SyedaArisha/predictive-maintenance-rag-system

1"""2Phase 2 Runner — FAISS Indexing, RAG Semantic Retrieval, LLM Explanation, and Scheduling3Author: Antigravity AI4Date: August 20265 6This script coordinates:71. Synthesizing logs and building/saving the local FAISS index82. Initializing the LLM Explainer and Log Retriever93. Mocking telemetry anomalies based on classification/RUL alerts104. Querying the FAISS index for past cases (RAG context)115. Generating natural language explanations for the production manager126. Executing scheduling reassignments to mitigate risk13"""14 15import os16import sys17import logging18 19# Ensure src/ is in the import path20BASE_DIR = os.path.abspath(os.path.dirname(__file__))21sys.path.insert(0, os.path.join(BASE_DIR, 'src'))22 23from utils import set_seeds, ProductionScheduler24from rag import build_and_save_index, LogRetriever, LLMExplainer25 26# Setup logging27logging.basicConfig(28    level=logging.INFO,29    format='%(asctime)s [%(levelname)s] %(message)s',30    handlers=[31        logging.StreamHandler()32    ]33)34logger = logging.getLogger(__name__)35 36 37def main():38    set_seeds(42)39    40    # Define directories according to clean architecture41    data_raw_dir = os.path.join(BASE_DIR, 'data', 'raw')42    data_processed_dir = os.path.join(BASE_DIR, 'data', 'processed')43    models_rag_dir = os.path.join(BASE_DIR, 'models', 'rag')44    reports_dir = os.path.join(BASE_DIR, 'reports')45    46    os.makedirs(data_raw_dir, exist_ok=True)47    os.makedirs(data_processed_dir, exist_ok=True)48    os.makedirs(models_rag_dir, exist_ok=True)49    os.makedirs(reports_dir, exist_ok=True)50    51    # Paths52    sch_path = os.path.join(data_raw_dir, 'production_schedule.csv')53    mach_path = os.path.join(data_raw_dir, 'PdM_machines.csv')54    adj_path = os.path.join(data_processed_dir, 'adjusted_schedule.csv')55    56    logger.info("=========================================")57    logger.info("STARTING PHASE 2: FAISS, LLM, AND OPERATION SCHEDULING")58    logger.info("=========================================")59    60    # ----------------------------------------------------61    # STEP 1: BUILD FAISS INDEX62    # ----------------------------------------------------63    logger.info("\n--- STEP 1: Building local FAISS vector database index ---")64    build_and_save_index(data_raw_dir, models_rag_dir)65    66    # ----------------------------------------------------67    # STEP 2: INITIALIZE RETRIEVER AND EXPLAINER68    # ----------------------------------------------------69    logger.info("\n--- STEP 2: Loading retriever and explainer modules ---")70    retriever = LogRetriever(models_rag_dir)71    72    # To run instantly without waiting to download the 270MB LLM locally,73    # set use_fallback=True. Set to False if you want to download and test SmolLM2-135M.74    use_fallback = True75    if use_fallback:76        logger.info("Using rule-based fallback explainer template for instant local test.")77    else:78        logger.info("Downloading and loading SmolLM2-135M locally...")79        80    explainer = LLMExplainer(force_fallback=use_fallback)81    82    # ----------------------------------------------------83    # STEP 3: SIMULATE ANOMALOUS TELEMETRY & RUN RAG84    # ----------------------------------------------------85    logger.info("\n--- STEP 3: Simulating machine anomaly alerts & executing RAG queries ---")86    87    critical_alerts = {88        3: {'failure_probability': 0.88, 'rul': 12.5},   # Machine 3 high failure risk, low RUL89        12: {'failure_probability': 0.94, 'rul': 8.0}     # Machine 12 high failure risk, low RUL90    }91    92    reports = {}93    for mid, alert in critical_alerts.items():94        query_str = f"Machine ID: {mid} component error telemetry anomalies"95        logger.info(f"Querying FAISS for: '{query_str}'")96        97        # Retrieve top 2 matching historical records98        hits = retriever.query(query_str, k=2)99        100        # Generate LLM explanation using retrieved records as context101        report = explainer.generate_explanation(102            machine_id=mid,103            failure_prob=alert['failure_probability'],104            rul=alert['rul'],105            historical_logs=hits106        )107        reports[mid] = report108        109        # Save report to reports folder110        report_path = os.path.join(reports_dir, f"machine_{mid}_explanation_report.md")111        with open(report_path, 'w', encoding='utf-8') as f:112            f.write(report)113            114        logger.info(f"Report for Machine {mid} saved to {report_path}")115        print("\n" + "="*50)116        print(f"EXPLANATION REPORT FOR MACHINE {mid}:")117        print("="*50)118        print(report)119        print("="*50 + "\n")120        121    # ----------------------------------------------------122    # STEP 4: RUN CLOSED-LOOP PRODUCTION SCHEDULER123    # ----------------------------------------------------124    logger.info("\n--- STEP 4: Running Production Scheduler to adjust operations ---")125    scheduler = ProductionScheduler(sch_path, mach_path)126    127    # Reroute jobs scheduled on at-risk machines (3 and 12) to healthy backups128    adjusted_schedule = scheduler.adjust_schedule(129        predictions=critical_alerts,130        output_adjusted_path=adj_path,131        failure_threshold=0.7,132        min_rul_hours=48.0133    )134    135    logger.info("=========================================")136    logger.info("PHASE 2 COMPLETED SUCCESSFULLY.")137    logger.info("=========================================")138 139 140if __name__ == '__main__':141    main()142