Skip to content

DeepTeam is a framework to red team LLMs and AI agents.

Fix ROT13

penguine1d4bc5b239ac
diff --git a/b.py b/b.py
index f134f1c1..8ddd2255 100644
--- a/b.py
+++ b/b.py
@@ -2,11 +2,17 @@ from deepteam import red_team
from deepteam.vulnerabilities import Bias
from deepteam.attacks.single_turn import PromptInjection
+
def model_callback(input: str) -> str:
# Replace this with your LLM application
return f"I'm sorry but I can't answer this: {input}"
+
bias = Bias(types=["race"])
prompt_injection = PromptInjection()
-red_team(model_callback=model_callback, vulnerabilities=[bias], attacks=[prompt_injection])
\ No newline at end of file
+red_team(
+ model_callback=model_callback,
+ vulnerabilities=[bias],
+ attacks=[prompt_injection],
+)
diff --git a/deepteam/attacks/multi_turn/tree_jailbreaking/tree_jailbreaking.py b/deepteam/attacks/multi_turn/tree_jailbreaking/tree_jailbreaking.py
index 9b0ac080..e42632dd 100644
--- a/deepteam/attacks/multi_turn/tree_jailbreaking/tree_jailbreaking.py
+++ b/deepteam/attacks/multi_turn/tree_jailbreaking/tree_jailbreaking.py
@@ -350,14 +350,10 @@ class TreeJailbreaking(BaseAttack):
##################################################
def _generate_schema(self, prompt: str, schema: BaseModel):
- return generate_schema(
- prompt, schema, self.simulator_model
- )
+ return generate_schema(prompt, schema, self.simulator_model)
async def _a_generate_schema(self, prompt: str, schema: BaseModel):
- return await a_generate_schema(
- prompt, schema, self.simulator_model
- )
+ return await a_generate_schema(prompt, schema, self.simulator_model)
def get_name(self) -> str:
return "Tree Jailbreaking"
diff --git a/deepteam/attacks/single_turn/multilingual/schema.py b/deepteam/attacks/single_turn/multilingual/schema.py
index e1038a73..cc1f5532 100644
--- a/deepteam/attacks/single_turn/multilingual/schema.py
+++ b/deepteam/attacks/single_turn/multilingual/schema.py
@@ -10,6 +10,7 @@ class ComplianceData(BaseModel):
Attributes:
non_compliant (bool): True if the prompt is non-compliant, False otherwise.
"""
+
non_compliant: bool
@@ -18,4 +19,5 @@ class IsTranslation(BaseModel):
Attributes:
is_translation (bool): True if the prompt is a translation, False otherwise.
"""
+
is_translation: bool
diff --git a/deepteam/attacks/single_turn/rot13/rot13.py b/deepteam/attacks/single_turn/rot13/rot13.py
index 37d689a2..df7c8766 100644
--- a/deepteam/attacks/single_turn/rot13/rot13.py
+++ b/deepteam/attacks/single_turn/rot13/rot13.py
@@ -2,6 +2,9 @@ from deepteam.attacks import BaseAttack
class ROT13(BaseAttack):
+ def __init__(self, weight: int = 1):
+ self.weight = weight
+
def enhance(self, attack: str) -> str:
"""Enhance the attack using ROT13 encoding."""
return attack.translate(
diff --git a/tests/test_deepteam.py b/tests/test_deepteam.py
index 7f182be6..58429998 100644
--- a/tests/test_deepteam.py
+++ b/tests/test_deepteam.py
@@ -1,5 +1,5 @@
from deepteam import red_team
+
def test_red_team():
assert red_team
-