From 4477d01fba797176a42101b89943efb23989e54f Mon Sep 17 00:00:00 2001
From: Rabah Abdul Khalek <rabah.khalek@gmail.com>
Date: Wed, 20 Dec 2023 13:12:41 +0100
Subject: [PATCH] updated readme, requirements, validation

---
 README.md        |  4 ++++
 requirements.txt |  6 ++++++
 validate.py      | 28 ++++++++++++++++++++++++++++
 3 files changed, 38 insertions(+)
 create mode 100644 requirements.txt
 create mode 100644 validate.py

diff --git a/README.md b/README.md
index 114bbd5..06c3e42 100644
--- a/README.md
+++ b/README.md
@@ -6,3 +6,7 @@ The `prompt_injections.csv` file is a concatenation of prompts from the followin
 - https://github.com/agencyenterprise/PromptInject
 
 with their respective licenses in the `licenses` directory.
+
+## Setup and validation
+
+The pip `requirement.txt` is only needed to run `validate.py`, a minimal validation script to ensure that the prompt injection data is generated in the correct format.
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..09844fe
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,6 @@
+numpy==1.26.2
+pandas==2.1.4
+python-dateutil==2.8.2
+pytz==2023.3.post1
+six==1.16.0
+tzdata==2023.3
diff --git a/validate.py b/validate.py
new file mode 100644
index 0000000..925a522
--- /dev/null
+++ b/validate.py
@@ -0,0 +1,28 @@
+import pandas as pd
+import ast
+
+INJECTION_DATA_PATH = "prompt_injections.csv"
+GISKARD_META_PATH = "giskard_meta_data.csv"
+
+def _check_matching_dfs_len(df1, df2):
+    if len(df1) != len(df2):
+        raise ValueError(
+            f"{__name__}: {INJECTION_DATA_PATH} and {GISKARD_META_PATH} should "
+            "have the same length and should be a one-to-one mapping of each other."
+        )
+
+def _check_meta_df_requirements(df):
+    if "expected_strings" not in df.columns:
+        raise ValueError(f"{__name__}: expected_strings are needed for the evaluation.")
+
+    if df.expected_strings.isnull().values.any():
+        raise ValueError(f"{__name__}: expected_strings column cannot have any NaN values.")
+    df.expected_strings = df.expected_strings.apply(ast.literal_eval)
+    
+    
+if __name__ == "__main__":
+    prompt_injections_df = pd.read_csv(INJECTION_DATA_PATH, index_col=["index"])
+    meta_df = pd.read_csv(GISKARD_META_PATH, index_col=["index"])
+    _check_matching_dfs_len(meta_df, prompt_injections_df)
+    _check_meta_df_requirements(meta_df)
+    print("Validation passed succesfully!")
\ No newline at end of file