kailash 0.3.2__py3-none-any.whl → 0.4.0__py3-none-any.whl

kailash/nodes/rag/workflows.py

@@ -0,0 +1,575 @@
+"""
+RAG Workflow Nodes
+
+Pre-built WorkflowNode components that combine multiple RAG strategies
+and operations into reusable workflow patterns.
+"""
+
+import logging
+from typing import Any, Dict, Optional
+
+from ...workflow.builder import WorkflowBuilder
+from ..base import Node, NodeParameter, register_node
+from ..logic import SwitchNode
+from ..logic.workflow import WorkflowNode
+from .strategies import (
+    RAGConfig,
+    create_hierarchical_rag_workflow,
+    create_hybrid_rag_workflow,
+    create_semantic_rag_workflow,
+    create_statistical_rag_workflow,
+)
+
+logger = logging.getLogger(__name__)
+
+
+@register_node()
+class SimpleRAGWorkflowNode(WorkflowNode):
+    """
+    Simple RAG Workflow Node
+
+    Basic chunk → embed → store → retrieve pipeline using semantic chunking.
+    Perfect for getting started with RAG or simple document Q&A.
+    """
+
+    def __init__(
+        self, name: str = "simple_rag_workflow", config: Optional[RAGConfig] = None
+    ):
+        self.rag_config = config or RAGConfig()
+
+        # Create semantic RAG workflow
+        workflow_node = create_semantic_rag_workflow(self.rag_config)
+
+        # Initialize as WorkflowNode
+        super().__init__(
+            workflow=workflow_node.workflow,
+            name=name,
+            description="Simple RAG workflow with semantic chunking and dense retrieval",
+        )
+
+
+@register_node()
+class AdvancedRAGWorkflowNode(WorkflowNode):
+    """
+    Advanced RAG Workflow Node
+
+    Multi-stage RAG pipeline with quality checks, multiple retrieval strategies,
+    and result validation. Includes monitoring and performance optimization.
+    """
+
+    def __init__(
+        self, name: str = "advanced_rag_workflow", config: Optional[RAGConfig] = None
+    ):
+        self.rag_config = config or RAGConfig()
+
+        # Build advanced workflow
+        workflow = self._create_advanced_workflow()
+
+        super().__init__(
+            workflow=workflow,
+            name=name,
+            description="Advanced RAG with quality checks and multi-stage processing",
+        )
+
+    def _create_advanced_workflow(self):
+        """Create advanced RAG workflow with quality checks and monitoring"""
+        builder = WorkflowBuilder()
+
+        # Document quality analyzer
+        quality_analyzer_id = builder.add_node(
+            "PythonCodeNode",
+            node_id="quality_analyzer",
+            config={
+                "code": """
+# Analyze document quality and determine best RAG strategy
+def analyze_documents(documents):
+    analysis = {
+        "total_docs": len(documents),
+        "avg_length": sum(len(doc.get("content", "")) for doc in documents) / len(documents) if documents else 0,
+        "has_structure": any("section" in doc or "heading" in doc for doc in documents),
+        "is_technical": any(keyword in doc.get("content", "").lower()
+                          for doc in documents
+                          for keyword in ["code", "function", "algorithm", "api", "class"]),
+        "recommended_strategy": "semantic"  # Default
+    }
+
+    # Determine best strategy based on analysis
+    if analysis["has_structure"] and analysis["avg_length"] > 2000:
+        analysis["recommended_strategy"] = "hierarchical"
+    elif analysis["is_technical"]:
+        analysis["recommended_strategy"] = "statistical"
+    elif analysis["total_docs"] > 100:
+        analysis["recommended_strategy"] = "hybrid"
+
+    return analysis
+
+result = {"analysis": analyze_documents(documents), "documents": documents}
+"""
+            },
+        )
+
+        # Strategy router using switch node
+        router_id = builder.add_node(
+            "SwitchNode",
+            node_id="strategy_router",
+            config={
+                "condition_field": "analysis.recommended_strategy",
+                "routes": {
+                    "semantic": "semantic_rag_pipeline",
+                    "statistical": "statistical_rag_pipeline",
+                    "hybrid": "hybrid_rag_pipeline",
+                    "hierarchical": "hierarchical_rag_pipeline",
+                },
+            },
+        )
+
+        # Add all RAG strategy pipelines
+        semantic_workflow = create_semantic_rag_workflow(self.rag_config)
+        statistical_workflow = create_statistical_rag_workflow(self.rag_config)
+        hybrid_workflow = create_hybrid_rag_workflow(self.rag_config)
+        hierarchical_workflow = create_hierarchical_rag_workflow(self.rag_config)
+
+        semantic_id = builder.add_node(
+            "WorkflowNode",
+            node_id="semantic_rag_pipeline",
+            config={"workflow": semantic_workflow.workflow},
+        )
+
+        statistical_id = builder.add_node(
+            "WorkflowNode",
+            node_id="statistical_rag_pipeline",
+            config={"workflow": statistical_workflow.workflow},
+        )
+
+        hybrid_id = builder.add_node(
+            "WorkflowNode",
+            node_id="hybrid_rag_pipeline",
+            config={"workflow": hybrid_workflow.workflow},
+        )
+
+        hierarchical_id = builder.add_node(
+            "WorkflowNode",
+            node_id="hierarchical_rag_pipeline",
+            config={"workflow": hierarchical_workflow.workflow},
+        )
+
+        # Quality validator
+        validator_id = builder.add_node(
+            "PythonCodeNode",
+            node_id="quality_validator",
+            config={
+                "code": """
+def validate_rag_results(results, analysis):
+    validation = {
+        "results_count": len(results.get("documents", [])),
+        "avg_score": sum(results.get("scores", [])) / len(results.get("scores", [])) if results.get("scores") else 0,
+        "quality_score": 0.0,
+        "passed": False
+    }
+
+    # Calculate quality score
+    if validation["results_count"] > 0:
+        validation["quality_score"] = validation["avg_score"] * (validation["results_count"] / 5.0)
+        validation["passed"] = validation["quality_score"] > 0.5
+
+    return {
+        "results": results,
+        "validation": validation,
+        "strategy_used": analysis.get("recommended_strategy"),
+        "final_status": "passed" if validation["passed"] else "needs_improvement"
+    }
+
+result = validate_rag_results(rag_results, analysis)
+"""
+            },
+        )
+
+        # Connect the advanced pipeline
+        builder.add_connection(quality_analyzer_id, "result", router_id, "input")
+
+        # Connect router to all strategy pipelines
+        builder.add_connection(router_id, semantic_id, route="semantic")
+        builder.add_connection(router_id, statistical_id, route="statistical")
+        builder.add_connection(router_id, hybrid_id, route="hybrid")
+        builder.add_connection(router_id, hierarchical_id, route="hierarchical")
+
+        # Connect all pipelines to validator
+        builder.add_connection(semantic_id, "output", validator_id, "rag_results")
+        builder.add_connection(statistical_id, "output", validator_id, "rag_results")
+        builder.add_connection(hybrid_id, "output", validator_id, "rag_results")
+        builder.add_connection(hierarchical_id, "output", validator_id, "rag_results")
+        builder.add_connection(quality_analyzer_id, "result", validator_id, "analysis")
+
+        return builder.build(name="advanced_rag_workflow")
+
+
+@register_node()
+class AdaptiveRAGWorkflowNode(WorkflowNode):
+    """
+    Adaptive RAG Workflow Node
+
+    AI-driven strategy selection that uses LLM to analyze documents and queries
+    to automatically choose the optimal RAG approach for each use case.
+    """
+
+    def __init__(
+        self,
+        name: str = "adaptive_rag_workflow",
+        llm_model: str = "gpt-4",
+        config: Optional[RAGConfig] = None,
+    ):
+        self.rag_config = config or RAGConfig()
+        self.llm_model = llm_model
+
+        # Build adaptive workflow
+        workflow = self._create_adaptive_workflow()
+
+        super().__init__(
+            workflow=workflow,
+            name=name,
+            description="AI-driven adaptive RAG with intelligent strategy selection",
+        )
+
+    def _create_adaptive_workflow(self):
+        """Create adaptive RAG workflow with LLM-driven strategy selection"""
+        builder = WorkflowBuilder()
+
+        # LLM Strategy Analyzer
+        llm_analyzer_id = builder.add_node(
+            "LLMAgentNode",
+            node_id="rag_strategy_analyzer",
+            config={
+                "model": self.llm_model,
+                "provider": "openai",
+                "system_prompt": """You are a RAG strategy expert. Analyze documents and queries to recommend the optimal RAG approach.
+
+Available strategies:
+- semantic: Best for narrative content, general Q&A, semantic similarity
+- statistical: Best for technical docs, code, structured content, keyword matching
+- hybrid: Best for mixed content, combines semantic + statistical (20-30% better performance)
+- hierarchical: Best for long documents, structured content with sections/headings
+
+Analyze the input and respond with ONLY a JSON object:
+{
+    "recommended_strategy": "semantic|statistical|hybrid|hierarchical",
+    "reasoning": "Brief explanation of why this strategy is optimal",
+    "confidence": 0.0-1.0,
+    "fallback_strategy": "backup strategy if primary fails"
+}""",
+                "prompt_template": """Analyze these documents for optimal RAG strategy:
+
+Document Analysis:
+- Count: {document_count}
+- Average length: {avg_length} characters
+- Has structure (headings/sections): {has_structure}
+- Technical content detected: {is_technical}
+- Content types: {content_types}
+
+Query (if provided): {query}
+
+Recommend the optimal RAG strategy:""",
+            },
+        )
+
+        # Document preprocessor for LLM analysis
+        preprocessor_id = builder.add_node(
+            "PythonCodeNode",
+            node_id="document_preprocessor",
+            config={
+                "code": """
+import re
+
+def analyze_for_llm(documents, query=""):
+    if not documents:
+        return {
+            "document_count": 0,
+            "avg_length": 0,
+            "has_structure": False,
+            "is_technical": False,
+            "content_types": [],
+            "query": query
+        }
+
+    # Analyze documents
+    total_length = sum(len(doc.get("content", "")) for doc in documents)
+    avg_length = total_length / len(documents)
+
+    # Check for structure
+    has_structure = any(
+        any(keyword in doc.get("content", "").lower()
+            for keyword in ["# ", "## ", "### ", "heading", "section", "chapter"])
+        for doc in documents
+    )
+
+    # Check for technical content
+    technical_keywords = ["code", "function", "class", "algorithm", "api", "import", "def ", "return", "variable"]
+    is_technical = any(
+        any(keyword in doc.get("content", "").lower()
+            for keyword in technical_keywords)
+        for doc in documents
+    )
+
+    # Determine content types
+    content_types = []
+    if has_structure:
+        content_types.append("structured")
+    if is_technical:
+        content_types.append("technical")
+    if avg_length > 2000:
+        content_types.append("long_form")
+    if len(documents) > 50:
+        content_types.append("large_collection")
+
+    return {
+        "document_count": len(documents),
+        "avg_length": int(avg_length),
+        "has_structure": has_structure,
+        "is_technical": is_technical,
+        "content_types": content_types,
+        "query": query,
+        "documents": documents
+    }
+
+result = analyze_for_llm(documents, query)
+"""
+            },
+        )
+
+        # Strategy executor with switch
+        executor_id = builder.add_node(
+            "SwitchNode",
+            node_id="strategy_executor",
+            config={
+                "condition_field": "recommended_strategy",
+                "routes": {
+                    "semantic": "semantic_pipeline",
+                    "statistical": "statistical_pipeline",
+                    "hybrid": "hybrid_pipeline",
+                    "hierarchical": "hierarchical_pipeline",
+                },
+            },
+        )
+
+        # Add strategy pipelines
+        semantic_workflow = create_semantic_rag_workflow(self.rag_config)
+        statistical_workflow = create_statistical_rag_workflow(self.rag_config)
+        hybrid_workflow = create_hybrid_rag_workflow(self.rag_config)
+        hierarchical_workflow = create_hierarchical_rag_workflow(self.rag_config)
+
+        semantic_pipeline_id = builder.add_node(
+            "WorkflowNode",
+            node_id="semantic_pipeline",
+            config={"workflow": semantic_workflow.workflow},
+        )
+
+        statistical_pipeline_id = builder.add_node(
+            "WorkflowNode",
+            node_id="statistical_pipeline",
+            config={"workflow": statistical_workflow.workflow},
+        )
+
+        hybrid_pipeline_id = builder.add_node(
+            "WorkflowNode",
+            node_id="hybrid_pipeline",
+            config={"workflow": hybrid_workflow.workflow},
+        )
+
+        hierarchical_pipeline_id = builder.add_node(
+            "WorkflowNode",
+            node_id="hierarchical_pipeline",
+            config={"workflow": hierarchical_workflow.workflow},
+        )
+
+        # Results aggregator
+        aggregator_id = builder.add_node(
+            "PythonCodeNode",
+            node_id="results_aggregator",
+            config={
+                "code": """
+def aggregate_adaptive_results(rag_results, llm_decision, preprocessed_data):
+    return {
+        "results": rag_results,
+        "strategy_used": llm_decision.get("recommended_strategy"),
+        "llm_reasoning": llm_decision.get("reasoning"),
+        "confidence": llm_decision.get("confidence"),
+        "document_analysis": {
+            "count": preprocessed_data.get("document_count"),
+            "avg_length": preprocessed_data.get("avg_length"),
+            "content_types": preprocessed_data.get("content_types")
+        },
+        "adaptive_metadata": {
+            "llm_model_used": "gpt-4",
+            "strategy_selection_method": "llm_analysis",
+            "fallback_available": llm_decision.get("fallback_strategy")
+        }
+    }
+
+result = aggregate_adaptive_results(rag_results, llm_decision, preprocessed_data)
+"""
+            },
+        )
+
+        # Connect adaptive pipeline
+        builder.add_connection(preprocessor_id, "result", llm_analyzer_id, "input")
+        builder.add_connection(llm_analyzer_id, "result", executor_id, "input")
+        builder.add_connection(
+            preprocessor_id, "result", executor_id, "preprocessed_data"
+        )
+
+        # Connect executor to strategy pipelines
+        builder.add_connection(executor_id, semantic_pipeline_id, route="semantic")
+        builder.add_connection(
+            executor_id, statistical_pipeline_id, route="statistical"
+        )
+        builder.add_connection(executor_id, hybrid_pipeline_id, route="hybrid")
+        builder.add_connection(
+            executor_id, hierarchical_pipeline_id, route="hierarchical"
+        )
+
+        # Connect all pipelines to aggregator
+        builder.add_connection(
+            semantic_pipeline_id, "output", aggregator_id, "rag_results"
+        )
+        builder.add_connection(
+            statistical_pipeline_id, "output", aggregator_id, "rag_results"
+        )
+        builder.add_connection(
+            hybrid_pipeline_id, "output", aggregator_id, "rag_results"
+        )
+        builder.add_connection(
+            hierarchical_pipeline_id, "output", aggregator_id, "rag_results"
+        )
+        builder.add_connection(llm_analyzer_id, "result", aggregator_id, "llm_decision")
+        builder.add_connection(
+            preprocessor_id, "result", aggregator_id, "preprocessed_data"
+        )
+
+        return builder.build(name="adaptive_rag_workflow")
+
+
+@register_node()
+class RAGPipelineWorkflowNode(WorkflowNode):
+    """
+    Configurable RAG Pipeline Workflow Node
+
+    Flexible RAG workflow that can be configured for different use cases
+    without code changes. Supports all strategies and custom configurations.
+    """
+
+    def __init__(
+        self,
+        name: str = "rag_pipeline",
+        default_strategy: str = "hybrid",
+        config: Optional[RAGConfig] = None,
+    ):
+        self.rag_config = config or RAGConfig()
+        self.default_strategy = default_strategy
+
+        # Build configurable workflow
+        workflow = self._create_configurable_workflow()
+
+        super().__init__(
+            workflow=workflow,
+            name=name,
+            description=f"Configurable RAG pipeline with {default_strategy} as default strategy",
+        )
+
+    def _create_configurable_workflow(self):
+        """Create configurable RAG workflow"""
+        builder = WorkflowBuilder()
+
+        # Configuration processor
+        config_processor_id = builder.add_node(
+            "PythonCodeNode",
+            node_id="config_processor",
+            config={
+                "code": f"""
+def process_config(documents, query="", strategy="{self.default_strategy}", **kwargs):
+    # Merge user config with defaults
+    processed_config = {{
+        "strategy": strategy,
+        "documents": documents,
+        "query": query,
+        "chunk_size": kwargs.get("chunk_size", {self.rag_config.chunk_size}),
+        "chunk_overlap": kwargs.get("chunk_overlap", {self.rag_config.chunk_overlap}),
+        "embedding_model": kwargs.get("embedding_model", "{self.rag_config.embedding_model}"),
+        "retrieval_k": kwargs.get("retrieval_k", {self.rag_config.retrieval_k})
+    }}
+
+    return processed_config
+
+result = process_config(documents, **kwargs)
+"""
+            },
+        )
+
+        # Strategy dispatcher
+        dispatcher_id = builder.add_node(
+            "SwitchNode",
+            node_id="strategy_dispatcher",
+            config={
+                "condition_field": "strategy",
+                "routes": {
+                    "semantic": "semantic_strategy",
+                    "statistical": "statistical_strategy",
+                    "hybrid": "hybrid_strategy",
+                    "hierarchical": "hierarchical_strategy",
+                },
+                "default_route": "hybrid_strategy",
+            },
+        )
+
+        # Add all strategy implementations
+        strategies = {
+            "semantic": create_semantic_rag_workflow(self.rag_config),
+            "statistical": create_statistical_rag_workflow(self.rag_config),
+            "hybrid": create_hybrid_rag_workflow(self.rag_config),
+            "hierarchical": create_hierarchical_rag_workflow(self.rag_config),
+        }
+
+        strategy_ids = {}
+        for strategy_name, workflow_node in strategies.items():
+            strategy_id = builder.add_node(
+                "WorkflowNode",
+                node_id=f"{strategy_name}_strategy",
+                config={"workflow": workflow_node.workflow},
+            )
+            strategy_ids[strategy_name] = strategy_id
+
+        # Results formatter
+        formatter_id = builder.add_node(
+            "PythonCodeNode",
+            node_id="results_formatter",
+            config={
+                "code": """
+def format_pipeline_results(results, config):
+    return {
+        "results": results,
+        "strategy_used": config.get("strategy"),
+        "configuration": config,
+        "pipeline_type": "configurable",
+        "success": True if results else False
+    }
+
+result = format_pipeline_results(strategy_results, processed_config)
+"""
+            },
+        )
+
+        # Connect configurable pipeline
+        builder.add_connection(config_processor_id, "result", dispatcher_id, "input")
+
+        # Connect dispatcher to all strategies
+        for strategy_name, strategy_id in strategy_ids.items():
+            builder.add_connection(
+                dispatcher_id, strategy_id, route=f"{strategy_name}_strategy"
+            )
+            builder.add_connection(
+                strategy_id, "output", formatter_id, "strategy_results"
+            )
+
+        builder.add_connection(
+            config_processor_id, "result", formatter_id, "processed_config"
+        )
+
+        return builder.build(name="configurable_rag_pipeline")

kailash/nodes/security/__init__.py

@@ -0,0 +1,19 @@
+"""Security-related nodes for the Kailash SDK."""
+
+from .abac_evaluator import ABACPermissionEvaluatorNode
+from .audit_log import AuditLogNode
+from .behavior_analysis import BehaviorAnalysisNode
+from .credential_manager import CredentialManagerNode
+from .rotating_credentials import RotatingCredentialNode
+from .security_event import SecurityEventNode
+from .threat_detection import ThreatDetectionNode
+
+__all__ = [
+    "CredentialManagerNode",
+    "RotatingCredentialNode",
+    "AuditLogNode",
+    "SecurityEventNode",
+    "ThreatDetectionNode",
+    "ABACPermissionEvaluatorNode",
+    "BehaviorAnalysisNode",
+]