agents.rag.simple.enhanced_v3.agent¶

SimpleRAG V3 - Enhanced MultiAgent Implementation.

This module implements SimpleRAG using Enhanced MultiAgent V3 with the pattern: SimpleRAG = EnhancedMultiAgent[RetrieverAgent, SimpleAnswerAgent]

The implementation provides: - Type-safe agent composition - Performance tracking and optimization - Debug support and monitoring - Adaptive routing capabilities - Comprehensive state management

Examples

Basic usage:

rag = SimpleRAGV3.from_documents(
    documents=documents,
    embedding_config=embedding_config,
    performance_mode=True
)

result = await rag.arun("What is machine learning?")

Advanced usage with monitoring:

rag = SimpleRAGV3(
    name="qa_system",
    vector_store_config=vector_store_config,
    performance_mode=True,
    debug_mode=True,
    adaptation_rate=0.2
)

result = await rag.arun("Complex query")
analysis = rag.analyze_agent_performance()

Classes¶

SimpleRAGV3

SimpleRAG V3 - Enhanced MultiAgent implementation.

Module Contents¶

class agents.rag.simple.enhanced_v3.agent.SimpleRAGV3¶

Bases: haive.agents.multi.enhanced_multi_agent_v3.EnhancedMultiAgent[RAGAgentCollection]

SimpleRAG V3 - Enhanced MultiAgent implementation.

This class implements SimpleRAG using Enhanced MultiAgent V3 with the pattern: SimpleRAGV3 = EnhancedMultiAgent[RetrieverAgent, SimpleAnswerAgent]

The sequential execution flow is: 1. RetrieverAgent: Retrieves relevant documents from vector store 2. SimpleAnswerAgent: Generates answer using retrieved documents

Key Features:

Type-safe agent composition using Enhanced MultiAgent V3
Performance tracking and adaptive optimization
Debug support with comprehensive monitoring
Automatic state management and transfer
Factory methods for easy creation
Backward compatibility with existing SimpleRAG

State Management:

Uses SimpleRAGState when enhanced features are enabled, falls back to EnhancedMultiAgentState for basic usage.

Examples

From documents:

rag = SimpleRAGV3.from_documents(
    documents=my_documents,
    embedding_config=embedding_config,
    performance_mode=True,
    debug_mode=True
)

result = await rag.arun("What is machine learning?")

From vector store:

rag = SimpleRAGV3.from_vectorstore(
    vector_store_config=vs_config,
    llm_config=AugLLMConfig(temperature=0.7),
    performance_mode=True
)

With structured output:

class QAResponse(BaseModel):
    answer: str
    sources: List[str]
    confidence: float

rag = SimpleRAGV3(
    name="structured_rag",
    vector_store_config=vs_config,
    structured_output_model=QAResponse,
    performance_mode=True
)

async arun(input_data, debug=False, **kwargs)¶

Execute RAG pipeline using Enhanced MultiAgent V3 sequential execution.

This leverages the Enhanced MultiAgent V3 infrastructure for: - Performance tracking and optimization - Debug support and monitoring - Adaptive routing capabilities - Comprehensive state management

Parameters:

input_data (str | dict[str, Any]) – Query string or structured input with ‘query’ field
debug (bool) – Enable debug logging and detailed output
**kwargs – Additional execution parameters

Returns:

Generated response from the answer generation agent

Raises:

ValueError – If input validation fails
RuntimeError – If pipeline execution fails

Return type:

Any

classmethod ensure_agents_is_list(values)¶

Ensure agents field starts as an empty list for our List type.

Parameters:: values (dict)
Return type:: dict

classmethod from_documents(documents, embedding_config, llm_config=None, name=None, **kwargs)¶

Create SimpleRAG V3 from a list of documents.

Parameters:

documents (list[langchain_core.documents.Document]) – List of documents to create vector store from
embedding_config (Any) – Embedding configuration for vector store
llm_config (haive.core.engine.aug_llm.AugLLMConfig | None) – LLM configuration for answer generation
name (str | None) – Name for the RAG system
**kwargs – Additional configuration parameters

Returns:

Configured SimpleRAGV3 instance

Return type:

SimpleRAGV3

Examples

Basic usage:

rag = SimpleRAGV3.from_documents(
    documents=my_documents,
    embedding_config=embedding_config
)

With enhanced features:

rag = SimpleRAGV3.from_documents(
    documents=my_documents,
    embedding_config=embedding_config,
    llm_config=AugLLMConfig(temperature=0.3),
    performance_mode=True,
    debug_mode=True,
    top_k=10
)

classmethod from_vectorstore(vector_store_config, llm_config=None, name=None, **kwargs)¶

Create SimpleRAG V3 from existing vector store configuration.

Parameters:

vector_store_config (haive.core.engine.vectorstore.VectorStoreConfig) – Vector store configuration
llm_config (haive.core.engine.aug_llm.AugLLMConfig | None) – LLM configuration for answer generation
name (str | None) – Name for the RAG system
**kwargs – Additional configuration parameters

Returns:

Configured SimpleRAGV3 instance

Return type:

SimpleRAGV3

Examples

Basic usage:

rag = SimpleRAGV3.from_vectorstore(
    vector_store_config=vs_config,
    llm_config=AugLLMConfig()
)

With monitoring:

rag = SimpleRAGV3.from_vectorstore(
    vector_store_config=vs_config,
    llm_config=AugLLMConfig(temperature=0.5),
    performance_mode=True,
    adaptation_rate=0.2
)

async generate_answer(query, documents, **kwargs)¶

Generate answer using the answer generation agent.

Parameters:

query (str) – Original query
documents (list[langchain_core.documents.Document]) – Retrieved documents for context
**kwargs – Additional generation parameters

Returns:

Generated answer (format depends on structured_output_model)

Return type:

Any

get_answer_agent()¶

Get the answer generation agent.

Return type:: agents.rag.simple.enhanced_v3.answer_generator_agent.SimpleAnswerAgent

get_rag_info()¶

Get comprehensive information about the RAG configuration.

Return type:: dict[str, Any]

get_retriever_agent()¶

Get the retriever agent.

Return type:: agents.rag.simple.enhanced_v3.retriever_agent.RetrieverAgent

async retrieve_documents(query, k=None, score_threshold=None, **kwargs)¶

Retrieve documents using the retriever agent.

Parameters:

query (str) – Query string for retrieval
k (int | None) – Number of documents to retrieve (defaults to self.top_k)
score_threshold (float | None) – Minimum similarity score (defaults to self.similarity_threshold)
**kwargs – Additional retrieval parameters

Returns:

Retrieval result with documents and metadata

Return type:

dict[str, Any]

setup_rag_pipeline()¶

Setup the RAG pipeline with RetrieverAgent and SimpleAnswerAgent.

Return type:: SimpleRAGV3

classmethod validate_citation_style(v)¶

Validate citation style.

Parameters:: v (str)
Return type:: str