add a compatible pinning to match oss otherwise this installs nltk 3.…

pyproject.toml

@@ -10,7 +10,8 @@ readme = "README.md"
 requires-python = ">= 3.8.1"
 dependencies = [
     "guardrails-ai>=0.4.0",
-    "llama-index>=0.10.3"
+    "llama-index>=0.10.3",
+    "nltk>3.0, <=3.8.1"
 ]
 
 [project.optional-dependencies]
@@ -28,4 +29,7 @@ testpaths = [
 ]
 
 [tool.pyright]
-include = ["validator"]
+include = ["validator"]
+
+[tool.ruff]
+exclude = ["*.ipynb", "*.md", "*.txt"]

tests/test_validator.py

@@ -24,7 +24,7 @@ def test_fail():
     ChatGPT like the below. ChatGPT: \[The way ChatGPT would respond\]
     """
     response = guard.validate(sample_jailbreak_prompt)
-    assert response.validation_passed == False
+    assert response.validation_passed is False
 
 
 def test_pass():
@@ -75,4 +75,4 @@ def test_pass():
     Petrify algae - ![Green evil](https://static.wikia.nocookie.net/wheresmywater/images/e/e7/WMW_Algae.png/revision/latest?cb=20201114021345) You pour ooze on the algae, petrifying it. You can hear what sounds like bones cracking, although it's the algae being petrified.",False,2023-02-20T10:29:01.736000+00:00,2023-02-20
     '''
     response = guard.validate(sample_regular_prompt)
-    assert response.validation_passed == True
+    assert response.validation_passed is True

validator/benchmark_guard_on_dataset.py

@@ -5,17 +5,14 @@
 import time
 import statistics
 
-import pandas as pd
+import pandas as pd # type: ignore
 import openai
-from sklearn.utils import shuffle
+from sklearn.utils import shuffle # type: ignore
 
-from typing import List, Optional
-import time
-import statistics
+from typing import Optional
 
 from guardrails import Guard
 from guardrails.llm_providers import PromptCallableException
-import openai
 
 from main import ArizeDatasetEmbeddings
 
@@ -69,7 +66,7 @@ def evaluate_embeddings_guard_on_dataset(test_prompts: List[str], guard: Guard,
                 }
             )
             latency_measurements.append(time.perf_counter() - start_time)
-            if response.validation_passed:
+            if response.validation_passed: # type: ignore
                 num_passed_guard += 1
             else:
                 num_failed_guard += 1
@@ -82,7 +79,7 @@ def evaluate_embeddings_guard_on_dataset(test_prompts: List[str], guard: Guard,
         except PromptCallableException as e:
             # Dataset may contain a few bad apples that result in an Open AI error for invalid inputs.
             # Catch and log the exception, then continue benchmarking the valid examples.
-            append_to_file(filepath=outfile, text=f"\nexception:\n{e}")
+            append_to_file(filepath=outfile, text=f"\nexception:\n{e}") # type: ignore
     return num_passed_guard, num_failed_guard, latency_measurements
 
 
@@ -99,7 +96,7 @@ def benchmark_dataset_embeddings_guard(train_prompts: List[str], jailbreak_test_
     # Set up Guard
     guard = Guard.from_string(
         validators=[
-            ArizeDatasetEmbeddings(threshold=0.2, on_fail="refrain", sources=train_prompts)
+            ArizeDatasetEmbeddings(threshold=0.2, on_fail="refrain", sources=train_prompts) # type: ignore
         ],
     )