kailash 0.3.1__py3-none-any.whl → 0.4.0__py3-none-any.whl

kailash/nodes/rag/__init__.py

@@ -0,0 +1,284 @@
+"""
+RAG (Retrieval Augmented Generation) Toolkit
+
+Comprehensive RAG implementation with swappable strategies, conditional routing,
+and enterprise-grade features. Supports semantic, statistical, hybrid, and
+hierarchical RAG approaches with advanced similarity methods.
+
+## Quick Start
+
+```python
+from kailash.nodes.rag import (
+    RAGStrategyRouterNode,
+    SemanticRAGWorkflowNode,
+    HybridRAGWorkflowNode,
+    AdaptiveRAGWorkflowNode,
+    ColBERTRetrievalNode,
+    AsyncParallelRAGNode
+)
+
+# Method 1: AI-driven strategy selection
+workflow.add_node("rag_router", RAGStrategyRouterNode())
+workflow.add_node("rag_processor", SwitchNode(
+    condition_field="strategy",
+    routes={
+        "semantic": "semantic_rag",
+        "hybrid": "hybrid_rag",
+        "hierarchical": "hierarchical_rag"
+    }
+))
+
+# Method 2: Direct strategy selection
+workflow.add_node("rag", HybridRAGWorkflowNode(
+    chunking_strategy="semantic",
+    retrieval_method="hybrid",
+    vector_db="postgresql"
+))
+
+# Method 3: Advanced similarity approaches
+workflow.add_node("colbert", ColBERTRetrievalNode())
+workflow.add_node("multi_vector", MultiVectorRetrievalNode())
+```
+
+## Available Components
+
+### Core Strategies
+- **SemanticRAGNode**: Semantic chunking + dense retrieval
+- **StatisticalRAGNode**: Statistical chunking + sparse retrieval
+- **HybridRAGNode**: Combines semantic + statistical approaches
+- **HierarchicalRAGNode**: Multi-level document processing
+
+### Advanced Techniques
+- **SelfCorrectingRAGNode**: Iterative verification and refinement
+- **RAGFusionNode**: Multi-query approach with result fusion
+- **HyDENode**: Hypothetical Document Embeddings
+- **StepBackRAGNode**: Abstract reasoning with background context
+
+### Similarity Approaches
+- **DenseRetrievalNode**: Advanced dense embeddings with instruction-awareness
+- **SparseRetrievalNode**: BM25, TF-IDF with query expansion
+- **ColBERTRetrievalNode**: Token-level late interaction
+- **MultiVectorRetrievalNode**: Multiple representations per document
+- **CrossEncoderRerankNode**: Two-stage retrieval with reranking
+- **HybridFusionNode**: RRF and advanced fusion methods
+- **PropositionBasedRetrievalNode**: Atomic fact extraction
+
+### Query Processing
+- **QueryExpansionNode**: Synonym and concept expansion
+- **QueryDecompositionNode**: Complex query breakdown
+- **QueryRewritingNode**: Query optimization for retrieval
+- **QueryIntentClassifierNode**: Intent-based routing
+- **MultiHopQueryPlannerNode**: Multi-step reasoning
+- **AdaptiveQueryProcessorNode**: Intelligent query enhancement
+
+### Performance Optimization
+- **CacheOptimizedRAGNode**: Multi-level caching with semantic similarity
+- **AsyncParallelRAGNode**: Concurrent strategy execution
+- **StreamingRAGNode**: Real-time progressive retrieval
+- **BatchOptimizedRAGNode**: High-throughput batch processing
+
+### Workflow Nodes
+- **SimpleRAGWorkflowNode**: Basic chunk → embed → store → retrieve
+- **AdvancedRAGWorkflowNode**: Multi-stage with quality checks
+- **AdaptiveRAGWorkflowNode**: AI-driven strategy selection
+
+### Router & Utilities
+- **RAGStrategyRouterNode**: LLM-powered strategy selection
+- **RAGQualityAnalyzerNode**: Quality assessment and optimization
+- **RAGPerformanceMonitorNode**: Performance tracking and metrics
+- **RAGWorkflowRegistry**: Component discovery and recommendations
+
+### Graph-Enhanced RAG
+- **GraphRAGNode**: Knowledge graph construction and querying
+- **GraphBuilderNode**: Build knowledge graphs from documents
+- **GraphQueryNode**: Execute complex graph queries
+
+### Agentic RAG
+- **AgenticRAGNode**: Autonomous reasoning with tool use
+- **ToolAugmentedRAGNode**: RAG with specialized tool integration
+- **ReasoningRAGNode**: Multi-step reasoning chains
+
+### Multimodal RAG
+- **MultimodalRAGNode**: Text + image retrieval and generation
+- **VisualQuestionAnsweringNode**: Answer questions about images
+- **ImageTextMatchingNode**: Cross-modal similarity matching
+
+### Real-time & Streaming
+- **RealtimeRAGNode**: Live data updates and monitoring
+- **IncrementalIndexNode**: Efficient incremental updates
+
+### Evaluation & Testing
+- **RAGEvaluationNode**: Comprehensive quality metrics (RAGAS)
+- **RAGBenchmarkNode**: Performance benchmarking
+- **TestDatasetGeneratorNode**: Synthetic test data generation
+
+### Privacy & Security
+- **PrivacyPreservingRAGNode**: Differential privacy and PII protection
+- **SecureMultiPartyRAGNode**: Federated computation without data sharing
+- **ComplianceRAGNode**: GDPR/HIPAA compliant RAG
+
+### Conversational RAG
+- **ConversationalRAGNode**: Multi-turn context management
+- **ConversationMemoryNode**: Long-term memory and personalization
+
+### Federated & Distributed
+- **FederatedRAGNode**: Distributed RAG across organizations
+- **EdgeRAGNode**: Optimized for edge devices
+- **CrossSiloRAGNode**: Cross-organizational data federation
+"""
+
+from .advanced import HyDENode, RAGFusionNode, SelfCorrectingRAGNode, StepBackRAGNode
+
+# Agentic RAG
+from .agentic import AgenticRAGNode, ReasoningRAGNode, ToolAugmentedRAGNode
+
+# Conversational RAG
+from .conversational import ConversationalRAGNode, ConversationMemoryNode
+
+# Evaluation Framework
+from .evaluation import RAGBenchmarkNode, RAGEvaluationNode, TestDatasetGeneratorNode
+
+# Federated RAG
+from .federated import CrossSiloRAGNode, EdgeRAGNode, FederatedRAGNode
+
+# Graph-based RAG
+from .graph import GraphBuilderNode, GraphQueryNode, GraphRAGNode
+
+# Multimodal RAG
+from .multimodal import (
+    ImageTextMatchingNode,
+    MultimodalRAGNode,
+    VisualQuestionAnsweringNode,
+)
+from .optimized import (
+    AsyncParallelRAGNode,
+    BatchOptimizedRAGNode,
+    CacheOptimizedRAGNode,
+    StreamingRAGNode,
+)
+
+# Privacy-preserving RAG
+from .privacy import (
+    ComplianceRAGNode,
+    PrivacyPreservingRAGNode,
+    SecureMultiPartyRAGNode,
+)
+from .query_processing import (
+    AdaptiveQueryProcessorNode,
+    MultiHopQueryPlannerNode,
+    QueryDecompositionNode,
+    QueryExpansionNode,
+    QueryIntentClassifierNode,
+    QueryRewritingNode,
+)
+
+# Real-time RAG
+from .realtime import (
+    IncrementalIndexNode,
+    RealtimeRAGNode,
+)
+from .realtime import (
+    StreamingRAGNode as RealtimeStreamingRAGNode,  # Avoid name conflict
+)
+from .registry import RAGWorkflowRegistry
+from .router import (
+    RAGPerformanceMonitorNode,
+    RAGQualityAnalyzerNode,
+    RAGStrategyRouterNode,
+)
+from .similarity import (
+    ColBERTRetrievalNode,
+    CrossEncoderRerankNode,
+    DenseRetrievalNode,
+    HybridFusionNode,
+    MultiVectorRetrievalNode,
+    PropositionBasedRetrievalNode,
+    SparseRetrievalNode,
+)
+from .strategies import (
+    HierarchicalRAGNode,
+    HybridRAGNode,
+    RAGConfig,
+    SemanticRAGNode,
+    StatisticalRAGNode,
+)
+from .workflows import (
+    AdaptiveRAGWorkflowNode,
+    AdvancedRAGWorkflowNode,
+    SimpleRAGWorkflowNode,
+)
+
+__all__ = [
+    # Core Strategy Nodes
+    "SemanticRAGNode",
+    "StatisticalRAGNode",
+    "HybridRAGNode",
+    "HierarchicalRAGNode",
+    "RAGConfig",
+    # Workflow Nodes
+    "SimpleRAGWorkflowNode",
+    "AdvancedRAGWorkflowNode",
+    "AdaptiveRAGWorkflowNode",
+    # Router & Analysis
+    "RAGStrategyRouterNode",
+    "RAGQualityAnalyzerNode",
+    "RAGPerformanceMonitorNode",
+    # Advanced RAG Techniques
+    "SelfCorrectingRAGNode",
+    "RAGFusionNode",
+    "HyDENode",
+    "StepBackRAGNode",
+    # Similarity Approaches
+    "DenseRetrievalNode",
+    "SparseRetrievalNode",
+    "ColBERTRetrievalNode",
+    "MultiVectorRetrievalNode",
+    "CrossEncoderRerankNode",
+    "HybridFusionNode",
+    "PropositionBasedRetrievalNode",
+    # Query Processing
+    "QueryExpansionNode",
+    "QueryDecompositionNode",
+    "QueryRewritingNode",
+    "QueryIntentClassifierNode",
+    "MultiHopQueryPlannerNode",
+    "AdaptiveQueryProcessorNode",
+    # Performance Optimization
+    "CacheOptimizedRAGNode",
+    "AsyncParallelRAGNode",
+    "StreamingRAGNode",
+    "BatchOptimizedRAGNode",
+    # Registry
+    "RAGWorkflowRegistry",
+    # Graph-based RAG
+    "GraphRAGNode",
+    "GraphBuilderNode",
+    "GraphQueryNode",
+    # Agentic RAG
+    "AgenticRAGNode",
+    "ToolAugmentedRAGNode",
+    "ReasoningRAGNode",
+    # Multimodal RAG
+    "MultimodalRAGNode",
+    "VisualQuestionAnsweringNode",
+    "ImageTextMatchingNode",
+    # Real-time RAG
+    "RealtimeRAGNode",
+    "RealtimeStreamingRAGNode",
+    "IncrementalIndexNode",
+    # Evaluation Framework
+    "RAGEvaluationNode",
+    "RAGBenchmarkNode",
+    "TestDatasetGeneratorNode",
+    # Privacy-preserving RAG
+    "PrivacyPreservingRAGNode",
+    "SecureMultiPartyRAGNode",
+    "ComplianceRAGNode",
+    # Conversational RAG
+    "ConversationalRAGNode",
+    "ConversationMemoryNode",
+    # Federated RAG
+    "FederatedRAGNode",
+    "EdgeRAGNode",
+    "CrossSiloRAGNode",
+]