app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
import pandas as pd
import numpy as np
from sklearn.linear_model import LinearRegression
from io import StringIO

# Load the model and tokenizer once
model_name = "ibm-granite/granite-3.3-2b-instruct"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", torch_dtype="auto")
llm = pipeline("text-generation", model=model, tokenizer=tokenizer)

# Module 1: Policy Summarization
def policy_summarizer(text):
    prompt = f"Summarize the following city policy in simple terms:\n{text}\nSummary:"
    result = llm(prompt, max_new_tokens=200)[0]["generated_text"]
    return result.replace(prompt, "").strip()

# Module 2: Citizen Feedback
def citizen_feedback(issue):
    return f"Thank you! Your issue '{issue}' has been logged and categorized appropriately."

# Module 3: KPI Forecasting
def kpi_forecasting(csv_file):
    df = pd.read_csv(csv_file.name)
    X = df.iloc[:, 0].values.reshape(-1, 1)
    y = df.iloc[:, 1].values
    model = LinearRegression().fit(X, y)
    next_year = [[X[-1][0] + 1]]
    prediction = model.predict(next_year)[0]
    return f"📈 Predicted KPI for {next_year[0][0]}: {round(prediction, 2)}"

# Module 4: Eco Tips Generator
def eco_tips(keyword):
    prompt = f"Give 3 actionable eco-friendly tips related to: {keyword}"
    result = llm(prompt, max_new_tokens=100)[0]["generated_text"]
    return result.replace(prompt, "").strip()

# Module 5: Anomaly Detection
def detect_anomaly(csv_file):
    df = pd.read_csv(csv_file.name)
    if 'value' not in df.columns:
        return "⚠️ CSV must contain a 'value' column."
    mean = df["value"].mean()
    std = df["value"].std()
    anomalies = df[np.abs(df["value"] - mean) > 2 * std]
    if anomalies.empty:
        return "✅ No significant anomalies detected."
    return "⚠️ Anomalies found:\\n" + anomalies.to_string(index=False)

# module 6: chat assistant
def chat_assistant(question):
    prompt = f"Answer this smart city sustainability question:\n\nQ: {question}\nA:"
    result = llm(prompt, max_new_tokens=200, temperature=0.7)[0]["generated_text"]
    return result.replace(prompt, "").strip()


# Gradio Interface
app = gr.TabbedInterface(
    interface_list=[
        gr.Interface(fn=policy_summarizer, inputs=gr.Textbox(lines=7, label="Paste Policy Text"), outputs="textbox", title="📝 Policy Summarization"),
        gr.Interface(fn=citizen_feedback, inputs=gr.Textbox(lines=2, label="Describe the Issue"), outputs="text", title="📣 Citizen Feedback"),
        gr.Interface(fn=kpi_forecasting, inputs=gr.File(label="Upload CSV with Year and Value columns"), outputs="text", title="📊 KPI Forecasting"),
        gr.Interface(fn=eco_tips, inputs=gr.Textbox(lines=2, label="Keyword (e.g. Plastic, Solar)"), outputs="text", title="🌱 Eco Tips Generator"),
        gr.Interface(fn=detect_anomaly, inputs=gr.File(label="Upload CSV with 'value' column"), outputs="text", title="🚨 Anomaly Detection"),
        gr.Interface(fn=chat_assistant, inputs=gr.Textbox(lines=2, label="Ask your question"), outputs="text", title="💬 Smart City Chat")
    ],
    tab_names=["Summarize", "Feedback", "Forecast", "Eco Tips", "Anomalies", "Chat"],
    title="🌆 Sustainable Smart City Assistant"
)

app.launch()