app.py

import os
import gradio as gr
import joblib
import logfire
import numpy as np
import pandas as pd
from openai import OpenAI
from pydantic import BaseModel

# Configure logging
logfire.configure(token=os.getenv("LOGFIRE_API_KEY"))
logfire.instrument_pydantic()

# Load pre-trained model and label names
model_data = joblib.load("model.joblib")
model = model_data["model"]
label_names = model_data["label_names"]

class Results(BaseModel):
    text: str
    hateful: float
    insults: float
    sexual: float
    violence: float
    self_harm: float
    aom: float

# Initialize OpenAI client
client = OpenAI()

def get_embedding(text: str, embedding_model: str = "text-embedding-3-large") -> np.ndarray:
    """
    Get embedding for the input text from OpenAI.
    Replaces newlines with spaces before calling the API.
    """
    text = text.replace("\n", " ")
    response = client.embeddings.create(input=[text], model=embedding_model)
    embedding = response.data[0].embedding
    return np.array(embedding)

def classify_text(text: str):
    """
    Get the OpenAI embedding for the provided text, classify it using your model,
    and return a DataFrame with the rounded probabilities and binary predictions.
    """
    embedding = get_embedding(text)
    X = embedding.reshape(1, -1)
    probabilities = model.predict(X)
    rounded_probs = np.round(probabilities[0], 4)
    
    # Optionally log the results (this doesn't affect the output)
    Results(
        text=text,
        hateful=rounded_probs[0],
        insults=rounded_probs[1],
        sexual=rounded_probs[2],
        violence=rounded_probs[3],
        self_harm=rounded_probs[4],
        aom=rounded_probs[5],
    )
    
    # Create DataFrame with rounded probabilities and binary predictions
    df = pd.DataFrame({
        "Label": label_names,
        "Probability": rounded_probs,
        "Prediction": (rounded_probs > 0.5).astype(int)
    })
    
    return gr.update(value=df, visible=True)

with gr.Blocks(title="Zoo Entry 001") as iface:
    input_text = gr.Textbox(lines=5, label="Input Text")
    submit_btn = gr.Button("Submit")
    output_table = gr.DataFrame(label="Classification Results", visible=False)
    
    submit_btn.click(fn=classify_text, inputs=input_text, outputs=output_table)

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