app.py

import os
import gradio as gr
from transformers import pipeline
import plotly.express as px
import pandas as pd

HF_TOKEN = os.getenv("HF_TOKEN")

MODEL_LIST = [
    "thethinkmachine/immune-resin",
    "thethinkmachine/bright-avocado",
    "thethinkmachine/kickass-supercluster",
]

pipeline_cache = {}

def get_pipeline(model_name):
    if model_name not in pipeline_cache:
        pipeline_cache[model_name] = pipeline(
            "text-classification",
            model=model_name,
            token=HF_TOKEN,
            return_all_scores=True
        )
    return pipeline_cache[model_name]


def classify(text, model_name, chart_type, threshold):
    if not text.strip():
        # return empty outputs for the table and plot
        return None, None

    clf = get_pipeline(model_name)
    results = clf(text)[0]
    df = pd.DataFrame(results)

    # Sort labels by probability
    df = df.sort_values(by="score", ascending=False).reset_index(drop=True)

    # Highlight labels above threshold
    df["highlight"] = df["score"].apply(lambda x: "High" if x >= threshold else "Low")

    # Chart selection
    if chart_type == "Radar Chart":
        fig = px.line_polar(
            df,
            r="score",
            theta="label",
            line_close=True,
            color="highlight",
            title=f"Label Probabilities - {model_name}"
        )
        fig.update_traces(fill='toself')
        fig.update_layout(polar=dict(radialaxis=dict(range=[0, 1])))
    else:
        fig = px.bar(
            df,
            x="label",
            y="score",
            color="highlight",
            title=f"Label Probabilities - {model_name}"
        )
        fig.update_layout(yaxis=dict(range=[0, 1]))

    return df, fig


with gr.Blocks(theme=gr.themes.Ocean()) as demo:
    gr.Markdown("# Ekman Emotions Playground 🤖👾🦜")
    gr.Markdown("### Why let humans play with your emotions when a robot can do it for you for free? ...Cheaper than therapy, 100% less effective!")

    with gr.Row():
        model_dropdown = gr.Dropdown(
            choices=MODEL_LIST,
            label="Select Model",
            value=MODEL_LIST[0]
        )
        chart_dropdown = gr.Dropdown(
            choices=["Radar Chart", "Bar Chart"],
            label="Chart Type",
            value="Bar Chart"
        )
        threshold_slider = gr.Slider(
            0, 1,
            value=0.5,
            step=0.01,
            label="Highlight Threshold"
        )

    text_input = gr.Textbox(label="Input Text")

    with gr.Row():
        output_table = gr.DataFrame(label="Scores Table")
        output_plot = gr.Plot(label="Probability Chart")

    classify_btn = gr.Button("Run Classification")

    classify_btn.click(
        classify,
        inputs=[text_input, model_dropdown, chart_dropdown, threshold_slider],
        outputs=[output_table, output_plot]
    )

    gr.Examples(
        examples=[
            "I only saw her once and I'm head over heels!",
            "Pay you for what, just standing there?!",
            "Dumbass Broncos fans circa December 2015.",
            "It's great that you're a recovering addict, that's cool. Have you ever tried DMT?",
            "I'm scared to even ask my mom ,I might get yelled at 😟",
            "hurr durr I like using reddit, and anyone who doesn't agree with me is a retard",
            "They're really pushing me into this... once I go, there's no coming back you know?",
            "Considering I haven't eaten or drunk anything in about twenty hours, my head hurts.",
        ],
        inputs=[text_input]
    )

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