app.py

from statistics import mean
import random

import torch
from transformers import BertModel, BertTokenizerFast
import numpy as np
import torch.nn.functional as F
import gradio as gr

threshold = 0.4

tokenizer = BertTokenizerFast.from_pretrained("setu4993/LaBSE")
model = BertModel.from_pretrained("setu4993/LaBSE")
model = model.eval()

order_food_ex = [
    "food",
    "I am hungry, I want to order food",
    "How do I order food",
    "What are the food options",
    "I need dinner",
    "I want lunch",
    "What are the menu options",
    "I want a hamburger"
        ]

talk_to_human_ex = [
    "I need to talk to someone",
    "Connect me with a human",
    "I need to speak with a person",
    "Put me on with a human",
    "Connect me with customer service",
    "human"
        ]

def embed(text, tokenizer, model):
    inputs = tokenizer(text, return_tensors="pt", padding=True)
    with torch.no_grad():
        outputs = model(**inputs)
    return outputs.pooler_output

def similarity(embeddings_1, embeddings_2):
    normalized_embeddings_1 = F.normalize(embeddings_1, p=2)
    normalized_embeddings_2 = F.normalize(embeddings_2, p=2)
    return torch.matmul(
        normalized_embeddings_1, normalized_embeddings_2.transpose(0, 1)
    )

order_food_embed = [embed(x, tokenizer, model) for x in order_food_ex]
talk_to_human_embed = [embed(x, tokenizer, model) for x in talk_to_human_ex]

def chat(message, history):
    history = history or []
    message_embed = embed(message, tokenizer, model)
    order_sim = []
    for em in order_food_embed:
        order_sim.append(float(similarity(em, message_embed)))
    human_sim = []
    for em in talk_to_human_embed:
        human_sim.append(float(similarity(em, message_embed)))
    if mean(order_sim) > threshold:
        response = random.choice([
            "We have hamburgers or pizza! Which one do you want?",
            "Do you want a hamburger or a pizza?"])
    elif mean(human_sim) > threshold:
        response = random.choice([
            "Sure, a customer service agent will jump into this convo shortly!", 
            "No problem. Let me forward on this conversation to a person that can respond."])
    else:
        response = "Sorry, I didn't catch that. Could your rephrase?"
    history.append((message, response))
    return history, history

iface = gr.Interface(
    chat,
    ["text", "state"],
    ["chatbot", "state"],
    allow_screenshot=False,
    allow_flagging="never",
)
iface.launch()