"""
SOTA Multi-Agentic RAG System

A production-grade RAG system following FAANG best practices:
- Query decomposition and planning
- Hybrid retrieval (dense + sparse)
- Cross-encoder reranking
- Grounded synthesis with citations
- Hallucination detection and self-correction
- LangGraph orchestration

Architecture:
    User Query
         |
    [QueryPlannerAgent] - Decomposes complex queries, identifies intent
         |
    [RetrieverAgent] - Hybrid search with query expansion
         |
    [RerankerAgent] - Cross-encoder scoring, filters low-quality
         |
    [SynthesizerAgent] - Generates grounded answer with citations
         |
    [CriticAgent] - Validates for hallucination, checks citations
         |
    (Loop back if critic fails)
         |
    Final Answer
"""

from .query_planner import QueryPlannerAgent, QueryPlan, SubQuery
from .retriever import RetrieverAgent, RetrievalResult, HybridSearchConfig
from .reranker import RerankerAgent, RankedResult, RerankerConfig
from .synthesizer import SynthesizerAgent, SynthesisResult, Citation
from .critic import CriticAgent, CriticResult, ValidationIssue
from .orchestrator import AgenticRAG, RAGConfig, RAGResponse

__all__ = [
    # Query Planner
    "QueryPlannerAgent",
    "QueryPlan",
    "SubQuery",
    # Retriever
    "RetrieverAgent",
    "RetrievalResult",
    "HybridSearchConfig",
    # Reranker
    "RerankerAgent",
    "RankedResult",
    "RerankerConfig",
    # Synthesizer
    "SynthesizerAgent",
    "SynthesisResult",
    "Citation",
    # Critic
    "CriticAgent",
    "CriticResult",
    "ValidationIssue",
    # Orchestrator
    "AgenticRAG",
    "RAGConfig",
    "RAGResponse",
]