app.py

import gradio as gr
import numpy as np
import matplotlib.pyplot as plt
import cv2
from io import BytesIO
import uuid
import os
import requests
import json


AIRTABLE_TOKEN = os.environ.get("AIRTABLE_TOKEN")
SUMMARY_TABLE_URL = os.environ.get("SUMMARY_TABLE_URL")
QUERIES_TABLE_URL = os.environ.get("QUERIES_TABLE_URL")

def send_to_airtable(data, url):
    headers = {
        "Authorization": f"Bearer {AIRTABLE_TOKEN}",
        "Content-Type": "application/json"
    }
    data = json.dumps(data)
    response = requests.post(url, headers=headers, data=data)
    print(response.text)
    return response.text

def log_final_decision(activity, objective, budget_left, session_id, url):
    data = {"fields": {"activity": activity, "objective": objective, "budget_left": budget_left, "session_id": session_id}}
    return send_to_airtable(data, url)


def log_query(session_id, x, y, url):
    data = {"fields": {"session_id": session_id, "x": x, "y": y}}
    return send_to_airtable(data, url)

def draw_points(points, query_count, session_id=""):
    plt.figure(figsize=(12, 8))
    if points:
        x_vals, y_vals = zip(*points)
        cmap = plt.get_cmap('viridis')
        for i, (x_val, y_val) in enumerate(points):
            plt.scatter(x_val, y_val, color=cmap(i / len(points)))
            plt.text(x_val, y_val + 50, str(i + 1), fontsize=7, ha='center')
        plt.xlabel('x')
        plt.ylabel('f(x)')
        plt.ylim(-1000, 1000)
        plt.grid(True)
        plt.axhline(0, color='black', linewidth=0.5)
        plt.axvline(0, color='black', linewidth=0.5)
        plt.title(f'Sampled Points - Query {query_count[0] + 1} ({session_id})')
        buf = BytesIO()
        plt.savefig(buf, format='png')
        plt.close()
        buf.seek(0)
        image = np.frombuffer(buf.getvalue(), dtype=np.uint8)
        image = cv2.imdecode(image, cv2.IMREAD_COLOR)
        return image

def f(x, scale=1000):
    x_scaled = x / scale
    polynomial = (x_scaled**4 - 10*x_scaled**3 + 35*x_scaled**2 - 50*x_scaled + 24) / 4000
    sinusoidal = 800 * np.sin(abs(x_scaled)/10) + 500 * np.cos(x_scaled/8) + 200 * np.sin(x_scaled/2)
    y_value = 800 * np.tanh((polynomial + sinusoidal) / 1000)
    return y_value

def finalize_decision(global_minimizer, points, session_state, query_count, nqueries=10, activity="group_1"):
    session_id = session_state['session_id'] if session_state['session_id'] else str(uuid.uuid4())[:6]
    remaining_queries = nqueries - query_count[0]
    try:
        x_star = float(global_minimizer)
        y_value = f(x_star)
        y_star = y_value - 10 * remaining_queries
        session_end_msg = f"Session ends with remaining query {remaining_queries}, and objective {y_value:.2f} - {remaining_queries}*10 = {y_star:.2f}."
        
        points.append((x_star, y_value))
        log_query(session_id, x_star, y_value, QUERIES_TABLE_URL)
        img = draw_points(points, query_count, session_id)        
        points_list = '\n'.join([f"x: {p[0]:.0f}, y: {p[1]:.3f}" for p in points])
        #add  "--- final decision" string to the final points_list element
        points_list += "--- final decision"
        
        log_final_decision(activity, y_value, remaining_queries, session_id, SUMMARY_TABLE_URL)
        
        return [session_end_msg, session_end_msg, gr.Button("Submit", interactive=False), gr.Button("Finalize Decision", interactive=False), img, points_list]
    except ValueError:
        img = draw_points(points, query_count, session_id)
        points_list = '\n'.join([f"x: {p[0]:.0f}, y: {p[1]:.3f}" for p in points])
        msg = "Invalid input: Please enter a numeric value for x*"
        return [msg, msg, gr.Button("Submit", interactive=True), gr.Button("Finalize Decision", interactive=True), img, points_list]

def submit_query(x, query_count, points, session_state, nqueries=10):
    session_id = session_state['session_id'] if session_state['session_id'] else str(uuid.uuid4())[:6]
    session_state['session_id'] = session_id  

    if query_count[0] >= nqueries:
        return None, query_count, "Query limit reached", "", session_state, "Query limit reached"

    try:
        x = float(x)
    except ValueError:
        msg = "Invalid input: Please enter a numeric value"
        return None, query_count, msg, "", session_state, msg

    y_value = f(x)
    points.append((x, y_value))
    log_query(session_id, x, y_value, QUERIES_TABLE_URL)
    img = draw_points(points, query_count, session_id)
    query_count[0] += 1
    remaining_queries = nqueries - query_count[0]
    points_list = '\n'.join([f"x: {p[0]:.0f}, y: {p[1]:.3f}" for p in points])
    msg = f"{remaining_queries} queries remaining"
    return img, query_count, msg, points_list, session_state, msg

with gr.Blocks() as app:
    query_count = gr.State([0])
    points = gr.State([])
    session_state = gr.State({'session_id': str(uuid.uuid4())[:6]})  
    
    with gr.Row():
        x_input = gr.Number(label="X")
        submit_button = gr.Button("Submit", interactive=True)

    with gr.Row():
        remaining_queries_label = gr.Label()
        points_display = gr.Textbox(label="Queried Points", lines=10, interactive=False, value="")
        
    with gr.Row():
        output_image = gr.Image(label="Plot", width="100vw")
        
    with gr.Row():        
        global_minimizer_input = gr.Textbox(label="Global Minimizer (x*)")
        finalize_button = gr.Button("Finalize Decision")
        
    with gr.Row():
        final_label = gr.Label()
   

    submit_button.click(
        submit_query,
        inputs=[x_input, query_count, points, session_state],
        outputs=[output_image, query_count, remaining_queries_label, points_display, session_state, final_label]
    )

    finalize_button.click(
        finalize_decision,
        inputs=[global_minimizer_input, points, session_state, query_count],
        outputs=[remaining_queries_label, final_label, submit_button, finalize_button, output_image, points_display]
    )

app.launch(share=False)

#save the output into airtable
#save the queries into the correct table