Update app.py

app.py

@@ -1,16 +1,18 @@
-import streamlit as st
 import torch
+import torch.nn as nn
 import torch.nn.functional as F
+import random
 from textblob import TextBlob
 import pandas as pd
 import requests
 from io import StringIO
+import gradio as gr
 import speech_recognition as sr
-import random
-
-# --- Your Dummy Model and Helpers (same as before) ---
+import json
 
-vocab = {'<PAD>': 0, '<UNK>': 1, 'i': 2, 'am': 3, 'feeling': 4, 'sad': 5, 'happy': 6, 'angry': 7, 'love': 8, 'stressed': 9, 'anxious': 10}
+# --- Dummy vocab and label encoder ---
+vocab = {'<PAD>': 0, '<UNK>': 1, 'i': 2, 'am': 3, 'feeling': 4, 'sad': 5, 'happy': 6,
+         'angry': 7, 'love': 8, 'stressed': 9, 'anxious': 10}
 MAX_LEN = 16
 
 class DummyLabelEncoder:
@@ -21,17 +23,18 @@ class DummyLabelEncoder:
 
 le = DummyLabelEncoder()
 
-class DummyModel(torch.nn.Module):
+class DummyModel(nn.Module):
     def __init__(self):
         super().__init__()
-        self.embedding = torch.nn.Embedding(len(vocab), 8)
-        self.fc = torch.nn.Linear(8, len(le.classes_))
+        self.embedding = nn.Embedding(len(vocab), 8)
+        self.fc = nn.Linear(8, len(le.classes_))
     def forward(self, x):
         x = self.embedding(x)
         x = x.mean(dim=1)
         return self.fc(x)
 
 model = DummyModel()
+model.eval()
 
 def preprocess_input(text):
     tokens = text.lower().split()
@@ -39,26 +42,16 @@ def preprocess_input(text):
     padded = encoded[:MAX_LEN] + [vocab['<PAD>']] * max(0, MAX_LEN - len(encoded))
     return torch.tensor([padded], dtype=torch.long)
 
-# Load CSV from Google Drive
-file_id = "1yVJh_NVL4Y4qEXGym47UCK5ZNZgVZYv"
+# --- Load solutions CSV from Google Drive ---
+file_id = "1yVJh_NVL4Y4YqEXGym47UCK5ZNZgVZYv"  # Replace with your file ID
 url = f"https://drive.google.com/uc?export=download&id={file_id}"
 response = requests.get(url)
 csv_text = response.text
 
-# Read CSV and clean column names
-solutions_df = pd.read_csv(StringIO(csv_text), header=0, on_bad_lines='skip')
-solutions_df.columns = solutions_df.columns.str.strip().str.lower()  # clean columns
-
-# Show debug info in Streamlit UI (remove later if you want)
-st.write("Columns found in CSV:", list(solutions_df.columns))
-st.write("First few rows of CSV:")
-st.write(solutions_df.head())
-
-# Stop if 'emotion' column missing
-if 'emotion' not in solutions_df.columns:
-    st.error("CSV is missing the 'emotion' column. Please check your file or rename the column.")
-    st.stop()
+if csv_text.strip().startswith('<'):
+    raise Exception("ERROR: Google Drive link is not returning CSV! Check your sharing settings.")
 
+solutions_df = pd.read_csv(StringIO(csv_text), header=0, on_bad_lines='skip')
 used_solutions = {emotion: set() for emotion in solutions_df['emotion'].unique()}
 
 negative_words = [
@@ -108,10 +101,6 @@ def get_unique_solution(emotion):
 def correct_spelling(text):
     return str(TextBlob(text).correct())
 
-def get_sentiment(text):
-    blob = TextBlob(text)
-    return blob.sentiment.polarity
-
 def is_negative_input(text):
     text_lower = text.lower()
     return any(word in text_lower for word in negative_words)
@@ -119,20 +108,14 @@ def is_negative_input(text):
 def get_emotion(user_input):
     if is_negative_input(user_input):
         return "sadness"
-    sentiment = get_sentiment(user_input)
     x = preprocess_input(user_input)
-    model.eval()
     with torch.no_grad():
-        probs = torch.stack([F.softmax(model(x), dim=1) for _ in range(5)])
-        avg_probs = probs.mean(dim=0)
-        prob, idx = torch.max(avg_probs, dim=1)
+        logits = model(x)
+        probs = F.softmax(logits, dim=1)
+        prob, idx = torch.max(probs, dim=1)
     pred_emotion = le.classes_[idx.item()]
     if prob.item() < 0.6:
         return "neutral"
-    if sentiment < -0.25 and pred_emotion == "happiness":
-        return "sadness"
-    if sentiment > 0.25 and pred_emotion == "sadness":
-        return "happiness"
     return pred_emotion
 
 def audio_to_text(audio_file):
@@ -147,59 +130,78 @@ def audio_to_text(audio_file):
     except Exception:
         return ""
 
-# --- Streamlit UI and Logic ---
+GLOBAL_CONVO_HISTORY = []
+USER_FEEDBACK_STATE = {}
 
-st.title("EmotiBot Connect (Streamlit)")
-
-if "history" not in st.session_state:
-    st.session_state.history = []
+def emoti_chat(audio, text, history_json=""):
+    # Get user input from voice or text
+    if text and text.strip():
+        user_input = text
+    elif audio is not None:
+        user_input = audio_to_text(audio)
+    else:
+        user_input = ""
 
-# User input
-audio_input = st.file_uploader("🎤 Upload audio message (wav, mp3)", type=["wav", "mp3"])
-text_input = st.text_input("💬 Or type your message here")
+    if not user_input.strip():
+        return "Please say something or type your message.", json.dumps(GLOBAL_CONVO_HISTORY[-5:], indent=2), ""
 
-if st.button("Send"):
+    user_input = correct_spelling(user_input)
 
-    # Get text from audio or text input
-    if text_input.strip():
-        user_input = text_input
-    elif audio_input is not None:
-        user_input = audio_to_text(audio_input)
-        if not user_input:
-            st.warning("Sorry, could not recognize speech from audio.")
-            user_input = ""
-    else:
-        user_input = ""
+    # Exit phrases
+    if user_input.lower().strip() in ["exit", "quit", "goodbye", "bye", "close"]:
+        return "Take care! I’m here whenever you want to talk. 👋", json.dumps(GLOBAL_CONVO_HISTORY[-5:], indent=2), gr.update(visible=False)
 
-    if user_input.strip() == "":
-        st.warning("Please say something or type your message.")
-    else:
-        # Correct spelling
-        user_input_corrected = correct_spelling(user_input)
+    # Feedback handling
+    user_id = "default_user"
+    state = USER_FEEDBACK_STATE.get(user_id, {"emotion": None, "pending": False})
 
-        # Handle exit phrases
-        if user_input_corrected.lower() in ["exit", "quit", "goodbye", "bye", "close"]:
-            st.success("Take care! I’m here whenever you want to talk. 👋")
+    if state["pending"]:
+        feedback = user_input.lower().strip()
+        GLOBAL_CONVO_HISTORY[-1]["feedback"] = feedback
+        if feedback == "no":
+            suggestion = get_unique_solution(state["emotion"])
+            reply = f"Here's another suggestion for you: {suggestion}\nDid this help? (yes/no/skip)"
+            USER_FEEDBACK_STATE[user_id]["pending"] = True
+            return reply, json.dumps(GLOBAL_CONVO_HISTORY[-5:], indent=2), ""
         else:
-            # Get emotion and response
-            pred_emotion = get_emotion(user_input_corrected)
-            support = random.choice(responses.get(pred_emotion, responses["neutral"]))
-            try:
-                suggestion = get_unique_solution(pred_emotion)
-            except Exception:
-                suggestion = get_unique_solution("neutral")
-
-            reply = f"{support}\n\nHere's a suggestion for you: {suggestion}\nDid this help? (yes/no/skip)"
-
-            # Save in history
-            st.session_state.history.append({
-                "user": user_input_corrected,
-                "emotion": pred_emotion,
-                "bot": reply,
-                "feedback": ""
-            })
-
-# Show conversation history (last 5)
-for chat in st.session_state.history[-5:]:
-    st.markdown(f"**You:** {chat['user']}")
-    st.markdown(f"**EmotiBot:** {chat['bot']}")
+            USER_FEEDBACK_STATE[user_id] = {"emotion": None, "pending": False}
+            return "How can I help you further?", json.dumps(GLOBAL_CONVO_HISTORY[-5:], indent=2), ""
+
+    # Normal user message
+    pred_emotion = get_emotion(user_input)
+    support = random.choice(responses.get(pred_emotion, responses["neutral"]))
+    try:
+        suggestion = get_unique_solution(pred_emotion)
+    except Exception:
+        suggestion = get_unique_solution("neutral")
+
+    reply = f"{support}\n\nHere's a suggestion for you: {suggestion}\nDid this help? (yes/no/skip)"
+
+    GLOBAL_CONVO_HISTORY.append({
+        "user_input": user_input,
+        "emotion": pred_emotion,
+        "bot_support": support,
+        "bot_suggestion": suggestion,
+        "feedback": ""
+    })
+    USER_FEEDBACK_STATE[user_id] = {"emotion": pred_emotion, "pending": True}
+    return reply, json.dumps(GLOBAL_CONVO_HISTORY[-5:], indent=2), ""
+
+import gradio as gr
+
+iface = gr.Interface(
+    fn=emoti_chat,
+    inputs=[
+        gr.Audio(type="filepath", label="🎤 Speak your message"),
+        gr.Textbox(lines=2, placeholder="Or type your message here...", label="💬 Type message"),
+        gr.Textbox(lines=1, value="", visible=False)  # Hidden, conversation history
+    ],
+    outputs=[
+        gr.Textbox(label="EmotiBot Reply"),
+        gr.Textbox(label="Conversation History (JSON)", visible=False)
+    ],
+    title="EmotiBot Connect",
+    description="Talk to EmotiBot using your voice or by typing. Detects your emotion, gives dynamic suggestions, and keeps conversation history!"
+)
+
+iface.launch()