v1 completed successfully

.dockerignore

@@ -0,0 +1,15 @@
+node_modules
+dist
+venv
+.env
+*.db
+.git
+.gitignore
+__pycache__
+*.pyc
+.pytest_cache
+.vscode
+.idea
+frontend/node_modules
+frontend/dist
+backend/__pycache__

.env.example

@@ -0,0 +1,11 @@
+# NVIDIA NIM API Key (Required for AI Advisor)
+NVIDIA_API_KEY=your_nvapi_key_here
+
+# Hugging Face Token (Required for Persistent DB Backups)
+HF_TOKEN=your_hf_token_here
+
+# Database URL (Default for Docker/HF)
+# SQLALCHEMY_DATABASE_URL=sqlite:////app/data/loan_db.db
+
+# Deployment environment (production/development)
+ENV=production

.gitignore

@@ -0,0 +1,26 @@
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+venv/
+.env
+
+# Database (Local)
+*.db
+!data/*.csv
+
+# Node / Frontend
+node_modules/
+dist/
+build/
+.next/
+
+# IDE / Editor
+.vscode/
+.idea/
+*.swp
+*.swo
+.DS_Store
+
+# Docker Local Testing
+docker-compose.override.yml

Dockerfile

@@ -0,0 +1,45 @@
+# Stage 1: Build React Frontend
+FROM node:20-alpine AS build
+WORKDIR /app/frontend
+COPY frontend/package*.json ./
+RUN npm install
+COPY frontend .
+RUN npm run build
+
+# Stage 2: Serve with FastAPI (Hugging Face setup)
+FROM python:3.11-slim
+
+# Create a non-root user matching Hugging Face defaults (1000)
+RUN useradd -m -u 1000 user
+WORKDIR /app
+
+# System dependencies for scientific packages
+RUN apt-get update && apt-get install -y gcc g++ \
+    && rm -rf /var/lib/apt/lists/*
+
+# Copy project metadata and backend source for package installation
+COPY pyproject.toml .
+COPY backend ./backend/
+RUN pip install --no-cache-dir .
+
+# Create necessary directories and set ownership
+RUN mkdir -p /app/backend /data /app/frontend
+RUN chown -R user:user /app /data
+
+# Switch to the non-root user
+USER user
+
+# Copy pre-built frontend from stage 1
+COPY --from=build --chown=user:user /app/frontend/dist /app/frontend/dist
+
+# Copy legacy data for ML Engine
+COPY --chown=user:user data ./data/
+
+# Copy backend source
+COPY --chown=user:user backend ./backend/
+
+# Hugging Face deployment assumes we serve on 7860
+EXPOSE 7860
+
+# FastAPI startup
+CMD ["uvicorn", "backend.main:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -1,51 +1,124 @@
-# 🏦 Loan Approval Prediction System
+# 🏦 Enterprise Loan AI: Adaptive Predictive Ecosystem
 
-A Machine Learning-based web application that predicts whether a loan will be approved or rejected based on applicant details.
+**Enterprise-grade financial decision engine combining Hybrid ML Logic with Mistral Large 3 Generative Insights.**
 
 ---
 
-## 🚀 Features
+## 🏗️ 1. System Architecture
+
+The ecosystem operates as a high-fidelity diagnostic pipeline, ensuring mathematical rigor before AI interpretation.
+
+```mermaid
+graph TD
+    User((User)) -->|Submit Application| UI(React/Vite Frontend)
+    UI -->|API POST /predict| API(FastAPI Backend)
+    
+    subgraph "Hybrid Inference Engine"
+        API -->|1. Deterministic Map| ML(sklearn Random Forest)
+        ML -->|Probabilities| ADVISOR(NVIDIA NIM Mistral-3)
+        ADVISOR -->|2. Structured Narrative| RESPONSE(Final JSON Packet)
+    end
+    
+    RESPONSE -->|Persistence| DB[(SQLite / PostgreSQL)]
+    RESPONSE -->|Visual Analytics| DASH(Radar & Distribution Charts)
+    DASH -->|Render| User
+```
 
-* Simple and user-friendly interface using Streamlit
-* Real-time loan approval prediction
-* Machine Learning model (Random Forest)
-* Data preprocessing and feature engineering
-* Displays prediction with confidence score
+---
+
+## 📂 2. Repository Structure
+
+```text
+├── backend/              # FastAPI Application Source
+│   ├── db/               # Persistence & Data Models
+│   ├── logic/            # Deterministic ML Engine (sklearn)
+│   ├── services/         # LLM Advisor (NVIDIA NIM Integration)
+│   ├── main.py           # API Entry Point & Startup Logic
+│   └── requirements.txt  # Python Dependencies
+├── data/                 # Training datasets (CSV)
+├── frontend/             # React (Vite) Application
+│   ├── src/
+│   │   ├── components/   # Modular UI Components (Charts, Form, Sidebar)
+│   │   ├── App.jsx       # State-Based Navigation & Layout
+│   │   └── index.css     # Premium UI Design System
+│   └── package.json      # Node.js Dependencies
+├── Dockerfile            # Multi-stage Production Build
+├── .dockerignore         # Docker Build Exclusions
+├── .gitignore            # Git Excluded Files
+└── .env.example          # Environment Management Template
+```
 
 ---
 
-## 🛠️ Tech Stack
+## 🚀 3. End-to-End Setup Guide
 
-* Python
-* Streamlit
-* Pandas
-* NumPy
-* Scikit-learn
+### **A. Local Development**
 
----
+#### **1. Backend (Python 3.11+)**
+```powershell
+# Create Virtual Environment
+python -m venv venv
+.\venv\Scripts\activate
 
-## ⚙️ How It Works
+# Install Project as Editable Package
+pip install -e .
 
-1. User enters loan details (income, loan amount, etc.)
-2. Data is processed and transformed
-3. Machine Learning model makes prediction
-4. Result is displayed as Approved or Rejected with confidence
+# Start Server
+uvicorn backend.main:app --reload --port 8000
+```
 
----
+#### **2. Frontend (Node.js 20+)**
+```powershell
+cd frontend
+npm install
+npm run dev
+```
+
+### **B. Environment Configuration**
+Create a `.env` file in the root directory:
+```env
+# Required for AI Analysis
+NVIDIA_API_KEY=your_key_here
 
-## 🌐 Future Scope
+# Required for Persistent Cloud Backups (Optional)
+HF_TOKEN=your_huggingface_write_token
+HF_REPO_ID=your_username/your_space_name
+```
 
-* Convert into full-stack application (React + Flask)
-* Add database integration
-* Deploy on cloud platform
-* Improve model accuracy
+### **C. Cloud Synchronization (New)**
+The system now features an **Automated Cloud Sync** service. 
+- When running in a Docker/Hugging Face environment with a valid `HF_TOKEN`, the system will automatically back up your assessment history to your Space's repository every time you make a new prediction or clear the history.
+- This ensures your data persists even if the Space's ephemeral container is restarted.
 
 ---
 
-## 📌 Conclusion
+## 🐳 4. Production Deployment (Hugging Face)
 
-This project demonstrates how Machine Learning can be used to automate and improve the loan approval process, making it faster and more efficient.
+This project is optimized for deployment as a **Hugging Face Space** using the **Docker** runtime.
+
+### **1. Build & Run Locally**
+```powershell
+# Build Image
+docker build -t loan-prediction-app .
+
+# Run Container (History persistence requires /app/data volume)
+docker run -d -p 7860:7860 --name loan-app loan-prediction-app
+```
+
+### **2. Deploy to Hugging Face**
+1. Create a new Space on [huggingface.co](https://huggingface.co/spaces) selecting the **Docker** SDK.
+2. In the Space **Settings**:
+   - Add your `NVIDIA_API_KEY` as a **Secret**.
+   - (Optional) Enable **Persistent Storage** and mount to `/app/data`.
+3. Push your code. The Space will automatically build and launch your dashboard.
 
 ---
 
-⭐ If you like this project, feel free to give it a star!
+## ✅ 5. Platform Features
+- **Deterministic Math**: Validated Random Forest scoring.
+- **AI Narrative Sub-Cards**: Readable, point-by-point financial insights.
+- **Radar Comparisons**: Real-time benchmarking against successful profiles.
+- **ChatGPT History Sidebar**: Persistent task tracking with "Clear History" support.
+
+---
+*Built for High-Trust Lending Environments.*

backend/db/database.py

@@ -0,0 +1,29 @@
+from sqlalchemy import create_engine
+from sqlalchemy.orm import sessionmaker, declarative_base
+
+import os
+
+# Persistent path for Docker/Hugging Face Spaces (Matches Bucket Mount)
+DATA_DIR = "/data"
+DB_PATH = os.path.join(DATA_DIR, "loan_db.db")
+
+# Fallback to local root if not in Docker environment
+if not os.path.exists(DATA_DIR):
+    DB_PATH = "./loan_db_v3.db"
+
+SQLALCHEMY_DATABASE_URL = f"sqlite:///{DB_PATH}"
+
+engine = create_engine(
+    SQLALCHEMY_DATABASE_URL, connect_args={"check_same_thread": False}
+)
+SessionLocal = sessionmaker(autocommit=False, autoflush=False, bind=engine)
+
+Base = declarative_base()
+
+# Dependency
+def get_db():
+    db = SessionLocal()
+    try:
+        yield db
+    finally:
+        db.close()

backend/db/models.py

@@ -0,0 +1,35 @@
+from sqlalchemy import Column, Integer, String, Float, DateTime, Text
+from sqlalchemy.sql import func
+from .database import Base
+
+class LoanApplication(Base):
+    __tablename__ = "loan_applications"
+
+    id = Column(Integer, primary_key=True, index=True)
+    applicant_name = Column(String, default="Applicant")
+    
+    # Financial Inputs (Stored as stringified JSON or separate columns, but columns are safer for SQLite)
+    gender = Column(String)
+    married = Column(String)
+    dependents = Column(String)
+    education = Column(String)
+    self_employed = Column(String)
+    applicant_income = Column(Float)
+    coapplicant_income = Column(Float)
+    loan_amount = Column(Float)
+    loan_amount_term = Column(Float)
+    credit_history = Column(Float)
+    property_area = Column(String)
+    
+    # Processed Results
+    prediction = Column(String) # "Y" or "N"
+    confidence = Column(Float)
+    dti_ratio = Column(Float)
+    
+    # AI & Explanation Outputs
+    explanation_text = Column(Text, nullable=True)
+    optimized_suggestion = Column(Text, nullable=True)
+    feature_importance_json = Column(Text, nullable=True) # stringified JSON
+    benchmarks_json = Column(Text, nullable=True) # stringified JSON
+    
+    created_at = Column(DateTime(timezone=True), server_default=func.now())

backend/db/schemas.py

@@ -0,0 +1,34 @@
+from pydantic import BaseModel, Field
+from typing import Optional
+from datetime import datetime
+
+class LoanApplicationBase(BaseModel):
+    applicant_name: Optional[str] = "Applicant"
+    gender: str = Field(..., description="Male or Female")
+    married: str = Field(..., description="Yes or No")
+    dependents: str = Field(..., description="0, 1, 2, 3+")
+    education: str = Field(..., description="Graduate or Not Graduate")
+    self_employed: str = Field(..., description="Yes or No")
+    applicant_income: float = Field(..., description="Monthly Applicant Income")
+    coapplicant_income: float = Field(..., description="Monthly Co-applicant Income")
+    loan_amount: float = Field(..., description="Total Loan AmountRequested")
+    loan_amount_term: float = Field(..., description="Loan Term in Months")
+    credit_history: float = Field(..., description="1.0 for Yes, 0.0 for No")
+    property_area: str = Field(..., description="Urban, Semiurban, Rural")
+
+class LoanApplicationCreate(LoanApplicationBase):
+    pass
+
+class LoanApplicationResponse(LoanApplicationBase):
+    id: int
+    prediction: str
+    confidence: float
+    dti_ratio: float
+    explanation_text: Optional[str] = None
+    optimized_suggestion: Optional[str] = None
+    feature_importance_json: Optional[str] = None
+    benchmarks_json: Optional[str] = None
+    created_at: datetime
+
+    class Config:
+        from_attributes = True

backend/logic/deterministic.py

@@ -0,0 +1,109 @@
+import copy
+from .ml_model import ml_engine
+import json
+
+KEY_MAP = {
+    "gender": "Gender",
+    "married": "Married",
+    "dependents": "Dependents",
+    "education": "Education",
+    "self_employed": "Self_Employed",
+    "applicant_income": "ApplicantIncome",
+    "coapplicant_income": "CoapplicantIncome",
+    "loan_amount": "LoanAmount",
+    "loan_amount_term": "Loan_Amount_Term",
+    "credit_history": "Credit_History",
+    "property_area": "Property_Area"
+}
+
+def map_keys(data: dict) -> dict:
+    return {KEY_MAP.get(k, k): v for k, v in data.items()}
+
+def prepare_input(base_data: dict) -> dict:
+    # Compute derived features for the ML Engine
+    total_income = base_data["ApplicantIncome"] + base_data["CoapplicantIncome"]
+    
+    # Avoid division by zero
+    term = base_data["Loan_Amount_Term"] if base_data["Loan_Amount_Term"] > 0 else 360
+    
+    emi = (base_data["LoanAmount"] * 1000) / term
+    balance_income = total_income - emi
+
+    base_data["Total_Income"] = total_income
+    base_data["EMI"] = emi
+    base_data["Balance_Income"] = balance_income
+    return base_data
+
+def process_pipeline(input_data: dict) -> dict:
+    """Evaluates the loan application and runs sensitivity analysis if rejected."""
+    
+    # 0. Map snake_case API keys to PascalCase Model keys
+    mapped_input = map_keys(input_data)
+    
+    # 1. Base Evaluation
+    processed_input = prepare_input(copy.deepcopy(mapped_input))
+    
+    result, confidence = ml_engine.predict(processed_input)
+    
+    dti_ratio = 0.0
+    if processed_input["Total_Income"] > 0:
+        dti_ratio = processed_input["EMI"] / processed_input["Total_Income"]
+    dti_ratio_pct = min(100.0, dti_ratio * 100)
+
+    response = {
+        "prediction": result,
+        "confidence": confidence,
+        "dti_ratio": dti_ratio_pct,
+        "optimized_suggestion": "Your profile meets all current thresholds. No optimization necessary.",
+        "feature_importances": ml_engine.get_feature_importances(),
+        "benchmarks": ml_engine.get_benchmarks()
+    }
+
+    # 2. What-If Optimization (If Rejected)
+    if result == "N":
+        # Simulate lower loan amount
+        sim_data_loan = copy.deepcopy(mapped_input)
+        loan_decrease_needed = 0
+        loan_approved = False
+        
+        while sim_data_loan["LoanAmount"] > 10 and not loan_approved:
+            sim_data_loan["LoanAmount"] -= 10 # Decrease by 10k
+            loan_decrease_needed += 10
+            
+            p_input = prepare_input(copy.deepcopy(sim_data_loan))
+            sim_res, sim_conf = ml_engine.predict(p_input)
+            if sim_res == "Y":
+                loan_approved = True
+                break
+                
+        # Simulate higher coapplicant income
+        sim_data_inc = copy.deepcopy(mapped_input)
+        inc_increase_needed = 0
+        inc_approved = False
+        
+        while inc_increase_needed < 20000 and not inc_approved:
+            sim_data_inc["CoapplicantIncome"] += 1000 # Increase by 1k
+            inc_increase_needed += 1000
+            
+            p_input_inc = prepare_input(copy.deepcopy(sim_data_inc))
+            sim_res_inc, sim_conf_inc = ml_engine.predict(p_input_inc)
+            if sim_res_inc == "Y":
+                inc_approved = True
+                break
+
+        # Generate rule-based optimization text (Determnistic)
+        suggestion = "Unfortunately, we couldn't find a minor adjustment to approve your loan. Consider improving your CIBIL score."
+        
+        if loan_approved and inc_approved:
+            suggestion = f"Reducing your loan request by ₹{loan_decrease_needed * 1000} OR adding a co-applicant income of ~₹{inc_increase_needed} would likely result in approval."
+        elif loan_approved:
+            suggestion = f"Reducing your loan request by ₹{loan_decrease_needed * 1000} would likely result in an approval."
+        elif inc_approved:
+            suggestion = f"Adding a co-applicant income of ~₹{inc_increase_needed} would likely result in an approval."
+            
+        if mapped_input["Credit_History"] == 0:
+             suggestion = "Your Credit History is 0. This is the primary blocking factor. Improving your credit standing is required before other adjustments will work."
+
+        response["optimized_suggestion"] = suggestion
+
+    return response

backend/logic/ml_model.py

@@ -0,0 +1,106 @@
+import os
+import pandas as pd
+import numpy as np
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.preprocessing import LabelEncoder
+import json
+
+class MLEngine:
+    def __init__(self):
+        self.model = None
+        self.encoders = {}
+        self.target_encoder = None
+        self.feature_columns = []
+        self.train_model()
+
+    def load_data(self):
+        path = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "..", "data", "train.csv")
+        try:
+            df = pd.read_csv(path)
+            for col in df.select_dtypes(include=['int64', 'float64']).columns:
+                df[col] = df[col].fillna(df[col].mean())
+            for col in df.select_dtypes(include=['object']).columns:
+                df[col] = df[col].fillna(df[col].mode()[0])
+            return df
+        except Exception as e:
+            print(f"Error loading {path}: {e}")
+            return None
+
+    def feature_engineering(self, df):
+        if "Loan_ID" in df.columns:
+            df = df.drop("Loan_ID", axis=1)
+        df["Total_Income"] = df["ApplicantIncome"] + df["CoapplicantIncome"]
+        df["EMI"] = (df["LoanAmount"] * 1000) / df["Loan_Amount_Term"]
+        df["Balance_Income"] = df["Total_Income"] - df["EMI"]
+        return df
+
+    def encode_data(self, df):
+        encoders = {}
+        target_encoder = LabelEncoder()
+        
+        # Avoid SettingWithCopyWarning by operating on frame directly if needed, but it's okay here
+        df["Loan_Status"] = target_encoder.fit_transform(df["Loan_Status"])
+        
+        cols = ["Gender", "Married", "Dependents", "Education", "Self_Employed", "Property_Area"]
+        for col in cols:
+            le = LabelEncoder()
+            df[col] = le.fit_transform(df[col].astype(str))
+            encoders[col] = le
+            
+        return df, encoders, target_encoder
+
+    def train_model(self):
+        print("Training Random Forest model on boot...")
+        df = self.load_data()
+        if df is not None:
+            df = self.feature_engineering(df)
+            df, self.encoders, self.target_encoder = self.encode_data(df)
+            
+            X = df.drop("Loan_Status", axis=1)
+            y = df["Loan_Status"]
+            self.feature_columns = X.columns.tolist()
+            
+            self.model = RandomForestClassifier(n_estimators=200, max_depth=10, random_state=42)
+            self.model.fit(X, y)
+            
+            # Calculate Benchmarks for 'Approved' (Status 1 if transformed or 'Y' if checking pre-encoded)
+            # Find the index of 'Y' in target_encoder
+            y_idx = list(self.target_encoder.classes_).index('Y')
+            approved_df = df[df['Loan_Status'] == y_idx]
+            self.benchmarks = approved_df[self.feature_columns].mean().to_dict()
+            
+            print("Model and Benchmarks generated successfully.")
+
+    def get_benchmarks(self):
+        return {k: float(v) for k, v in getattr(self, 'benchmarks', {}).items()}
+            
+    def get_feature_importances(self):
+        if not self.model: return {}
+        importances = self.model.feature_importances_
+        # Convert numpy types to standard python floats for JSON serialization
+        return {k: float(v) for k, v in zip(self.feature_columns, importances)}
+
+    def predict(self, input_data: dict):
+        if not self.model:
+            raise Exception("Model is not trained.")
+            
+        df = pd.DataFrame([input_data])
+        for col in self.feature_columns:
+            if col not in df.columns:
+                df[col] = 0
+                
+        df = df[self.feature_columns]
+        
+        for col, le in self.encoders.items():
+            if col in df.columns:
+                df[col] = le.transform(df[col].astype(str))
+                
+        pred = self.model.predict(df)
+        prob = self.model.predict_proba(df)
+        
+        result = self.target_encoder.inverse_transform(pred)[0]
+        confidence = float(np.max(prob))
+        
+        return result, confidence
+
+ml_engine = MLEngine()

backend/main.py

@@ -0,0 +1,98 @@
+from fastapi import FastAPI, Depends, HTTPException, BackgroundTasks
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.staticfiles import StaticFiles
+from fastapi.responses import FileResponse
+from sqlalchemy.orm import Session
+from sqlalchemy import desc
+import json
+import os
+
+from backend.db import database, models, schemas
+from backend.logic.deterministic import process_pipeline
+from backend.services.llm_advisor import generate_explanation
+from backend.services.hf_sync import sync_db_to_hf
+
+# Create tables
+models.Base.metadata.create_all(bind=database.engine)
+
+app = FastAPI(title="Loan Prediction System API")
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"], # Allow all for demo
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+@app.post("/api/predict", response_model=schemas.LoanApplicationResponse)
+def create_prediction(app_in: schemas.LoanApplicationCreate, background_tasks: BackgroundTasks, db: Session = Depends(database.get_db)):
+    # 1. Convert to dictionary
+    input_data = app_in.model_dump()
+    
+    # 2. Run Deterministic Layer (Prediction + Wait-If Simulation)
+    packet = process_pipeline(input_data)
+    
+    # 3. Pass Packet to LLM Layer
+    explanation = generate_explanation(packet)
+    
+    # 4. Save to DB
+    db_record = models.LoanApplication(
+        applicant_name=app_in.applicant_name,
+        gender=app_in.gender,
+        married=app_in.married,
+        dependents=app_in.dependents,
+        education=app_in.education,
+        self_employed=app_in.self_employed,
+        applicant_income=app_in.applicant_income,
+        coapplicant_income=app_in.coapplicant_income,
+        loan_amount=app_in.loan_amount,
+        loan_amount_term=app_in.loan_amount_term,
+        credit_history=app_in.credit_history,
+        property_area=app_in.property_area,
+        
+        prediction=packet["prediction"],
+        confidence=packet["confidence"],
+        dti_ratio=packet["dti_ratio"],
+        
+        explanation_text=explanation,
+        optimized_suggestion=packet["optimized_suggestion"],
+        feature_importance_json=json.dumps(packet["feature_importances"]),
+        benchmarks_json=json.dumps(packet["benchmarks"])
+    )
+    
+    db.add(db_record)
+    db.commit()
+    db.refresh(db_record)
+    
+    # Trigger HF Backup in background
+    background_tasks.add_task(sync_db_to_hf)
+    
+    return db_record
+
+@app.get("/api/history", response_model=list[schemas.LoanApplicationResponse])
+def get_history(limit: int = 10, db: Session = Depends(database.get_db)):
+    records = db.query(models.LoanApplication).order_by(desc(models.LoanApplication.created_at)).limit(limit).all()
+    return records
+
+@app.delete("/api/history")
+def clear_history(background_tasks: BackgroundTasks, db: Session = Depends(database.get_db)):
+    db.query(models.LoanApplication).delete()
+    db.commit()
+    
+    # Trigger HF Backup in background
+    background_tasks.add_task(sync_db_to_hf)
+    
+    return {"message": "History cleared successfully"}
+
+# Serve Frontend
+frontend_build_path = os.path.join(os.path.dirname(__file__), "..", "frontend", "dist")
+if os.path.isdir(frontend_build_path):
+    app.mount("/assets", StaticFiles(directory=os.path.join(frontend_build_path, "assets")), name="assets")
+    
+    @app.get("/{full_path:path}")
+    def catch_all(full_path: str):
+        index_path = os.path.join(frontend_build_path, "index.html")
+        if os.path.isfile(index_path):
+            return FileResponse(index_path)
+        return {"error": "Frontend build not found."}

backend/requirements.txt

@@ -0,0 +1,10 @@
+fastapi>=0.100.0
+uvicorn>=0.23.0
+pydantic>=2.0.0
+sqlalchemy>=2.0.0
+openai>=1.0.0
+scikit-learn>=1.2.0
+pandas>=2.0.0
+numpy>=1.24.0
+python-dotenv>=1.0.0
+huggingface_hub>=0.19.0

backend/services/hf_sync.py

@@ -0,0 +1,49 @@
+import os
+from huggingface_hub import HfApi
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def sync_db_to_hf():
+    """
+    Syncs the local SQLite database to a Hugging Face repository.
+    This is useful for persisting data on ephemeral Hugging Face Spaces.
+    """
+    token = os.getenv("HF_TOKEN")
+    repo_id = os.getenv("HF_REPO_ID") # e.g., "username/space-name"
+    
+    # Hugging Face Spaces automatically provide SPACE_ID
+    if not repo_id:
+        repo_id = os.getenv("SPACE_ID")
+        
+    if not token or not repo_id:
+        print("HF_Sync: Missing HF_TOKEN or HF_REPO_ID. Skipping sync.")
+        return False
+
+    # Get the db path from database.py logic
+    # We'll just use the constant path if it exists
+    db_path = "/app/data/loan_db.db"
+    
+    # Fallback to local if not in docker
+    if not os.path.exists("/app/data"):
+        db_path = "./loan_db_v3.db"
+
+    if not os.path.exists(db_path):
+        print(f"HF_Sync: Database file not found at {db_path}. Skipping.")
+        return False
+
+    try:
+        api = HfApi(token=token)
+        print(f"HF_Sync: Uploading {db_path} to {repo_id}...")
+        
+        api.upload_file(
+            path_or_fileobj=db_path,
+            path_in_repo="data/loan_db_backup.db",
+            repo_id=repo_id,
+            repo_type="space" # Assuming it's a space, but could be 'dataset'
+        )
+        print("HF_Sync: Database successfully synced to Hugging Face Space.")
+        return True
+    except Exception as e:
+        print(f"HF_Sync: Error during synchronization: {e}")
+        return False

backend/services/llm_advisor.py

@@ -0,0 +1,94 @@
+import os
+from openai import OpenAI
+from dotenv import load_dotenv
+
+load_dotenv()
+
+SYSTEM_PROMPT = """
+You are a senior banking AI advisor for a high-end lending firm. 
+Your goal is to provide a very professional, encouraging, and data-driven explanation for a loan decision.
+
+Rules:
+1. Divide your explanation into exactly 3 or 4 clear, distinct points.
+2. Prefix EVERY point with the delimiter: @@POINT@@
+3. Keep each point concise (1-2 sentences).
+4. Do NOT use markdown headers like ###. Use **Bold** for emphasis.
+5. Do NOT invent new financial data. Only use what is provided in the packet.
+6. If the loan is rejected, ensure the points explain the 'Why' and refer to the specific 'Actionable Step' provided in the deterministic packet.
+
+Format:
+@@POINT@@ [First insight]
+@@POINT@@ [Second insight]
+...etc
+"""
+
+# Prioritized models for fallback logic
+LLM_MODELS = [
+    "mistralai/mistral-large-3-675b-instruct-2512",
+    "meta/llama-3.1-405b-instruct",
+    "mistralai/mixtral-8x7b-instruct-v0.1"
+]
+
+def generate_explanation(packet: dict) -> str:
+    api_key = os.getenv("NVIDIA_API_KEY")
+    
+    # Check if we should use Deterministic Mode
+    result_label = "Approved" if packet.get('prediction') == 'Y' else "Rejected"
+    confidence_val = packet.get('confidence', 0) * 100
+    dti_val = packet.get('dti_ratio', 0)
+    suggestion = packet.get('optimized_suggestion', '')
+
+    default_narrative = f"""
+### **AI Advisor Analysis (Deterministic Mode)**
+*Note: This is an automated diagnostic analysis.*
+
+Based on our core financial engine, your application has been **{result_label}** with a confidence score of **{confidence_val:.2f}%**. 
+
+**Key Insights:**
+- Your current **Debt-to-Income (DTI)** ratio is **{dti_val:.2f}%**. 
+- **Actionable Step:** {suggestion}
+"""
+
+    if not api_key or "YOUR_KEY" in api_key:
+        return default_narrative
+        
+    client = OpenAI(
+        base_url="https://integrate.api.nvidia.com/v1",
+        api_key=api_key
+    )
+
+    user_content = f"""
+    Result: {packet.get('prediction', 'Unknown')}
+    Confidence: {packet.get('confidence', 0):.2f}%
+    DTI Ratio: {packet.get('dti_ratio', 0):.2f}%
+    Primary Suggestion: {packet.get('optimized_suggestion', 'None')}
+    """
+
+    # Multi-model Fallback Loop
+    for model_name in LLM_MODELS:
+        try:
+            print(f"Attempting inference with model: {model_name}...")
+            completion = client.chat.completions.create(
+                model=model_name,
+                messages=[
+                    {"role": "system", "content": SYSTEM_PROMPT},
+                    {"role": "user", "content": user_content}
+                ],
+                temperature=0.15,
+                max_tokens=1024
+            )
+            
+            content = completion.choices[0].message.content
+            # Basic validation that model followed instructions
+            if "@@POINT@@" in content:
+                print(f"Success with model: {model_name}")
+                return content
+            else:
+                print(f"Model {model_name} failed to provide structured output. Trying next...")
+                
+        except Exception as e:
+            print(f"Error calling {model_name}: {e}. Retrying with fallback...")
+
+    # Final Fallback if all AI models fail
+    print("CRITICAL: All AI models failed. Returning deterministic narrative.")
+    return default_narrative

backend/tests/test_llm_fallback.py

@@ -0,0 +1,77 @@
+import os
+import sys
+from dotenv import load_dotenv
+from openai import OpenAI
+
+# Add project root to path
+sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+from services.llm_advisor import generate_explanation, LLM_MODELS, SYSTEM_PROMPT
+
+load_dotenv()
+
+def test_single_model(model_name, packet):
+    print(f"\n--- Testing Model: {model_name} ---")
+    api_key = os.getenv("NVIDIA_API_KEY")
+    client = OpenAI(base_url="https://integrate.api.nvidia.com/v1", api_key=api_key)
+    
+    user_content = f"Result: {packet['prediction']}, Confidence: {packet['confidence']}, DTI: {packet['dti_ratio']}"
+    
+    try:
+        completion = client.chat.completions.create(
+            model=model_name,
+            messages=[
+                {"role": "system", "content": SYSTEM_PROMPT},
+                {"role": "user", "content": user_content}
+            ],
+            temperature=0.1,
+            max_tokens=512
+        )
+        content = completion.choices[0].message.content
+        print(f"Response received ({len(content)} chars)")
+        if "@@POINT@@" in content:
+            print("SUCCESS: Output contains required @@POINT@@ delimiters.")
+        else:
+            print("FAILURE: Output missing @@POINT@@ delimiters.")
+        return True
+    except Exception as e:
+        print(f"ERROR: Model {model_name} failed: {e}")
+        return False
+
+def test_fallback_logic(packet):
+    print("\n--- Testing Full Fallback Workflow ---")
+    # Forcing a failure by temporarily corrupting the model list in memory
+    original_models = list(LLM_MODELS)
+    # Inject a fake non-existent model at the start
+    import services.llm_advisor as advisor
+    advisor.LLM_MODELS = ["invalid/non-existent-model-1234"] + original_models
+    
+    print("Injected invalid model. Expecting fail-over to the first valid model...")
+    response = advisor.generate_explanation(packet)
+    
+    if "@@POINT@@" in response:
+        print("SUCCESS: Fallback logic correctly skipped the invalid model and used a secondary model.")
+    else:
+        print("FAILURE: Fallback logic did not return a valid AI summary.")
+    
+    # Reset models
+    advisor.LLM_MODELS = original_models
+
+if __name__ == "__main__":
+    sample_packet = {
+        "prediction": "Y",
+        "confidence": 0.88,
+        "dti_ratio": 32.5,
+        "optimized_suggestion": "Excellent profile."
+    }
+    
+    print("STARTING COMPREHENSIVE LLM VERIFICATION")
+    
+    # 1. Test every model in the list
+    results = []
+    for m in LLM_MODELS:
+        results.append(test_single_model(m, sample_packet))
+        
+    # 2. Test the fallback mechanism
+    test_fallback_logic(sample_packet)
+    
+    print("\nVERIFICATION COMPLETE")

frontend/.gitignore

@@ -0,0 +1,24 @@
+# Logs
+logs
+*.log
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+pnpm-debug.log*
+lerna-debug.log*
+
+node_modules
+dist
+dist-ssr
+*.local
+
+# Editor directories and files
+.vscode/*
+!.vscode/extensions.json
+.idea
+.DS_Store
+*.suo
+*.ntvs*
+*.njsproj
+*.sln
+*.sw?

frontend/README.md

@@ -0,0 +1,16 @@
+# React + Vite
+
+This template provides a minimal setup to get React working in Vite with HMR and some ESLint rules.
+
+Currently, two official plugins are available:
+
+- [@vitejs/plugin-react](https://github.com/vitejs/vite-plugin-react/blob/main/packages/plugin-react) uses [Oxc](https://oxc.rs)
+- [@vitejs/plugin-react-swc](https://github.com/vitejs/vite-plugin-react/blob/main/packages/plugin-react-swc) uses [SWC](https://swc.rs/)
+
+## React Compiler
+
+The React Compiler is not enabled on this template because of its impact on dev & build performances. To add it, see [this documentation](https://react.dev/learn/react-compiler/installation).
+
+## Expanding the ESLint configuration
+
+If you are developing a production application, we recommend using TypeScript with type-aware lint rules enabled. Check out the [TS template](https://github.com/vitejs/vite/tree/main/packages/create-vite/template-react-ts) for information on how to integrate TypeScript and [`typescript-eslint`](https://typescript-eslint.io) in your project.

frontend/eslint.config.js

@@ -0,0 +1,29 @@
+import js from '@eslint/js'
+import globals from 'globals'
+import reactHooks from 'eslint-plugin-react-hooks'
+import reactRefresh from 'eslint-plugin-react-refresh'
+import { defineConfig, globalIgnores } from 'eslint/config'
+
+export default defineConfig([
+  globalIgnores(['dist']),
+  {
+    files: ['**/*.{js,jsx}'],
+    extends: [
+      js.configs.recommended,
+      reactHooks.configs.flat.recommended,
+      reactRefresh.configs.vite,
+    ],
+    languageOptions: {
+      ecmaVersion: 2020,
+      globals: globals.browser,
+      parserOptions: {
+        ecmaVersion: 'latest',
+        ecmaFeatures: { jsx: true },
+        sourceType: 'module',
+      },
+    },
+    rules: {
+      'no-unused-vars': ['error', { varsIgnorePattern: '^[A-Z_]' }],
+    },
+  },
+])

frontend/index.html

@@ -0,0 +1,13 @@
+<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="UTF-8" />
+    <link rel="icon" type="image/svg+xml" href="/favicon.svg" />
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <title>frontend</title>
+  </head>
+  <body>
+    <div id="root"></div>
+    <script type="module" src="/src/main.jsx"></script>
+  </body>
+</html>

frontend/package.json

@@ -0,0 +1,32 @@
+{
+  "name": "frontend",
+  "private": true,
+  "version": "0.0.0",
+  "type": "module",
+  "scripts": {
+    "dev": "vite",
+    "build": "vite build",
+    "lint": "eslint .",
+    "preview": "vite preview"
+  },
+  "dependencies": {
+    "axios": "^1.15.0",
+    "chart.js": "^4.5.1",
+    "lucide-react": "^1.8.0",
+    "react": "^19.2.4",
+    "react-chartjs-2": "^5.3.1",
+    "react-dom": "^19.2.4",
+    "react-markdown": "^10.1.0"
+  },
+  "devDependencies": {
+    "@eslint/js": "^9.39.4",
+    "@types/react": "^19.2.14",
+    "@types/react-dom": "^19.2.3",
+    "@vitejs/plugin-react": "^6.0.1",
+    "eslint": "^9.39.4",
+    "eslint-plugin-react-hooks": "^7.0.1",
+    "eslint-plugin-react-refresh": "^0.5.2",
+    "globals": "^17.4.0",
+    "vite": "^8.0.4"
+  }
+}

frontend/src/App.css

@@ -0,0 +1,184 @@
+.counter {
+  font-size: 16px;
+  padding: 5px 10px;
+  border-radius: 5px;
+  color: var(--accent);
+  background: var(--accent-bg);
+  border: 2px solid transparent;
+  transition: border-color 0.3s;
+  margin-bottom: 24px;
+
+  &:hover {
+    border-color: var(--accent-border);
+  }
+  &:focus-visible {
+    outline: 2px solid var(--accent);
+    outline-offset: 2px;
+  }
+}
+
+.hero {
+  position: relative;
+
+  .base,
+  .framework,
+  .vite {
+    inset-inline: 0;
+    margin: 0 auto;
+  }
+
+  .base {
+    width: 170px;
+    position: relative;
+    z-index: 0;
+  }
+
+  .framework,
+  .vite {
+    position: absolute;
+  }
+
+  .framework {
+    z-index: 1;
+    top: 34px;
+    height: 28px;
+    transform: perspective(2000px) rotateZ(300deg) rotateX(44deg) rotateY(39deg)
+      scale(1.4);
+  }
+
+  .vite {
+    z-index: 0;
+    top: 107px;
+    height: 26px;
+    width: auto;
+    transform: perspective(2000px) rotateZ(300deg) rotateX(40deg) rotateY(39deg)
+      scale(0.8);
+  }
+}
+
+#center {
+  display: flex;
+  flex-direction: column;
+  gap: 25px;
+  place-content: center;
+  place-items: center;
+  flex-grow: 1;
+
+  @media (max-width: 1024px) {
+    padding: 32px 20px 24px;
+    gap: 18px;
+  }
+}
+
+#next-steps {
+  display: flex;
+  border-top: 1px solid var(--border);
+  text-align: left;
+
+  & > div {
+    flex: 1 1 0;
+    padding: 32px;
+    @media (max-width: 1024px) {
+      padding: 24px 20px;
+    }
+  }
+
+  .icon {
+    margin-bottom: 16px;
+    width: 22px;
+    height: 22px;
+  }
+
+  @media (max-width: 1024px) {
+    flex-direction: column;
+    text-align: center;
+  }
+}
+
+#docs {
+  border-right: 1px solid var(--border);
+
+  @media (max-width: 1024px) {
+    border-right: none;
+    border-bottom: 1px solid var(--border);
+  }
+}
+
+#next-steps ul {
+  list-style: none;
+  padding: 0;
+  display: flex;
+  gap: 8px;
+  margin: 32px 0 0;
+
+  .logo {
+    height: 18px;
+  }
+
+  a {
+    color: var(--text-h);
+    font-size: 16px;
+    border-radius: 6px;
+    background: var(--social-bg);
+    display: flex;
+    padding: 6px 12px;
+    align-items: center;
+    gap: 8px;
+    text-decoration: none;
+    transition: box-shadow 0.3s;
+
+    &:hover {
+      box-shadow: var(--shadow);
+    }
+    .button-icon {
+      height: 18px;
+      width: 18px;
+    }
+  }
+
+  @media (max-width: 1024px) {
+    margin-top: 20px;
+    flex-wrap: wrap;
+    justify-content: center;
+
+    li {
+      flex: 1 1 calc(50% - 8px);
+    }
+
+    a {
+      width: 100%;
+      justify-content: center;
+      box-sizing: border-box;
+    }
+  }
+}
+
+#spacer {
+  height: 88px;
+  border-top: 1px solid var(--border);
+  @media (max-width: 1024px) {
+    height: 48px;
+  }
+}
+
+.ticks {
+  position: relative;
+  width: 100%;
+
+  &::before,
+  &::after {
+    content: '';
+    position: absolute;
+    top: -4.5px;
+    border: 5px solid transparent;
+  }
+
+  &::before {
+    left: 0;
+    border-left-color: var(--border);
+  }
+  &::after {
+    right: 0;
+    border-right-color: var(--border);
+  }
+}

frontend/src/App.jsx

@@ -0,0 +1,72 @@
+import { useState } from 'react';
+import Form from './components/Form';
+import Dashboard from './components/Dashboard';
+import Sidebar from './components/Sidebar';
+import Home from './components/Home';
+
+import { Home as HomeIcon } from 'lucide-react';
+
+function App() {
+  const [view, setView] = useState('home'); // 'home', 'form', 'result'
+  const [result, setResult] = useState(null);
+  const [loading, setLoading] = useState(false);
+  const [historyKey, setHistoryKey] = useState(0);
+
+  const handleProceed = () => setView('form');
+  
+  const handleNewAssessment = () => {
+    setResult(null);
+    setView('form');
+  };
+  
+  const handleSelectResult = (pastResult) => {
+    setResult(pastResult);
+    setView('result');
+  };
+  
+  const handleNewResult = (newResult) => {
+    setResult(newResult);
+    setView('result');
+    setHistoryKey(k => k + 1); // Refresh sidebar history
+  };
+
+  const handleBackToHome = () => setView('home');
+
+  return (
+    <div className="app-layout">
+      {/* Sidebar only visible on Form and Result pages */}
+      {(view === 'form' || view === 'result') && (
+        <Sidebar 
+          key={historyKey} 
+          onSelectResult={handleSelectResult} 
+          onNewAssessment={handleNewAssessment} 
+        />
+      )}
+      
+      <div className="main-content" style={{ position: 'relative' }}>
+        {/* Navigation Button at Top Right corner of all pages except Home */}
+        {view !== 'home' && (
+          <button className="back-to-home-btn" onClick={handleBackToHome}>
+            <HomeIcon size={16} /> Back to Home
+          </button>
+        )}
+
+        <div className="content-container">
+          {loading ? (
+            <div style={{ textAlign: 'center', padding: '6rem 0' }}>
+              <div style={{ fontSize: '1.5rem', color: 'var(--primary)', fontWeight: '600' }}>Analyzing application data...</div>
+            </div>
+          ) : view === 'home' ? (
+            <Home onProceed={handleProceed} />
+          ) : view === 'form' ? (
+            <Form setResult={handleNewResult} setLoading={setLoading} />
+          ) : view === 'result' && result ? (
+            <Dashboard result={result} onReset={handleNewAssessment} />
+          ) : null}
+        </div>
+      </div>
+    </div>
+  );
+}
+
+export default App;

frontend/src/components/Charts.jsx

@@ -0,0 +1,173 @@
+import {
+  Chart as ChartJS,
+  ArcElement,
+  Tooltip,
+  Legend,
+  CategoryScale,
+  LinearScale,
+  BarElement,
+  Title,
+  RadialLinearScale,
+  PointElement,
+  LineElement,
+  Filler,
+} from 'chart.js';
+import { Doughnut, Bar, Radar } from 'react-chartjs-2';
+
+ChartJS.register(
+  ArcElement,
+  Tooltip,
+  Legend,
+  CategoryScale,
+  LinearScale,
+  BarElement,
+  Title,
+  RadialLinearScale,
+  PointElement,
+  LineElement,
+  Filler
+);
+
+export function DTIGauge({ dti }) {
+  const data = {
+    labels: ['DTI', 'Remaining Capacity'],
+    datasets: [
+      {
+        data: [dti, Math.max(0, 100 - dti)],
+        backgroundColor: [
+          dti > 50 ? '#ef4444' : dti > 30 ? '#f59e0b' : '#10b981',
+          '#f3f4f6'
+        ],
+        borderWidth: 0,
+        cutout: '85%'
+      }
+    ]
+  };
+
+  const options = {
+    rotation: -90,
+    circumference: 180,
+    plugins: {
+      legend: { display: false },
+      tooltip: { enabled: false }
+    },
+    maintainAspectRatio: true,
+  };
+
+  return (
+    <div style={{ position: 'relative', width: '220px', margin: '0 auto' }}>
+      <Doughnut data={data} options={options} />
+      <div style={{
+        position: 'absolute',
+        top: '70%',
+        left: '50%',
+        transform: 'translate(-50%, -50%)',
+        textAlign: 'center'
+      }}>
+        <div style={{ fontSize: '1.75rem', fontWeight: '700' }}>{dti.toFixed(1)}%</div>
+        <div style={{ fontSize: '0.75rem', color: '#6b7280', fontWeight: '600' }}>DEBT-TO-INCOME</div>
+      </div>
+    </div>
+  );
+}
+
+export function ComparisonRadar({ userResults }) {
+  const benchmarks = JSON.parse(userResults.benchmarks_json || "{}");
+  
+  // Clean keys for display
+  const labelMap = {
+    'ApplicantIncome': 'Income',
+    'LoanAmount': 'Loan Size',
+    'Total_Income': 'Total Rev',
+    'EMI': 'EMI Load',
+    'Credit_History': 'Credit Score'
+  };
+
+  const keys = Object.keys(labelMap);
+  
+  // Normalize values roughly for radar (0-100 scale)
+  const normalize = (val, max) => Math.min(100, (val / max) * 100);
+
+  const userData = [
+    normalize(userResults.applicant_income, benchmarks.ApplicantIncome * 2 || 10000),
+    normalize(userResults.loan_amount, benchmarks.LoanAmount * 2 || 500),
+    normalize(userResults.applicant_income + userResults.coapplicant_income, benchmarks.Total_Income * 2 || 15000),
+    normalize((userResults.loan_amount * 1000) / userResults.loan_amount_term, benchmarks.EMI * 2 || 2000),
+    userResults.credit_history * 100
+  ];
+
+  const benchData = [50, 50, 50, 50, 50]; // Benchmarks are normalized to center
+
+  const data = {
+    labels: keys.map(k => labelMap[k]),
+    datasets: [
+      {
+        label: 'Your Profile',
+        data: userData,
+        backgroundColor: 'rgba(242, 98, 47, 0.2)',
+        borderColor: '#f2622f',
+        borderWidth: 2,
+        pointBackgroundColor: '#f2622f',
+      },
+      {
+        label: 'Approved Average',
+        data: benchData,
+        backgroundColor: 'rgba(107, 114, 128, 0.1)',
+        borderColor: '#9ca3af',
+        borderWidth: 1,
+        borderDash: [5, 5],
+        pointRadius: 0
+      }
+    ]
+  };
+
+  const options = {
+    scales: {
+      r: {
+        angleLines: { display: false },
+        suggestedMin: 0,
+        suggestedMax: 100,
+        ticks: { display: false }
+      }
+    },
+    plugins: {
+      legend: { position: 'bottom', labels: { boxWidth: 12, font: { size: 10 } } }
+    }
+  };
+
+  return <Radar data={data} options={options} />;
+}
+
+export function FeatureImportanceBar({ featuresJson }) {
+  let features = {};
+  try {
+    features = JSON.parse(featuresJson);
+  } catch(e) {
+    return null;
+  }
+
+  const sorted = Object.entries(features)
+    .sort(([, a], [, b]) => b - a)
+    .slice(0, 5);
+
+  const data = {
+    labels: sorted.map(([k]) => k.replace(/_/g, ' ')),
+    datasets: [{
+      label: 'Impact Score',
+      data: sorted.map(([,v]) => v),
+      backgroundColor: '#f2622f',
+      borderRadius: 6
+    }]
+  };
+
+  const options = {
+    indexAxis: 'y',
+    plugins: { legend: { display: false } },
+    scales: {
+      x: { display: false },
+      y: { grid: { display: false }, ticks: { font: { size: 11 } } }
+    }
+  };
+
+  return <Bar data={data} options={options} />;
+}

frontend/src/components/Dashboard.jsx

@@ -0,0 +1,104 @@
+import { RefreshCw, CheckCircle, XCircle, BrainCircuit, ShieldCheck, TrendingDown, Target, Lightbulb } from 'lucide-react';
+import ReactMarkdown from 'react-markdown';
+import { DTIGauge, FeatureImportanceBar, ComparisonRadar } from './Charts';
+
+export default function Dashboard({ result, onReset }) {
+  const isApproved = result.prediction === 'Y';
+  
+  // Parse structured AI points
+  const explanationPoints = result.explanation_text 
+    ? result.explanation_text.split('@@POINT@@').filter(p => p.trim() !== '')
+    : [];
+
+  return (
+    <div className="animate-in">
+      <div style={{ display: 'flex', justifyContent: 'space-between', alignItems: 'center', marginBottom: '2rem' }}>
+        <div className={`status-badge ${isApproved ? 'status-approved' : 'status-rejected'}`}>
+          {isApproved ? <CheckCircle size={24} /> : <XCircle size={24} />}
+          {isApproved ? 'Application Approved' : 'Action Required: Rejection'}
+        </div>
+        <button onClick={onReset} className="btn" style={{ background: 'white', color: 'var(--text-main)', border: '1px solid var(--border-color)', boxShadow: 'none' }}>
+          <RefreshCw size={16} /> New Assessment
+        </button>
+      </div>
+
+      {/* Top Metric Grid */}
+      <div className="metric-grid">
+        <div className="metric-tile">
+          <span className="metric-label">Approval Confidence</span>
+          <span className="metric-value" style={{ color: isApproved ? 'var(--success)' : 'var(--danger)' }}>
+            {(result.confidence * 100).toFixed(0)}%
+          </span>
+        </div>
+        <div className="metric-tile">
+          <span className="metric-label">DTI Ratio</span>
+          <span className="metric-value">{result.dti_ratio.toFixed(1)}%</span>
+        </div>
+        <div className="metric-tile">
+          <span className="metric-label">Credit Level</span>
+          <span className="metric-value">{result.credit_history === 1 ? 'High' : 'At-Risk'}</span>
+        </div>
+      </div>
+
+      <div className="glass-card" style={{ marginBottom: '2rem' }}>
+        <h2 style={{ marginBottom: '0.5rem', display: 'flex', alignItems: 'center', gap: '0.75rem', color: 'var(--primary)', fontWeight: '700' }}>
+          <BrainCircuit size={28} />
+          AI Advisor Narrative
+        </h2>
+        <p style={{ color: 'var(--text-muted)', marginBottom: '1.5rem', fontSize: '1rem' }}>
+          Structured multi-point analysis from the Mistral Reasoning Engine.
+        </p>
+        
+        <div className="narrative-grid">
+          {explanationPoints.length > 0 ? (
+            explanationPoints.map((point, index) => (
+              <div key={index} className="narrative-sub-card">
+                <div style={{ display: 'flex', gap: '1rem' }}>
+                   <div style={{ color: 'var(--primary)', marginTop: '0.2rem' }}><Lightbulb size={20} /></div>
+                   <div className="narrative-content">
+                     <ReactMarkdown>{point.trim()}</ReactMarkdown>
+                   </div>
+                </div>
+              </div>
+            ))
+          ) : (
+            <div className="narrative-sub-card" style={{ gridColumn: '1 / -1' }}>
+              <p>{result.explanation_text || "Gathering advisor narrative..."}</p>
+            </div>
+          )}
+        </div>
+        
+        {!isApproved && result.optimized_suggestion && (
+          <div className="ai-advice-box">
+            <h3><Target size={20} /> Actionable Path to Approval</h3>
+            <p style={{ fontSize: '1.1rem', color: '#4b5563' }}>{result.optimized_suggestion}</p>
+          </div>
+        )}
+      </div>
+
+      <div className="dashboard-layout">
+        <div className="glass-card">
+          <h3 style={{ marginBottom: '2rem', display: 'flex', alignItems: 'center', gap: '0.5rem', borderBottom: '1px solid var(--border-color)', paddingBottom: '1rem' }}>
+            <ShieldCheck size={20} color="var(--success)" />
+            Peer Comparison Radar
+          </h3>
+          <div style={{ height: '300px' }}>
+            <ComparisonRadar userResults={result} />
+          </div>
+        </div>
+
+        <div className="glass-card">
+          <h3 style={{ marginBottom: '2rem', display: 'flex', alignItems: 'center', gap: '0.5rem', borderBottom: '1px solid var(--border-color)', paddingBottom: '1rem' }}>
+            <TrendingDown size={20} color="var(--primary)" />
+            Impact Vector Analysis
+          </h3>
+          <FeatureImportanceBar featuresJson={result.feature_importance_json} />
+        </div>
+      </div>
+      
+      <div style={{ textAlign: 'center', marginTop: '3rem', opacity: 0.6 }}>
+         <DTIGauge dti={result.dti_ratio} />
+      </div>
+    </div>
+  );
+}

frontend/src/components/Form.jsx

@@ -0,0 +1,127 @@
+import { useState } from 'react';
+import axios from 'axios';
+import { ArrowRight } from 'lucide-react';
+
+export default function Form({ setResult, setLoading }) {
+  const [formData, setFormData] = useState({
+    gender: 'Male',
+    married: 'No',
+    dependents: '0',
+    education: 'Graduate',
+    self_employed: 'No',
+    applicant_income: 5000,
+    coapplicant_income: 0,
+    loan_amount: 100,
+    loan_amount_term: 360,
+    credit_history: 1.0,
+    property_area: 'Urban'
+  });
+
+  const handleChange = (e) => {
+    const { name, value } = e.target;
+    setFormData(prev => ({
+      ...prev,
+      [name]: ['applicant_income', 'coapplicant_income', 'loan_amount', 'loan_amount_term', 'credit_history'].includes(name) 
+        ? parseFloat(value) 
+        : value
+    }));
+  };
+
+  const handleSubmit = async (e) => {
+    e.preventDefault();
+    setLoading(true);
+    try {
+      const response = await axios.post('http://localhost:8000/api/predict', formData);
+      setResult(response.data);
+    } catch (error) {
+      console.error("Prediction error", error);
+      alert('Error predicting. Make sure backend is running.');
+    } finally {
+      setLoading(false);
+    }
+  };
+
+  return (
+    <div className="glass-card" style={{ maxWidth: '800px', margin: '0 auto' }}>
+      <h2 style={{ marginBottom: '1.5rem' }}>Application Details</h2>
+      <form onSubmit={handleSubmit}>
+        <div className="form-grid">
+          <div className="form-group">
+            <label>Gender</label>
+            <select name="gender" value={formData.gender} onChange={handleChange}>
+              <option>Male</option>
+              <option>Female</option>
+            </select>
+          </div>
+          <div className="form-group">
+            <label>Married</label>
+            <select name="married" value={formData.married} onChange={handleChange}>
+              <option>No</option>
+              <option>Yes</option>
+            </select>
+          </div>
+          <div className="form-group">
+            <label>Dependents</label>
+            <select name="dependents" value={formData.dependents} onChange={handleChange}>
+              <option>0</option>
+              <option>1</option>
+              <option>2</option>
+              <option>3+</option>
+            </select>
+          </div>
+          <div className="form-group">
+            <label>Education</label>
+            <select name="education" value={formData.education} onChange={handleChange}>
+              <option>Graduate</option>
+              <option>Not Graduate</option>
+            </select>
+          </div>
+          <div className="form-group">
+            <label>Self Employed</label>
+            <select name="self_employed" value={formData.self_employed} onChange={handleChange}>
+              <option>No</option>
+              <option>Yes</option>
+            </select>
+          </div>
+          <div className="form-group">
+            <label>Applicant Income (Monthly)</label>
+            <input type="number" name="applicant_income" value={formData.applicant_income} onChange={handleChange} required />
+          </div>
+          <div className="form-group">
+            <label>Co-Applicant Income</label>
+            <input type="number" name="coapplicant_income" value={formData.coapplicant_income} onChange={handleChange} required />
+          </div>
+          <div className="form-group">
+            <label>Loan Amount (in Thousands)</label>
+            <input type="number" name="loan_amount" value={formData.loan_amount} onChange={handleChange} required />
+          </div>
+          <div className="form-group">
+            <label>Loan Term (Months)</label>
+            <input type="number" name="loan_amount_term" value={formData.loan_amount_term} onChange={handleChange} required />
+          </div>
+          <div className="form-group">
+            <label>Credit History</label>
+            <select name="credit_history" value={formData.credit_history} onChange={handleChange}>
+              <option value="1.0">Good (1.0)</option>
+              <option value="0.0">Bad (0.0)</option>
+            </select>
+          </div>
+          <div className="form-group">
+            <label>Property Area</label>
+            <select name="property_area" value={formData.property_area} onChange={handleChange}>
+              <option>Urban</option>
+              <option>Semiurban</option>
+              <option>Rural</option>
+            </select>
+          </div>
+        </div>
+        
+        <div style={{ marginTop: '2rem', textAlign: 'right' }}>
+          <button type="submit" className="btn">
+            Analyze Application <ArrowRight size={18} />
+          </button>
+        </div>
+      </form>
+    </div>
+  );
+}

frontend/src/components/Home.jsx

@@ -0,0 +1,76 @@
+import { ArrowRight, BrainCircuit, LineChart, ShieldCheck, Database, Zap, FileSpreadsheet } from 'lucide-react';
+
+export default function Home({ onProceed }) {
+  return (
+    <div className="home-container" style={{ padding: '2rem 0' }}>
+      {/* Hero Section */}
+      <div style={{ textAlign: 'center', marginBottom: '5rem', display: 'flex', flexDirection: 'column', alignItems: 'center' }}>
+        <div style={{ background: 'var(--primary-glow)', padding: '1.5rem', borderRadius: '50%', color: 'var(--primary)', marginBottom: '2rem' }}>
+          <BrainCircuit size={64} />
+        </div>
+        
+        <h1 style={{ fontSize: '4.5rem', fontWeight: '800', letterSpacing: '-0.04em', marginBottom: '1.5rem', color: 'var(--text-main)', lineHeight: '1.1' }}>
+          Next-Generation <br/>
+          <span style={{ color: 'var(--primary)' }}>Loan Intelligence</span>
+        </h1>
+        
+        <p style={{ fontSize: '1.25rem', color: 'var(--text-muted)', maxWidth: '650px', marginBottom: '3rem', lineHeight: '1.6' }}>
+          Evaluate loan applications instantly. Our hybrid architecture combines deterministic mathematical modeling with Mistral Large 3 AI to provide human-readable, actionable financial decisions.
+        </p>
+
+        <button onClick={onProceed} className="btn" style={{ fontSize: '1.15rem', padding: '1.2rem 3.5rem', borderRadius: '40px' }}>
+          Proceed to Application <ArrowRight size={22} />
+        </button>
+      </div>
+
+      {/* Feature Grid Section */}
+      <div style={{ marginBottom: '4rem' }}>
+        <h2 style={{ textAlign: 'center', fontSize: '2.5rem', marginBottom: '3rem', fontWeight: '700' }}>Platform Capabilities</h2>
+        
+        <div style={{ display: 'grid', gridTemplateColumns: 'repeat(auto-fit, minmax(300px, 1fr))', gap: '2rem' }}>
+          {/* Card 1 */}
+          <div className="glass-card" style={{ padding: '2rem', display: 'flex', flexDirection: 'column', gap: '1rem' }}>
+            <div style={{ background: '#ecfdf5', width: '50px', height: '50px', borderRadius: '12px', display: 'flex', alignItems: 'center', justifyContent: 'center', color: 'var(--success)' }}>
+              <ShieldCheck size={28} />
+            </div>
+            <h3 style={{ fontSize: '1.5rem', fontWeight: '700' }}>Deterministic Engine</h3>
+            <p style={{ color: 'var(--text-muted)', lineHeight: '1.6' }}>
+              Base evaluations are driven by a strictly enforced Random Forest pipeline. No hallucinations—just pure math calculating Debt-to-Income ratios and credit health.
+            </p>
+          </div>
+
+          {/* Card 2 */}
+          <div className="glass-card" style={{ padding: '2rem', display: 'flex', flexDirection: 'column', gap: '1rem' }}>
+            <div style={{ background: 'var(--primary-glow)', width: '50px', height: '50px', borderRadius: '12px', display: 'flex', alignItems: 'center', justifyContent: 'center', color: 'var(--primary)' }}>
+              <Zap size={28} />
+            </div>
+            <h3 style={{ fontSize: '1.5rem', fontWeight: '700' }}>Generative Explanations</h3>
+            <p style={{ color: 'var(--text-muted)', lineHeight: '1.6' }}>
+              We pipe the mathematical packet into NVIDIA NIM (Mistral 3). It translates complex metrics into a multi-point, easy-to-read narrative for applicants.
+            </p>
+          </div>
+
+          {/* Card 3 */}
+          <div className="glass-card" style={{ padding: '2rem', display: 'flex', flexDirection: 'column', gap: '1rem' }}>
+            <div style={{ background: '#eff6ff', width: '50px', height: '50px', borderRadius: '12px', display: 'flex', alignItems: 'center', justifyContent: 'center', color: '#3b82f6' }}>
+              <LineChart size={28} />
+            </div>
+            <h3 style={{ fontSize: '1.5rem', fontWeight: '700' }}>Visual Benchmarking</h3>
+            <p style={{ color: 'var(--text-muted)', lineHeight: '1.6' }}>
+              Dynamically compare incoming applications against a "Typical Approved Profile" using our built-in Radar comparisons and internal ML weigh distribution charts.
+            </p>
+          </div>
+        </div>
+      </div>
+      
+      {/* Footer Banner */}
+      <div style={{ borderRadius: '20px', background: 'linear-gradient(135deg, #1a1b1e 0%, #374151 100%)', color: 'white', padding: '3rem', textAlign: 'center', display: 'flex', flexDirection: 'column', alignItems: 'center', gap: '1.5rem' }}>
+         <Database size={40} color="var(--primary)" />
+         <h2 style={{ fontSize: '2rem' }}>Ready to analyze your next portfolio?</h2>
+         <button onClick={onProceed} className="btn" style={{ background: 'white', color: 'var(--text-main)', boxShadow: 'none' }}>
+           Start Prediction Pipeline
+         </button>
+      </div>
+    </div>
+  );
+}

frontend/src/components/Sidebar.jsx

@@ -0,0 +1,95 @@
+import { useState, useEffect } from 'react';
+import axios from 'axios';
+import { Plus, Clock, CheckCircle2, XCircle, ChevronRight, Trash2 } from 'lucide-react';
+
+export default function Sidebar({ onSelectResult, onNewAssessment }) {
+  const [history, setHistory] = useState([]);
+  const [loading, setLoading] = useState(true);
+
+  const fetchHistory = async () => {
+    try {
+      setLoading(true);
+      const res = await axios.get('http://localhost:8000/api/history?limit=20');
+      setHistory(res.data);
+    } catch (error) {
+      console.error("Failed to load history", error);
+    } finally {
+      setLoading(false);
+    }
+  };
+
+  const handleClearHistory = async () => {
+    if (window.confirm("Are you sure you want to clear all history? This action cannot be undone.")) {
+      try {
+        await axios.delete('http://localhost:8000/api/history');
+        setHistory([]);
+      } catch (error) {
+        console.error("Failed to clear history", error);
+        alert("Could not clear history");
+      }
+    }
+  };
+
+  useEffect(() => {
+    fetchHistory();
+  }, []);
+  
+  return (
+    <div className="sidebar">
+      <button 
+        className="btn" 
+        onClick={onNewAssessment} 
+        style={{ width: '100%', marginBottom: '1.5rem', justifyContent: 'center' }}
+      >
+        <Plus size={18} /> New Assessment
+      </button>
+
+      <div style={{ fontSize: '0.85rem', fontWeight: '600', color: 'var(--text-muted)', marginBottom: '1rem', display: 'flex', alignItems: 'center', justifyContent: 'space-between' }}>
+        <div style={{ display: 'flex', alignItems: 'center', gap: '0.5rem' }}>
+           <Clock size={14} /> RECENT ASSESSMENTS
+        </div>
+        {history.length > 0 && (
+          <button 
+             onClick={handleClearHistory} 
+             style={{ background: 'none', border: 'none', color: 'var(--danger)', cursor: 'pointer', display: 'flex', alignItems: 'center', gap: '0.2rem', fontSize: '0.75rem', fontWeight: '600' }}
+             title="Clear All History"
+          >
+             <Trash2 size={14} />
+             Clear
+          </button>
+        )}
+      </div>
+
+      <div className="history-list">
+        {loading ? (
+          <div style={{ padding: '1rem', textAlign: 'center', color: 'var(--text-muted)' }}>Loading...</div>
+        ) : history.length === 0 ? (
+          <div style={{ padding: '1rem', textAlign: 'center', color: 'var(--text-muted)', fontSize: '0.9rem' }}>No history found.</div>
+        ) : (
+          history.map((item) => (
+            <div 
+              key={item.id} 
+              className="history-item"
+              onClick={() => onSelectResult(item)}
+            >
+              <div style={{ display: 'flex', alignItems: 'center', gap: '0.5rem', marginBottom: '0.25rem' }}>
+                {item.prediction === 'Y' ? (
+                  <CheckCircle2 size={16} color="var(--success)" />
+                ) : (
+                  <XCircle size={16} color="var(--danger)" />
+                )}
+                <span style={{ fontWeight: '600', fontSize: '0.95rem' }}>
+                  {item.prediction === 'Y' ? 'Approved' : 'Rejected'}
+                </span>
+              </div>
+              <div style={{ fontSize: '0.8rem', color: 'var(--text-muted)', display: 'flex', justifyContent: 'space-between' }}>
+                <span>Income: ${item.applicant_income}</span>
+                <ChevronRight size={14} />
+              </div>
+            </div>
+          ))
+        )}
+      </div>
+    </div>
+  );
+}

frontend/src/index.css

@@ -0,0 +1,271 @@
+@import url('https://fonts.googleapis.com/css2?family=Outfit:wght@300;400;500;600;700&display=swap');
+
+:root {
+  --bg-color: #fcfbf9;
+  --text-main: #1a1b1e;
+  --text-muted: #6b7280;
+  --primary: #f2622f;
+  --primary-glow: rgba(242, 98, 47, 0.15);
+  --card-bg: #ffffff;
+  --border-color: #efedeb;
+  --success: #10b981;
+  --danger: #ef4444;
+  --warning: #f59e0b;
+  --light-bg: #fff9f5;
+}
+
+* {
+  box-sizing: border-box;
+  margin: 0;
+  padding: 0;
+}
+
+body {
+  font-family: 'Outfit', sans-serif;
+  background-color: var(--bg-color);
+  color: var(--text-main);
+  line-height: 1.6;
+  -webkit-font-smoothing: antialiased;
+}
+
+.app-layout {
+  display: flex;
+  height: 100vh;
+  width: 100vw;
+  overflow: hidden;
+  background-color: var(--bg-color);
+}
+
+.sidebar {
+  width: 280px;
+  background-color: #fcfbf9; /* Matching current theme */
+  border-right: 1px solid var(--border-color);
+  display: flex;
+  flex-direction: column;
+  padding: 1.5rem;
+  overflow-y: auto;
+  flex-shrink: 0;
+}
+
+.main-content {
+  flex: 1;
+  overflow-y: auto;
+  padding: 3rem 2rem;
+  background-color: #ffffff;
+}
+
+.content-container {
+  max-width: 1100px;
+  margin: 0 auto;
+}
+
+.history-list {
+  display: flex;
+  flex-direction: column;
+  gap: 0.75rem;
+}
+
+.history-item {
+  background: #ffffff;
+  border: 1px solid var(--border-color);
+  padding: 1rem;
+  border-radius: 12px;
+  cursor: pointer;
+  transition: all 0.2s;
+}
+
+.history-item:hover {
+  border-color: var(--primary);
+  transform: translateY(-2px);
+  box-shadow: 0 4px 8px rgba(0,0,0,0.03);
+}
+
+.back-to-home-btn {
+  position: absolute;
+  top: 1.5rem;
+  right: 2rem;
+  background: white;
+  border: 1px solid var(--border-color);
+  color: var(--text-main);
+  padding: 0.6rem 1.2rem;
+  border-radius: 30px;
+  font-size: 0.9rem;
+  font-weight: 600;
+  display: flex;
+  align-items: center;
+  gap: 0.5rem;
+  cursor: pointer;
+  z-index: 100;
+  transition: all 0.2s;
+  box-shadow: 0 4px 12px rgba(0,0,0,0.03);
+}
+
+.back-to-home-btn:hover {
+  border-color: var(--primary);
+  color: var(--primary);
+  transform: translateY(-1px);
+  box-shadow: 0 6px 16px rgba(0,0,0,0.06);
+}
+
+.header {
+  text-align: center;
+  margin-bottom: 4rem;
+}
+
+.header h1 {
+  font-size: 3.5rem;
+  font-weight: 700;
+  letter-spacing: -0.03em;
+  margin-bottom: 0.75rem;
+  background: linear-gradient(135deg, #1a1b1e 0%, #4b5563 100%);
+  -webkit-background-clip: text;
+  -webkit-text-fill-color: transparent;
+}
+
+.header p {
+  color: var(--text-muted);
+  font-size: 1.2rem;
+}
+
+.glass-card {
+  background: var(--card-bg);
+  border: 1px solid var(--border-color);
+  border-radius: 20px;
+  padding: 2.5rem;
+  box-shadow: 0 10px 30px -10px rgba(0,0,0,0.05);
+}
+
+.metric-grid {
+  display: grid;
+  grid-template-columns: repeat(3, 1fr);
+  gap: 1.5rem;
+  margin-bottom: 2rem;
+}
+
+.metric-tile {
+  background: white;
+  border: 1px solid var(--border-color);
+  padding: 1.5rem;
+  border-radius: 16px;
+  text-align: center;
+}
+
+.metric-value {
+  font-size: 1.75rem;
+  font-weight: 700;
+  display: block;
+}
+
+.metric-label {
+  font-size: 0.85rem;
+  color: var(--text-muted);
+  text-transform: uppercase;
+  letter-spacing: 0.05em;
+  font-weight: 600;
+}
+
+/* NARRATIVE SUB-CARDS */
+.narrative-grid {
+  display: grid;
+  grid-template-columns: repeat(auto-fit, minmax(300px, 1fr));
+  gap: 1.25rem;
+  margin-top: 1.5rem;
+}
+
+.narrative-sub-card {
+  background: white;
+  border: 1px solid var(--border-color);
+  padding: 1.5rem;
+  border-radius: 16px;
+  border-left: 5px solid var(--primary);
+  box-shadow: 0 2px 8px rgba(0,0,0,0.02);
+  transition: all 0.2s ease;
+}
+
+.narrative-sub-card:hover {
+  transform: translateY(-2px);
+  box-shadow: 0 10px 20px rgba(0,0,0,0.04);
+}
+
+.btn {
+  background-color: var(--primary);
+  color: white;
+  border: none;
+  padding: 0.85rem 2rem;
+  font-size: 1rem;
+  font-weight: 600;
+  border-radius: 30px;
+  cursor: pointer;
+  transition: all 0.3s;
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  gap: 0.6rem;
+}
+
+.btn:hover {
+  background-color: #df5320;
+  transform: translateY(-2px);
+}
+
+.form-grid {
+  display: grid;
+  grid-template-columns: repeat(auto-fit, minmax(240px, 1fr));
+  gap: 2rem;
+}
+
+.form-group {
+  display: flex;
+  flex-direction: column;
+  gap: 0.6rem;
+}
+
+.form-group label {
+  font-weight: 600;
+  font-size: 0.95rem;
+}
+
+.form-group input, 
+.form-group select {
+  padding: 0.9rem;
+  border: 1px solid var(--border-color);
+  border-radius: 12px;
+  font-family: inherit;
+  font-size: 1rem;
+  background-color: #fdfdfd;
+}
+
+.dashboard-layout {
+  display: grid;
+  grid-template-columns: 1.2fr 1fr;
+  gap: 2rem;
+  margin-top: 2rem;
+}
+
+@media (max-width: 900px) {
+  .metric-grid { grid-template-columns: 1fr; }
+  .dashboard-layout { grid-template-columns: 1fr; }
+}
+
+.status-badge {
+  padding: 0.5rem 1.25rem;
+  border-radius: 30px;
+  font-weight: 700;
+  font-size: 1rem;
+  display: inline-flex;
+  align-items: center;
+  gap: 0.6rem;
+  margin-bottom: 2rem;
+}
+
+.status-approved { background-color: #ecfdf5; color: var(--success); }
+.status-rejected { background-color: #fef2f2; color: var(--danger); }
+
+.ai-advice-box {
+  background-color: #fffaf8;
+  border: 1px solid #fee2d5;
+  border-left: 6px solid var(--primary);
+  padding: 2rem;
+  border-radius: 16px;
+  margin-top: 2rem;
+}

frontend/src/main.jsx

@@ -0,0 +1,10 @@
+import { StrictMode } from 'react'
+import { createRoot } from 'react-dom/client'
+import './index.css'
+import App from './App.jsx'
+
+createRoot(document.getElementById('root')).render(
+  <StrictMode>
+    <App />
+  </StrictMode>,
+)

frontend/vite.config.js

@@ -0,0 +1,7 @@
+import { defineConfig } from 'vite'
+import react from '@vitejs/plugin-react'
+
+// https://vite.dev/config/
+export default defineConfig({
+  plugins: [react()],
+})

pyproject.toml

@@ -0,0 +1,32 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "loan-prediction-system"
+version = "0.1.0"
+description = "Enterprise Loan AI Dashboard with Deterministic ML and LLM Advisor"
+readme = "README.md"
+requires-python = ">=3.11"
+authors = [
+    { name = "Siddharaj Shirke" },
+]
+license = { text = "MIT" }
+dependencies = [
+    "fastapi>=0.100.0",
+    "uvicorn>=0.23.0",
+    "pydantic>=2.0.0",
+    "sqlalchemy>=2.0.0",
+    "openai>=1.0.0",
+    "scikit-learn>=1.2.0",
+    "pandas>=2.0.0",
+    "numpy>=1.24.0",
+    "python-dotenv>=1.0.0",
+    "huggingface_hub>=0.19.0",
+]
+
+[project.scripts]
+loan-app = "uvicorn backend.main:app --port 8000 --reload"
+
+[tool.hatch.build.targets.wheel]
+packages = ["backend"]

scratch/reset_db.py

@@ -0,0 +1,28 @@
+import os
+from backend.db.database import engine
+from backend.db import models
+
+def reset_db():
+    print("Attempting to reset database...")
+    try:
+        # Drop all tables
+        models.Base.metadata.drop_all(bind=engine)
+        print("Dropped all tables.")
+        
+        # Recreate all tables
+        models.Base.metadata.create_all(bind=engine)
+        print("Recreated all tables with new schema.")
+        
+        # Alternatively, try to delete the file if it exists
+        db_path = "loan_system.db"
+        if os.path.exists(db_path):
+            try:
+                os.remove(db_path)
+                print(f"Deleted {db_path} file.")
+            except Exception as e:
+                print(f"Could not delete file (might be locked): {e}")
+    except Exception as e:
+        print(f"Error during reset: {e}")
+
+if __name__ == "__main__":
+    reset_db()