# -*- coding: utf-8 -*-
"""AI Loan Officer Assistant.ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/1Om043ogOg3FZB4mkhCZY2gcOapGFM-5q
"""

import os
import time
import pandas as pd
import gradio as gr
from google import genai
from google.genai import types
from google.api_core import exceptions

# 1. Initialize Client WITHOUT api_version
# This specifically fixes the 'INVALID_ARGUMENT' error you received

# Initialize the client with explicit API version 'v1'
import os
import time
import pandas as pd
import gradio as gr
from google import genai
from google.genai import types
from google.api_core import exceptions

# SECURE FIX: Use os.environ for Hugging Face Spaces
# The SDK will automatically use 'GOOGLE_API_KEY' if it exists in environment
client = genai.Client(api_key=os.environ.get('GOOGLE_API_KEY'))

# Use gemini-1.5-flash for the best free-tier stability
STABLE_MODEL = "gemini-flash-latest"

# 2. Data Preparation (Structured Systems)
df_credit = pd.DataFrame({'ID': [1111, 2222, 3333, 4444, 5555], 'Credit_Score': [455, 685, 825, 840, 350]})
df_account = pd.DataFrame({
    'ID': [1111, 2222, 3333, 4444, 5555],
    'Name': ['Loren', 'Matt', 'Hilda', 'Andy', 'Kit'],
    'Email': ['loren@gmail.com', 'matt@yahoo.com', 'halida@gmail.com', 'andy@gmail.com', 'kit@yahho.com'],
    'Status': ['Good-standing', 'Closed', 'Delinquent', 'Good-standing', 'Delinquent'],
    'Nationality': ['Singaporean', 'NonSingaporean', 'Singaporean', 'NonSingaporean', 'Singaporean']
})
df_gov = pd.DataFrame({'ID': [2222, 4444], 'PR_Status': [True, False]})
df_merged = df_account.merge(df_credit, on="ID").merge(df_gov, on="ID", how="left").fillna(False)

# 3. Upload Policy Files once on startup
risk_policy_file = client.files.upload(file="Bank Loan Overall Risk Policy.pdf")
interest_policy_file = client.files.upload(file="Bank Loan Interest Rate Policy.pdf")

# 4. Standardized Assessment Logic
def process_loan_request(applicant_id, customer_name_optional):
    try:
        # Retrieve data from the structured systems
        row = df_merged[df_merged['ID'] == int(applicant_id)]
        if row.empty: return "Error: Applicant ID not found."
        applicant = row.iloc[0]

        # THE STANDARDIZED PROMPT: Ensuring strict adherence to provided tables
        prompt = f"""
        You are a Senior Loan Officer. Assess the application for the individual provided. 
        Stick strictly to the information and tables provided in the attached PDF policies.

        APPLICANT DATA:
        Name: {applicant['Name']}
        ID: {applicant_id}
        Email: {applicant['Email']}
        Credit Score: {applicant['Credit_Score']}
        Account Status: {applicant['Status']}
        Nationality: {applicant['Nationality']}
        PR Status: {applicant['PR_Status']}

        REQUIRED OUTPUT PATTERN:

        # Loan Assessment Report: {applicant['Name']} (ID: {applicant_id}, Email: {applicant['Email']})

        ### **1. Info Retrieval**
        | Applicant Detail | Value |
        | :--- | :--- |
        | Name (ID) | {applicant['Name']} ({applicant_id}) |
        | Credit Score | {applicant['Credit_Score']} |
        | Account Status | {applicant['Status']} |
        | Nationality | {applicant['Nationality']} |
        | PR Status | {applicant['PR_Status']} |

        ### **2. PR Status Check**
        - For Non-Singaporean applicants: State result as "PR Status -> [True/False]".
        - For Singaporean applicants: State "Check not required for Singaporean".

        ### **3. Overall Risk Assessment**
        - Reference 'Bank Loan Overall Risk Policy.pdf'.
        - Calculation: Credit Score [{applicant['Credit_Score']}] + Account Status [{applicant['Status']}] -> Overall Risk: [Value from Table]

        ### **4. Interest Rate Determination**
        - Reference 'Bank Loan Interest Rate Policy.pdf'.
        - Calculation: Overall Risk [Value from Table] -> [Rate from Table]
        - Apply conservative rule if risk falls between categories.

        ### **5. Final Report & Recommendation**
        - Recommend the loan interest rate OR Not recommend.
        - Rationale: Based on risk profile and compliance (PR false for Non-Singaporean leads to 'Not recommend').
        """

        # Generate content using the multimodal input
        response = client.models.generate_content(
            model=STABLE_MODEL,
            contents=[risk_policy_file, interest_policy_file, prompt]
        )
        return response.text

    except exceptions.ResourceExhausted:
        return "Error: Quota reached. Please wait 60 seconds."
    except Exception as e:
        return f"System Error: {str(e)}"

# 5. Launch Gradio Interface
demo = gr.Interface(
    fn=process_loan_request,
    inputs=[gr.Textbox(label="Applicant ID"), gr.Textbox(label="Name (Optional)")],
    outputs=gr.Markdown(),
    title="🏦 Enterprise AI Loan Risk Assessment Tool"
)

if __name__ == "__main__":
    demo.launch()