Upload 2 files

app.py

@@ -0,0 +1,137 @@
+import os
+import streamlit as st
+from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassification
+import numpy as np
+from groq import Groq
+import io
+import faiss
+
+# Groq API Key (and other initializations)
+groq_api_key = os.environ.get("GROQ_API_KEY")
+if groq_api_key is None:
+    st.error("GROQ_API_KEY environment variable not set.")
+    st.stop()
+
+try:
+    client = Groq(api_key=groq_api_key)
+except Exception as e:
+    st.error(f"Error initializing Groq client: {e}")
+    st.stop()
+
+try:
+    pubmedbert_tokenizer = AutoTokenizer.from_pretrained("microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext")
+    pubmedbert_model = AutoModelForSequenceClassification.from_pretrained("microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext")
+    pubmedbert_pipeline = pipeline('feature-extraction', model=pubmedbert_model, tokenizer=pubmedbert_tokenizer, device=-1)
+except Exception as e:
+    st.error(f"Error loading PubMedBERT: {e}")
+    st.stop()
+
+embedding_dim = 768
+index = faiss.IndexFlatL2(embedding_dim)
+
+if "all_conversations" not in st.session_state:
+    st.session_state.all_conversations = {}
+if "current_conversation_id" not in st.session_state:
+    st.session_state.current_conversation_id = 0
+if "current_conversation_messages" not in st.session_state:
+    st.session_state.current_conversation_messages = []
+if "embeddings" not in st.session_state:
+    st.session_state.embeddings = []
+
+# Functions
+def preprocess_query(query):
+    tokens = query.lower().split()
+    keywords = [keyword for keyword in tokens if keyword in ["seizure", "symptoms", "jerks", "confusion", "epilepsy"]]
+    is_medical_related = any(keyword in keywords for keyword in ["seizure", "symptoms", "jerks", "confusion", "epilepsy", "medical"])
+    return tokens, keywords, is_medical_related
+
+def generate_response(user_query):
+    tokens, keywords, is_medical_related = preprocess_query(user_query)
+    enhanced_query = " ".join(tokens)
+    symptom_insights = ""
+
+    conversation_history = "\n".join([f"{msg['role'].capitalize()}: {msg['content']}" for msg in st.session_state.current_conversation_messages])
+
+    if is_medical_related:
+        try:
+            pubmedbert_embeddings = pubmedbert_pipeline(user_query)
+            embedding_mean = np.mean(pubmedbert_embeddings[0], axis=0)
+            st.session_state.embeddings.append(embedding_mean)
+            index.add(np.array([embedding_mean]))
+            pubmedbert_insights = "PubMedBERT analysis..."
+            model_name = "PubMedBERT"
+            model_response = pubmedbert_insights
+            if "seizure" in keywords or "symptoms" in keywords:
+                remedy_recommendations = "\n\n**General Recommendations:**\n..."
+            else:
+                remedy_recommendations = ""
+        except Exception as e:
+            model_response = f"Error during PubMedBERT: {e}"
+            remedy_recommendations = ""
+    else:
+        model_name = "LLaMA 2 / Mistral 7B (via Groq)"
+        try:
+            prompt = f"""
+            Conversation History:
+            {conversation_history}
+
+            User: {user_query}
+            Bot:
+            """
+            chat_completion = client.chat.completions.create(
+                messages=[{"role": "user", "content": prompt}],
+                model="llama-3.3-70b-versatile",
+                stream=False,
+            )
+            model_response = chat_completion.choices[0].message.content.strip()
+        except Exception as e:
+            model_response = f"Error from Groq: {e}"
+        remedy_recommendations = ""
+
+    final_response = f"**Enhanced Query:** {enhanced_query}\n\nChatbot Analysis:...\n\nModel Response/Insights:\n{model_response}\n{remedy_recommendations}"
+    return final_response, model_response
+# Streamlit Interface (and other parts)
+st.set_page_config(page_title="Epilepsy Chatbot", layout="wide")
+st.markdown("<style>.chat-message.user {background-color: #e6f7ff; padding: 8px; border-radius: 8px; margin-bottom: 8px;}.chat-message.bot {background-color: #f0f0f0; padding: 8px; border-radius: 8px; margin-bottom: 8px;}.stTextArea textarea {background-color: #f8f8f8;}</style>", unsafe_allow_html=True)
+
+with st.sidebar:
+    st.title("Conversations")
+    if st.button("New Conversation"):
+        st.session_state.current_conversation_id += 1
+        st.session_state.current_conversation_messages = []
+        st.session_state.embeddings = []
+        index.reset()
+    for conv_id in st.session_state.all_conversations:
+        if st.button(f"Conversation {conv_id}"):
+            st.session_state.current_conversation_id = conv_id
+            st.session_state.current_conversation_messages = st.session_state.all_conversations[conv_id]
+            st.session_state.embeddings = []
+            index.reset()
+
+st.title("Epilepsy & Seizure Chatbot")
+st.write("Ask questions related to epilepsy and seizures.")
+
+for message in st.session_state.current_conversation_messages:
+    with st.chat_message(message["role"]):
+        st.markdown(message["content"])
+
+if prompt := st.chat_input("Enter your query here:"):
+    st.session_state.current_conversation_messages.append({"role": "user", "content": prompt})
+    with st.chat_message("user"):
+        st.markdown(prompt)
+
+    with st.chat_message("bot"):
+        with st.spinner("Generating response..."):
+            try:
+                full_response, model_only_response = generate_response(prompt)
+                st.markdown(model_only_response)
+                st.session_state.current_conversation_messages.append({"role": "bot", "content": model_only_response})
+            except Exception as e:
+                st.error(f"Error processing query: {e}")
+
+    st.session_state.all_conversations[st.session_state.current_conversation_id] = st.session_state.current_conversation_messages
+
+# Download Chat
+if st.session_state.current_conversation_messages:
+    conversation_text = "\n".join([f"{message['role'].capitalize()}: {message['content']}" for message in st.session_state.current_conversation_messages])
+    st.download_button("Download Chat", data=conversation_text, file_name=f"chat_history_{st.session_state.current_conversation_id}.txt")

requirements.txt

@@ -0,0 +1,6 @@
+transformers
+streamlit
+numpy
+torch
+groq
+faiss-cpu