(
+    #     f"http://{MONITORING_METRICS_HOST_NAME}:{MONITORING_METRICS_PORT}/embedding_drift",
+    #     json=payload,
+    # )
 
-    logger.info(response.text)
+    # logger.info(response.text)
 
     return float(distance)
diff --git a/steps/data_preparation_steps/clean_data_step/clean_data_step.py b/steps/data_preparation_steps/clean_data_step/clean_data_step.py
index 4b23e7b..8bcd790 100644
--- a/steps/data_preparation_steps/clean_data_step/clean_data_step.py
+++ b/steps/data_preparation_steps/clean_data_step/clean_data_step.py
@@ -267,7 +267,7 @@ def clean_data(data: pd.DataFrame) -> pd.DataFrame:
     data["text_scraped"] = data["text_scraped"].map(lambda text: remove_pattern(r"Or see our page[^\.]+[\.]", text))
     data["text_scraped"] = data["text_scraped"].map(lambda text: remove_pattern(r"Read more about[^\.]+[\.]", text))
 
-    data = data.drop(data[data.text_scraped == ""].index)
+    data = data[data.text_scraped != ""]
     data = data.drop_duplicates()
     data = data.drop(columns=["html_scraped"])
 
diff --git a/steps/data_preparation_steps/split_nhs_pages_step/split_nhs_pages_step.py b/steps/data_preparation_steps/split_nhs_pages_step/split_nhs_pages_step.py
new file mode 100644
index 0000000..c1c05c7
--- /dev/null
+++ b/steps/data_preparation_steps/split_nhs_pages_step/split_nhs_pages_step.py
@@ -0,0 +1,110 @@
+"""Split NHS pages step."""
+from typing import List, Dict
+
+import pandas as pd
+from bs4 import BeautifulSoup
+from zenml import step
+
+
+SOURCE_MAPPING = {
+    "nhs": {
+        "tag": "section",
+        "kwargs": {}
+    },
+    "mind": {
+        "tag": "div",
+        "kwargs": {
+            "class_": "column"
+        }
+    }
+}
+
+
+def split_html(html: str, tag: str, kwargs: Dict[str, str]) -> List[str]:
+    """Split html using the tag and some other optional Beautiful Soup arguments.
+
+    Args:
+        html (str): HTML text to split
+        tag (str): tag to split by
+        kwargs (Dict[str, str]): Beautiful Soup keyword arguments
+
+    Returns:
+        List[str]: list of HTML strings
+    """
+    soup = BeautifulSoup(html, "lxml")
+    sections = soup.find_all(tag, **kwargs)
+    return [str(section) for section in sections]
+
+
+def split_page(data: pd.Series, source: str) -> pd.DataFrame:
+    """Split a page.
+
+    Preserve other metadata.
+    * url is appended with an anchor suffix of the form '#section-{n}'
+    * uuid is appended with an anchor suffix of the form '-{n}'
+    * timestamp is kept as is
+
+    Args:
+        data (pd.Series): The scraped NHS data.
+            Index:
+                Name: uuid, dtype: object
+                Name: html_scraped, dtype: object
+                Name: timestamp, dtype: datetime64[ns]
+                Name: url, dtype: object
+        source (str): source of the page; determines how the page is split
+
+    Returns:
+        pd.DataFrame: The split page.
+            Index:
+                RangeIndex
+            Columns:
+                Name: uuid, dtype: object
+                Name: html_scraped, dtype: object
+                Name: timestamp, dtype: datetime64[ns]
+                Name: url, dtype: object
+    """
+    params = SOURCE_MAPPING[source]
+    if data.html_scraped:
+        sections = split_html(data.html_scraped, params["tag"], params["kwargs"])
+    else:
+        sections = []
+    return pd.DataFrame(
+        [{
+            "uuid": f"{data.uuid}-{i}",
+            "html_scraped": section,
+            "timestamp": data.timestamp,
+            "url": f"{data.url}#section-{i}"
+        } for i, section in enumerate(sections)],
+        columns=["uuid", "html_scraped", "timestamp", "url"]
+    )
+
+
+@step
+def split_pages(data: pd.DataFrame, source: str) -> pd.DataFrame:
+    """Split the NHS pages by the <section> tag.
+
+    Preserve other metadata.
+
+    Args:
+        data (pd.DataFrame): The scraped NHS data.
+            Index:
+                RangeIndex
+            Columns:
+                Name: uuid, dtype: object
+                Name: html_scraped, dtype: object
+                Name: timestamp, dtype: datetime64[ns]
+                Name: url, dtype: object
+        source (str): source of the page; determines how the page is split
+
+    Returns:
+        pd.DataFrame: The split data in the format described above.
+            Index:
+                RangeIndex
+            Columns:
+                Name: uuid, dtype: object
+                Name: html_scraped, dtype: object
+                Name: timestamp, dtype: datetime64[ns]
+                Name: url, dtype: object
+    """
+    frames = data.apply(split_page, args=(source,), axis=1)
+    return pd.concat(frames.tolist())
diff --git a/tests/test_steps/test_data_preparation_steps/test_split_nhs_pages_step.py b/tests/test_steps/test_data_preparation_steps/test_split_nhs_pages_step.py
new file mode 100644
index 0000000..45f4dbd
--- /dev/null
+++ b/tests/test_steps/test_data_preparation_steps/test_split_nhs_pages_step.py
@@ -0,0 +1,161 @@
+"""Test split NHS pages step."""
+from typing import List
+
+import pandas as pd
+import pytest
+from pandas._testing import assert_frame_equal
+
+from steps.data_preparation_steps.split_nhs_pages_step.split_nhs_pages_step import split_pages, split_page
+
+
+@pytest.fixture
+def nhs_pages_split() -> List[pd.DataFrame]:
+    """A fixture with NHS pages after splitting.
+
+    Returns:
+        List[pd.DataFrame]: The list of mock scraped NHS data.
+            Index:
+                RangeIndex
+            Columns:
+                Name: uuid, dtype: object
+                Name: html_scraped, dtype: object
+                Name: timestamp, dtype: datetime64[ns]
+                Name: url, dtype: object
+    """
+    return [
+        pd.DataFrame([], columns=["uuid", "html_scraped", "timestamp", "url"]),
+        pd.DataFrame([
+            {
+                "uuid": "1-0",
+                "html_scraped": "<section><p>Section1</p></section>",
+                "timestamp": pd.to_datetime(0),
+                "url": f"https://example.com/1#section-0",
+            }
+        ]),
+        pd.DataFrame([
+            {
+                "uuid": "2-0",
+                "html_scraped": "<section><p>Section1</p></section>",
+                "timestamp": pd.to_datetime(0),
+                "url": f"https://example.com/2#section-0",
+            },
+            {
+                "uuid": "2-1",
+                "html_scraped": "<section><p>Section2</p></section>",
+                "timestamp": pd.to_datetime(0),
+                "url": f"https://example.com/2#section-1",
+            }
+        ]),
+        pd.DataFrame([
+            {
+                "uuid": "3-0",
+                "html_scraped": "<section><h1>Heading1</h1><p>Section1</p></section>",
+                "timestamp": pd.to_datetime(0),
+                "url": f"https://example.com/3#section-0",
+            },
+            {
+                "uuid": "3-1",
+                "html_scraped": "<section><p>Section2</p></section>",
+                "timestamp": pd.to_datetime(0),
+                "url": f"https://example.com/3#section-1",
+            }
+        ]),
+        pd.DataFrame([
+            {
+                "uuid": "4-0",
+                "html_scraped": "<section><h1>Heading1</h1><p>Section1</p></section>",
+                "timestamp": pd.to_datetime(0),
+                "url": f"https://example.com/4#section-0",
+            },
+            {
+                "uuid": "4-1",
+                "html_scraped": "<section><p>Section2</p></section>",
+                "timestamp": pd.to_datetime(0),
+                "url": f"https://example.com/4#section-1",
+            }
+        ])
+    ]
+
+@pytest.fixture
+def nhs_pages_raw() -> pd.DataFrame:
+    """A fixture with NHS pages before splitting.
+
+    Returns:
+        pd.DataFrame: The mock scraped NHS data.
+            Index:
+                RangeIndex
+            Columns:
+                Name: uuid, dtype: object
+                Name: html_scraped, dtype: object
+                Name: timestamp, dtype: datetime64[ns]
+                Name: url, dtype: object
+    """
+    html_scraped = [
+        "<p>No section</p>",
+        "<section><p>Section1</p></section>",
+        "<section><p>Section1</p></section><section><p>Section2</p></section>",
+        "<section><h1>Heading1</h1><p>Section1</p></section><section><p>Section2</p></section>",
+        "<h1>Outside</h1><section><h1>Heading1</h1><p>Section1</p></section><section><p>Section2</p></section>",
+    ]
+    cases = [{
+        "uuid": i,
+        "url": f"https://example.com/{i}",
+        "html_scraped": text,
+        "timestamp": pd.to_datetime(0)
+    } for i, text in enumerate(html_scraped)]
+
+    return pd.DataFrame(cases)
+
+
+def test_split_page(nhs_pages_raw: pd.DataFrame, nhs_pages_split: List[pd.DataFrame]):
+    """Test splitting individual pages.
+
+    Args:
+        nhs_pages_raw (pd.DataFrame): The mock raw scraped NHS data.
+            Index:
+                RangeIndex
+            Columns:
+                Name: uuid, dtype: object
+                Name: html_scraped, dtype: object
+                Name: timestamp, dtype: datetime64[ns]
+                Name: url, dtype: object
+        nhs_pages_split (List[pd.DataFrame]): The list of mock split scraped NHS data.
+            Index:
+                RangeIndex
+            Columns:
+                Name: uuid, dtype: object
+                Name: html_scraped, dtype: object
+                Name: timestamp, dtype: datetime64[ns]
+                Name: url, dtype: object
+    """
+    nhs_pages_raw_rows = [row for _, row in nhs_pages_raw.iterrows()]
+    for expected, nhs_page in zip(nhs_pages_split, nhs_pages_raw_rows):
+        got = split_page(nhs_page, "NHS")
+        assert_frame_equal(expected, got)
+
+
+def test_split_nhs_pages(nhs_pages_raw: pd.DataFrame, nhs_pages_split: List[pd.DataFrame]):
+    """Test split NHS pages step.
+
+    Args:
+        nhs_pages_raw (pd.DataFrame): The mock raw scraped NHS data.
+            Index:
+                RangeIndex
+            Columns:
+                Name: uuid, dtype: object
+                Name: html_scraped, dtype: object
+                Name: timestamp, dtype: datetime64[ns]
+                Name: url, dtype: object
+        nhs_pages_split (List[pd.DataFrame]): The list of mock split scraped NHS data.
+            Index:
+                RangeIndex
+            Columns:
+                Name: uuid, dtype: object
+                Name: html_scraped, dtype: object
+                Name: timestamp, dtype: datetime64[ns]
+                Name: url, dtype: object
+
+    """
+    expected = pd.concat(nhs_pages_split)
+    got = split_pages.entrypoint(nhs_pages_raw)
+    assert_frame_equal(expected, got)
\ No newline at end of file