rag_utils.py

from langchain.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import FAISS
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.chains import RetrievalQA
from langchain.prompts import PromptTemplate
from langchain.llms import HuggingFaceHub
import os
import logging

# Configure logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')

class RAGSystem:
    def __init__(self):
        try:
            # Initialize embeddings
            self.embeddings = HuggingFaceEmbeddings(
                model_name="sentence-transformers/all-mpnet-base-v2"
            )
            self.vector_store = None
            self.text_splitter = RecursiveCharacterTextSplitter(
                chunk_size=500,
                chunk_overlap=50
            )
            # Initialize HuggingFace model for text generation
            self.llm = HuggingFaceHub(
                repo_id="google/flan-t5-large",
                task="text-generation",
                model_kwargs={"temperature": 0.7, "max_length": 512}
            )
            logging.info("RAG system initialized successfully.")
        except Exception as e:
            logging.error(f"Failed to initialize RAG system: {str(e)}")
            raise e

    def initialize_knowledge_base(self, knowledge_base):
        """Initialize vector store with enhanced construction knowledge"""
        try:
            documents = []
            # Validate knowledge base
            self._validate_knowledge_base(knowledge_base)
            
            # Generate insights and case studies
            expert_insights = self._generate_expert_insights(knowledge_base)
            case_studies = self._generate_case_studies()
            
            for damage_type, cases in knowledge_base.items():
                for idx, case in enumerate(cases):
                    try:
                        # Combine insights into document text
                        relevant_insight = expert_insights.get(damage_type, "")
                        relevant_cases = case_studies.get(damage_type, "")
                        
                        doc_text = f"""
                        Damage Type: {damage_type}
                        Severity: {case['severity']}
                        Description: {case['description']}
                        Technical Details: {case['description']}
                        Expert Insight: {relevant_insight}
                        Case Studies: {relevant_cases}
                        Repair Methods: {', '.join(case['repair_method'])}
                        Cost Considerations: {case['estimated_cost']}
                        Implementation Timeline: {case['timeframe']}
                        Location Specifics: {case['location']}
                        Required Expertise Level: {case['required_expertise']}
                        Emergency Protocol: {case['immediate_action']}
                        Preventive Measures: {case['prevention']}
                        """
                        documents.append(doc_text)
                    except KeyError as e:
                        logging.warning(f"Missing key {str(e)} in {damage_type}, case {idx + 1}. Skipping.")
            
            if not documents:
                raise ValueError("No valid documents to process.")
            
            splits = self.text_splitter.create_documents(documents)
            self.vector_store = FAISS.from_documents(splits, self.embeddings)
            
            # Initialize QA chain
            self.qa_chain = RetrievalQA.from_chain_type(
                llm=self.llm,
                chain_type="stuff",
                retriever=self.vector_store.as_retriever(),
                chain_type_kwargs={
                    "prompt": self._get_qa_prompt()
                }
            )
            logging.info("Knowledge base initialized successfully.")
        except Exception as e:
            logging.error(f"Failed to initialize knowledge base: {str(e)}")
            raise e

    def _validate_knowledge_base(self, knowledge_base):
        """Validate the structure of the knowledge base."""
        required_keys = ['severity', 'description', 'repair_method', 'estimated_cost', 'timeframe', 'location', 'required_expertise', 'immediate_action', 'prevention']
        for damage_type, cases in knowledge_base.items():
            for idx, case in enumerate(cases):
                for key in required_keys:
                    if key not in case:
                        logging.error(f"Missing required field '{key}' in {damage_type}, case {idx + 1}")
                        raise ValueError(f"Missing required field '{key}' in {damage_type}, case {idx + 1}")
        logging.info("Knowledge base validation passed.")

    def _get_qa_prompt(self):
        """Create a custom prompt template for the QA chain"""
        template = """
        Context: {context}
        
        Question: {question}
        
        Provide a detailed analysis considering:
        1. Technical aspects
        2. Safety implications
        3. Cost-benefit analysis
        4. Long-term considerations
        5. Best practices and recommendations
        
        Answer:
        """
        return PromptTemplate(
            template=template,
            input_variables=["context", "question"]
        )

    def get_enhanced_analysis(self, damage_type, confidence, custom_query=None):
        """Get enhanced analysis with dynamic content generation"""
        try:
            if not self.vector_store:
                raise ValueError("Vector store is not initialized.")

            if not custom_query:
                base_query = f"""
                Provide a comprehensive analysis for {damage_type} damage with {confidence}% confidence level.
                Include technical assessment, safety implications, and expert recommendations.
                """
            else:
                base_query = custom_query
            
            # Get relevant documents
            results = self.qa_chain.run(base_query)
            if not results:
                logging.warning("No results returned for query.")
                return {"technical_details": [], "safety_considerations": [], "expert_recommendations": []}
            
            # Process and categorize the response
            enhanced_info = {
                "technical_details": self._extract_technical_details(results, damage_type),
                "safety_considerations": self._extract_safety_considerations(results),
                "expert_recommendations": self._extract_recommendations(results, confidence)
            }
            return enhanced_info
        except Exception as e:
            logging.error(f"Failed to generate enhanced analysis: {str(e)}")
            return {"technical_details": [], "safety_considerations": [], "expert_recommendations": []}