Create app.py

app.py

@@ -0,0 +1,199 @@
+import streamlit as st
+import os, json, datetime
+from langchain_community.vectorstores import FAISS
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain_groq import ChatGroq
+from langchain.chains import LLMChain
+from langchain.prompts import PromptTemplate
+from gtts import gTTS
+from pathlib import Path
+from dotenv import load_dotenv
+from sentence_transformers import SentenceTransformer, util
+import altair as alt
+import speech_recognition as sr
+from transformers import pipeline
+import torch
+
+# Load environment variables
+load_dotenv()
+GROQ_API_KEY = os.getenv("GROQ_API_KEY")
+
+CRISIS_KEYWORDS = ["suicide", "kill myself", "end it all", "worthless", "can't go on", "hurt myself", "self harm", "want to disappear", "no reason to live"]
+
+# Emotion detection
+def detect_emotion(text):
+    emotion_pipeline = pipeline(
+        "text-classification",
+        model="j-hartmann/emotion-english-distilroberta-base",
+        top_k=1,
+        device=-1
+    )
+    prediction = emotion_pipeline(text)[0][0]
+    return prediction['label'].lower(), prediction['score']
+
+# Streamlit UI
+st.set_page_config(page_title="DilBot - Emotional AI", page_icon="🧠")
+st.title("🧠 DilBot - Emotional AI Companion")
+st.markdown("A GenAI app that feels your words, finds comforting quotes, and replies with empathy.")
+
+# Quote categories
+quote_categories = {
+    "Grief": ["Grief is the price we pay for love.", "Tears are the silent language of grief."],
+    "Motivation": ["Believe in yourself and all that you are.", "Tough times never last, but tough people do."],
+    "Healing": ["Every wound has its own time to heal.", "It's okay to take your time to feel better."],
+    "Relationships": ["The best relationships are built on trust.", "Love is not about possession but appreciation."]
+}
+selected_category = st.selectbox(" Choose a quote theme:", list(quote_categories.keys()))
+uploaded_quotes = st.file_uploader(" Upload your own quotes (.txt)", type=["txt"])
+uploaded_audio = st.file_uploader("Upload a voice message (.wav)", type=["wav"])
+
+# Vectorstore
+@st.cache_resource(show_spinner=False)
+def build_vectorstore(quotes):
+    embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+    return FAISS.from_texts(quotes, embedding=embeddings)
+
+if uploaded_quotes:
+    custom_quotes = uploaded_quotes.read().decode("utf-8").splitlines()
+    vectorstore = build_vectorstore(custom_quotes)
+    current_quotes = custom_quotes
+    st.success("Custom quotes uploaded.")
+else:
+    default_quotes = quote_categories[selected_category]
+    vectorstore = build_vectorstore(default_quotes)
+    current_quotes = default_quotes
+
+# Crisis check
+def is_crisis(text):
+    return any(phrase in text.lower() for phrase in CRISIS_KEYWORDS)
+
+# Prompt setup
+prompt_template = PromptTemplate(
+    input_variables=["context", "user_input"],
+    template="""
+You are DilBot, an empathetic emotional support AI. Use the following emotional quote context to respond gently and supportively.
+
+Context:
+{context}
+
+User:
+{user_input}
+
+DilBot:"""
+)
+
+# LLM
+groq_llm = ChatGroq(api_key=GROQ_API_KEY, model="llama3-70b-8192")
+
+def get_response(user_input):
+    similar_docs = vectorstore.similarity_search(user_input, k=2)
+    context = "\n".join([doc.page_content for doc in similar_docs])
+    chain = LLMChain(llm=groq_llm, prompt=prompt_template)
+    return chain.run(context=context, user_input=user_input)
+
+# Save journal
+def save_journal(user_input, emotion, score, response, path="journal.json"):
+    entry = {
+        "date": str(datetime.date.today()),
+        "user_input": user_input,
+        "emotion": emotion,
+        "confidence": round(score * 100, 2),
+        "response": response
+    }
+    journal = []
+    if os.path.exists(path):
+        with open(path, "r") as f:
+            journal = json.load(f)
+    journal.append(entry)
+    with open(path, "w") as f:
+        json.dump(journal, f, indent=4)
+
+# Speak response
+def speak(text):
+    tts = gTTS(text)
+    path = Path("response.mp3")
+    tts.save(path)
+    st.audio(str(path), format="audio/mp3")
+
+# Transcribe voice
+def transcribe_audio_file(uploaded_audio):
+    recognizer = sr.Recognizer()
+    try:
+        with sr.AudioFile(uploaded_audio) as source:
+            audio_data = recognizer.record(source)
+            text = recognizer.recognize_google(audio_data)
+            return text
+    except Exception as e:
+        return f"Error: {str(e)}"
+
+# Store transcription
+if "transcribed_text" not in st.session_state:
+    st.session_state.transcribed_text = ""
+
+if uploaded_audio and st.button("Transcribe Voice"):
+    transcribed = transcribe_audio_file(uploaded_audio)
+    if transcribed.startswith("Error:"):
+        st.error(transcribed)
+    else:
+        st.session_state.transcribed_text = transcribed
+        st.success("Voice transcribed successfully!")
+
+# Input field
+user_input = st.text_area(
+    "💬 What's on your mind?",
+    value=st.session_state.transcribed_text,
+    height=100
+)
+
+final_input = user_input.strip() or st.session_state.transcribed_text.strip()
+
+# Single "Talk" Button
+if st.button("🧠 Talk to DilBot"):
+    if not final_input:
+        st.warning("Please enter or upload something to share.")
+    else:
+        with st.spinner("DilBot is thinking..."):
+            emotion, score = detect_emotion(final_input)
+            response = get_response(final_input)
+            save_journal(final_input, emotion, score, response)
+
+            st.success(f"**Emotion Detected:** {emotion.capitalize()} ({round(score*100)}% confidence)")
+            if is_crisis(final_input):
+                st.error("Crisis phrase detected! Please reach out to a mental health professional.")
+
+            if current_quotes:
+                model = SentenceTransformer("all-MiniLM-L6-v2")
+                quote_embeddings = model.encode(current_quotes, convert_to_tensor=True)
+                user_embedding = model.encode(final_input, convert_to_tensor=True)
+                sims = util.pytorch_cos_sim(user_embedding, quote_embeddings)[0]
+                best_match = sims.argmax().item()
+                selected_quote = current_quotes[best_match]
+                st.info(f"Quote for you: “{selected_quote}”")
+
+            st.markdown(f"**DilBot Says:** {response}")
+            speak(response)
+
+# Mood tracker
+if os.path.exists("journal.json"):
+    with open("journal.json", "r") as f:
+        journal_data = json.load(f)
+
+    df = {
+        "date": [entry["date"] for entry in journal_data],
+        "emotion": [entry["emotion"] for entry in journal_data],
+    }
+
+    st.markdown("**Daily Mood Tracker**")
+    chart = alt.Chart(alt.Data(values=df)).mark_bar().encode(
+        x='date:N',
+        y='count():Q',
+        color='emotion:N'
+    )
+    st.altair_chart(chart, use_container_width=True)
+
+    st.markdown("**Recent Conversations**")
+    for item in reversed(journal_data[-5:]):
+        st.markdown(f"`{item['date']}`\n- **You:** {item['user_input']}\n- **Emotion:** {item['emotion'].capitalize()} ({item['confidence']}%)\n- **DilBot:** {item['response']}")
+
+st.markdown("---")
+st.caption("Built Ahmad Sana Farooq( Member of CSG Hackathon Team)")