Skip to content

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

New workflow for framework testing

A-Vamshib945706b78dd
diff --git a/deepteam/red_teamer/red_teamer.py b/deepteam/red_teamer/red_teamer.py
index 76a1398f..50ff7d59 100644
--- a/deepteam/red_teamer/red_teamer.py
+++ b/deepteam/red_teamer/red_teamer.py
@@ -6,7 +6,8 @@ from rich.table import Table
import inspect
from rich import box
from enum import Enum
-
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from collections import defaultdict
from deepeval.models import DeepEvalBaseLLM
from deepeval.metrics.utils import initialize_model
@@ -15,6 +16,7 @@ from deepteam.test_case import RTTestCase
from deepeval.utils import get_or_create_event_loop
from deepteam.frameworks.frameworks import AISafetyFramework
+from deepteam.frameworks.risk_category import RiskCategory
from deepteam.telemetry import capture_red_teamer_run
from deepteam.attacks import BaseAttack
from deepteam.utils import validate_model_callback_signature
@@ -56,6 +58,97 @@ class RedTeamer:
max_concurrent=max_concurrent,
)
+ def _framework_assessment(
+ self,
+ model_callback: CallbackType,
+ simulator_model: DeepEvalBaseLLM = None,
+ evaluation_model: DeepEvalBaseLLM = None,
+ framework: Optional[AISafetyFramework] = None,
+ attacks_per_vulnerability_type: int = 1,
+ ignore_errors: bool = False,
+ reuse_simulated_test_cases: bool = False,
+ metadata: Optional[dict] = None,
+ ):
+ if not framework or framework._has_dataset:
+ raise ValueError("Please pass in a valid framework that does not rely on a dataset.")
+
+ def assess_risk_category(category: RiskCategory):
+ return self.red_team(
+ model_callback=model_callback,
+ attacks=category.attacks,
+ vulnerabilities=category.vulnerabilities,
+ simulator_model=simulator_model,
+ evaluation_model=evaluation_model,
+ ignore_errors=ignore_errors,
+ reuse_simulated_test_cases=reuse_simulated_test_cases,
+ metadata=metadata,
+ attacks_per_vulnerability_type=attacks_per_vulnerability_type,
+ )
+
+ results = {}
+
+ with ThreadPoolExecutor(
+ max_workers=min(len(framework.risk_categories), 32)
+ ) as executor:
+ future_map = {
+ executor.submit(assess_risk_category, category): category.name
+ for category in framework.risk_categories
+ }
+
+ pbar = tqdm(total=len(framework.risk_categories), desc=f"⏳ Running red-teaming for {framework.get_name()} Framework")
+ for future in as_completed(future_map):
+ name = future_map[future]
+ results[name] = future.result()
+ pbar.update(1)
+ pbar.close()
+
+ return results
+
+ async def _a_framework_assessment(
+ self,
+ model_callback: CallbackType,
+ simulator_model: DeepEvalBaseLLM = None,
+ evaluation_model: DeepEvalBaseLLM = None,
+ framework: Optional[AISafetyFramework] = None,
+ attacks_per_vulnerability_type: int = 1,
+ ignore_errors: bool = False,
+ reuse_simulated_test_cases: bool = False,
+ metadata: Optional[dict] = None,
+ )-> Dict[str, RiskAssessment]:
+ if not framework or framework._has_dataset:
+ raise ValueError("Please pass in a valid framework that does not rely on a dataset.")
+
+ async def assess_risk_category(category: RiskCategory):
+ return await self.a_red_team(
+ model_callback=model_callback,
+ attacks=category.attacks,
+ vulnerabilities=category.vulnerabilities,
+ simulator_model=simulator_model,
+ evaluation_model=evaluation_model,
+ ignore_errors=ignore_errors,
+ reuse_simulated_test_cases=reuse_simulated_test_cases,
+ metadata=metadata,
+ attacks_per_vulnerability_type=attacks_per_vulnerability_type,
+ )
+
+ tasks = {
+ asyncio.create_task(assess_risk_category(category)): category.name
+ for category in framework.risk_categories
+ }
+
+ results = {}
+
+ pbar = tqdm(total=len(tasks), desc=f"⏳ Running red-teaming for {framework.get_name()} Framework")
+
+ for task in asyncio.as_completed(tasks):
+ result = await task
+ name = tasks[task]
+ results[name] = result
+ pbar.update(1)
+
+ pbar.close()
+ return results
+
def red_team(
self,
model_callback: CallbackType,
@@ -79,6 +172,20 @@ class RedTeamer:
"You can only pass either 'framework' or 'attacks' and 'vulnerabilities' at the same time"
)
+ if framework and not framework._has_dataset:
+ framework_assessment = self._framework_assessment(
+ model_callback=model_callback,
+ simulator_model=simulator_model,
+ evaluation_model=evaluation_model,
+ framework=framework,
+ attacks_per_vulnerability_type=attacks_per_vulnerability_type,
+ ignore_errors=ignore_errors,
+ reuse_simulated_test_cases=reuse_simulated_test_cases,
+ metadata=metadata
+ )
+ self._print_framework_overview_table(framework_results=framework_assessment)
+ return framework_assessment
+
if self.async_mode:
validate_model_callback_signature(
model_callback=model_callback,
@@ -251,6 +358,20 @@ class RedTeamer:
raise ValueError(
"You can only pass either 'framework' or 'attacks' and 'vulnerabilities' at the same time"
)
+
+ if framework and not framework._has_dataset:
+ framework_assessment = await self._a_framework_assessment(
+ model_callback=model_callback,
+ simulator_model=simulator_model,
+ evaluation_model=evaluation_model,
+ framework=framework,
+ attacks_per_vulnerability_type=attacks_per_vulnerability_type,
+ ignore_errors=ignore_errors,
+ reuse_simulated_test_cases=reuse_simulated_test_cases,
+ metadata=metadata
+ )
+ self._print_framework_overview_table(framework_results=framework_assessment)
+ return framework_assessment
if framework:
if framework._has_dataset:
@@ -722,3 +843,79 @@ class RedTeamer:
console.print("\n" + "=" * 80)
console.print("[bold magenta]LLM red teaming complete.[/bold magenta]")
console.print("=" * 80 + "\n")
+
+ def _print_framework_overview_table(self, framework_results: dict):
+ console = Console()
+
+ console.print("\n" + "=" * 80)
+ console.print("[bold magenta]🏛 Framework-Level Risk Category Overview[/bold magenta]")
+ console.print("=" * 80)
+
+ table = Table(
+ show_header=True,
+ header_style="bold magenta",
+ border_style="blue",
+ box=box.HEAVY,
+ title="Risk Categories Overview",
+ title_style="bold magenta",
+ expand=True,
+ padding=(0, 1),
+ show_lines=True,
+ )
+
+ table.add_column("Risk Category", style="cyan", width=16)
+ table.add_column("Pass Rate", style="green", width=10, justify="center")
+ table.add_column("Passing", style="green", width=8, justify="center")
+ table.add_column("Failing", style="red", width=8, justify="center")
+ table.add_column("Errored", style="yellow", width=8, justify="center")
+ table.add_column("Vulnerabilities Tested", style="white", width=25)
+ table.add_column("Attack Methods Used", style="white", width=25)
+
+ for category_name in sorted(framework_results.keys()):
+ assessment = framework_results[category_name]
+
+ overview = assessment.overview
+ passing = sum(result.passing for result in overview.vulnerability_type_results)
+ failing = sum(result.failing for result in overview.vulnerability_type_results)
+ errored = sum(result.errored for result in overview.vulnerability_type_results)
+
+ total = passing + failing
+ pass_rate = passing / total if total > 0 else 0.0
+
+ vulnerability_groups = defaultdict(list)
+ for result in overview.vulnerability_type_results:
+ vulnerability_groups[result.vulnerability].append(result.vulnerability_type.value)
+
+ vuln_lines = []
+ for vulnerability_name, vulnerability_types in vulnerability_groups.items():
+ vuln_lines.append(f"[bold]{vulnerability_name}[/bold]")
+ for vulnerability_type in vulnerability_types:
+ vuln_lines.append(f" - {vulnerability_type}")
+
+ vulnerabilty_names = "\n".join(vuln_lines) if vuln_lines else "N/A"
+
+ # ----- Attack methods (simple list) -----
+ attack_list = [amr.attack_method for amr in overview.attack_method_results]
+ attack_names = "\n".join(attack_list) if attack_list else "N/A"
+
+ # Color-coded pass rate
+ if pass_rate >= 0.8:
+ pass_rate_str = f"[bold green]{pass_rate:.0%}[/bold green]"
+ elif pass_rate >= 0.5:
+ pass_rate_str = f"[bold yellow]{pass_rate:.0%}[/bold yellow]"
+ else:
+ pass_rate_str = f"[bold red]{pass_rate:.0%}[/bold red]"
+
+ table.add_row(
+ category_name,
+ pass_rate_str,
+ str(passing),
+ str(failing),
+ str(errored),
+ vulnerabilty_names,
+ attack_names,
+ )
+
+ console.print("\n")
+ console.print(table)
+ console.print("\n" + "=" * 80)