agent-evaluator 0.5.3__py3-none-any.whl

agent_evaluator/__init__.py

@@ -0,0 +1,144 @@
+"""
+Agent Evaluator SDK v0.5.3
+===========================
+
+Production-ready evaluation framework for AI agents.
+
+Quick Start (Basic):
+    >>> from agent_evaluator import PerformanceMonitor, create_taskresult
+    >>>
+    >>> monitor = PerformanceMonitor()
+    >>> task = create_taskresult(
+    ...     task_id="task_001",
+    ...     question="What is the capital?",
+    ...     response="Seoul",
+    ...     ground_truth="Seoul",
+    ...     execution_time=1.2
+    ... )
+    >>> monitor.record_task(task)
+    >>> monitor.save_to_file("results.json")  # Auto-includes full report!
+
+Quick Start (Context Manager):
+    >>> from agent_evaluator import evaluation_session
+    >>>
+    >>> with evaluation_session("results.json") as monitor:
+    ...     task = create_taskresult(...)
+    ...     monitor.record_task(task)
+    >>> # Auto-saved!
+
+Quick Start (LLM Integration):
+    >>> from agent_evaluator import PerformanceMonitor, LLMHelper
+    >>>
+    >>> monitor = PerformanceMonitor()
+    >>> llm = LLMHelper(monitor)
+    >>> task = llm.evaluate_openai("qa_001", "What is AI?", ground_truth="...")
+    >>> # Auto-recorded in monitor!
+"""
+
+__version__ = "0.5.3"
+__author__ = "Sungwoo Kim"
+
+# Import from core module
+# Config & init helpers (cli.main 임포트 없이 제공 — import-time side-effect 없음)
+from .config import get_settings, init_from_app, load_env
+
+# Transparency
+from .utils.test_transparency_manager import (
+    AnnotationType,
+    TestStepStatus,
+    TestTransparencyManager,
+)
+from .core.agent_evaluator import (
+    AccuracyEvaluator,
+    AgentCoordinationTracker,
+    EvaluationReport,
+    HallucinationDetector,
+    # Security Metrics (Layer 1 & 2)
+    InputSanitizationTracker,
+    LatencyTracker,
+    OutputLeakageDetector,
+    PerformanceMonitor,
+    PrivilegeEscalationDetector,
+    ResponseQualityEvaluator,
+    RetryCorrectionTracker,
+    TaskCompletionTracker,
+    TaskResult,
+    TaskType,
+    TokenEconomyTracker,
+    ToolAuthorizationTracker,
+    ToolCallAnalyzer,
+    ToolChainAttackDetector,
+    ToolSelectionTracker,
+    WorkflowExecutionTracker,
+)
+
+# Import Hybrid Monitor (with external library integration)
+from .core.hybrid_monitor import (
+    ExtendedTaskResult,
+    HybridEvaluationReport,
+    HybridPerformanceMonitor,
+)
+
+# Import context managers
+from .core.monitor_context import evaluation_session, hybrid_evaluation_session
+
+# Import helpers with simplified names
+from .helpers.taskresult_helpers import create_taskresult_from_execution as create_taskresult
+from .integrations.llm_helpers import (
+    AnthropicEvaluationHelper as ClaudeHelper,
+)
+
+# Import LLM helpers
+from .integrations.llm_helpers import (
+    LLMEvaluationHelper as LLMHelper,
+)
+
+__all__ = [
+    # Core
+    'PerformanceMonitor',
+    'TaskResult',
+    'TaskType',
+    'EvaluationReport',
+
+    # Hybrid Monitor
+    'HybridPerformanceMonitor',
+    'ExtendedTaskResult',
+    'HybridEvaluationReport',
+
+    # Helpers (simplified API)
+    'create_taskresult',  # Simplified name
+    'evaluation_session',  # Context manager
+    'hybrid_evaluation_session',  # Context manager
+    'LLMHelper',  # Simplified name
+    'ClaudeHelper',  # Simplified name
+
+    # Config & setup
+    'load_env',       # Smart .env loader (priority: system > CWD .env > global)
+    'get_settings',   # Settings singleton
+    'init_from_app',  # Programmatic init for library callers
+
+    # Trackers (for advanced users)
+    'TaskCompletionTracker',
+    'AccuracyEvaluator',
+    'HallucinationDetector',
+    'ResponseQualityEvaluator',
+    'LatencyTracker',
+    'TokenEconomyTracker',
+    'ToolCallAnalyzer',
+    'RetryCorrectionTracker',
+    'ToolSelectionTracker',
+    'AgentCoordinationTracker',
+    'WorkflowExecutionTracker',
+
+    # Security Metrics
+    'InputSanitizationTracker',
+    'OutputLeakageDetector',
+    'ToolAuthorizationTracker',
+    'PrivilegeEscalationDetector',
+    'ToolChainAttackDetector',
+
+    # Transparency
+    'TestTransparencyManager',
+    'AnnotationType',
+    'TestStepStatus',
+]

agent_evaluator/cli/__init__.py

@@ -0,0 +1 @@
+# agent_evaluator CLI package