import openai
import pandas as pd
import faiss
import numpy as np
from sentence_transformers import SentenceTransformer
import gradio as gr
from gtts import gTTS
import tempfile
import os
import speech_recognition as sr

openai.api_key = os.getenv("OPENAI_API_KEY") 

# Load embedding model
embed_model = SentenceTransformer("all-MiniLM-L6-v2")

# Load FAQ CSV
faq_df = pd.read_csv("fact-base-tesco.csv")
questions = faq_df["Question"].tolist()
answers = faq_df["Answer"].tolist()

# Create FAISS index
faq_embeddings = embed_model.encode(questions, convert_to_numpy=True)
index = faiss.IndexFlatL2(faq_embeddings.shape[1])
index.add(faq_embeddings)

def ask_faq(query):
    query_vec = embed_model.encode([query], convert_to_numpy=True)
    D, I = index.search(query_vec, k=1)
    similarity = D[0][0]

    if similarity < 0.5:
        answer = answers[I[0][0]]
        response = answer
    else:
        try:
            completion = openai.ChatCompletion.create(
                model="gpt-3.5-turbo",
                messages=[{"role": "user", "content": query}],
                max_tokens=200
            )
            response = completion.choices[0].message.content
        except:
            response = "Sorry, I couldn't generate a response right now."

    tts = gTTS(response)
    tts_fp = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
    tts.save(tts_fp.name)

    return response, tts_fp.name

def process_input(text, audio):
    if audio is not None:
        recognizer = sr.Recognizer()
        with sr.AudioFile(audio) as source:
            audio_data = recognizer.record(source)
            try:
                text = recognizer.recognize_google(audio_data)
            except:
                return "Sorry, could not understand audio", None

    if text:
        return ask_faq(text)
    else:
        return "Please provide a question.", None

with gr.Blocks() as demo:
    gr.Markdown("# 🎙️ FAQ + GPT Chatbot with Voice")
    gr.Markdown("Ask via text or mic. Falls back to GPT if FAQ match is not found.")

    text_input = gr.Textbox(label="Type your question or click mic below ⬇")
    mic_input = gr.Audio(type="filepath", label="🎤 Speak your question")
    submit_btn = gr.Button("Ask")
    text_output = gr.Textbox(label="Answer")
    audio_output = gr.Audio(label="🔊 Voice Answer", autoplay=True)

    submit_btn.click(fn=process_input, inputs=[text_input, mic_input], outputs=[text_output, audio_output])

demo.launch()