Update app.py

app.py

@@ -1,10 +1,38 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.vectorstores import FAISS
+from langchain.text_splitter import CharacterTextSplitter
+from langchain.document_loaders import PyPDFLoader
+import os
 
-
+# Load the model client
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
+# Initialize vector store
+vector_store = None
+
+# Preload and process the PDF document
+PDF_PATH = "C:/Users/palanive/Documents/generalsymptoms.pdf"  # Path to the pre-defined PDF document
+
+def preload_pdf():
+    global vector_store
+
+    # Load PDF and extract text
+    loader = PyPDFLoader(PDF_PATH)
+    documents = loader.load()
+
+    # Split the text into smaller chunks for retrieval
+    text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
+    docs = text_splitter.split_documents(documents)
+
+    # Compute embeddings for the chunks
+    embeddings = HuggingFaceEmbeddings()
+    vector_store = FAISS.from_documents(docs, embeddings)
 
+    print(f"PDF '{PDF_PATH}' loaded and indexed successfully.")
+
+# Response generation
 def respond(
     message,
     history: list[tuple[str, str]],
@@ -13,8 +41,21 @@ def respond(
     temperature,
     top_p,
 ):
-    system_message = "You are a good listener. Hear the symptoms user are reporting and try to diagnose what kind of issue user might have and suggest this might be the issue and try to give some inputs to overcome this kind of issue and some good habits to avoid such issues , suggest avoiding negative thoughts, and guide through steps to manage the health issue. Discuss what's on your mind. limit the reply with 3 to 4 sentences"
-    messages = [{"role": "system", "content": system_message}]
+    global vector_store
+
+    if vector_store is None:
+        return "The PDF document is not loaded. Please check the code setup."
+
+    # Retrieve relevant chunks from the PDF
+    relevant_docs = vector_store.similarity_search(message, k=3)
+    context = "\n".join([doc.page_content for doc in relevant_docs])
+
+    # Combine system message, context, and user message
+    full_system_message = (
+        f"{system_message}\n\nContext from the document:\n{context}\n\n"
+    )
+
+    messages = [{"role": "system", "content": full_system_message}]
 
     for val in history:
         if val[0]:
@@ -34,34 +75,44 @@ def respond(
         top_p=top_p,
     ):
         token = message.choices[0].delta.content
-
         response += token
         yield response
 
+# Gradio interface
+demo = gr.Blocks()
 
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value = "You are going to act like a medical practitioner, and concise and point-specific speaker. You hear the symptoms and guide tips to overcome the issue and suggest some good habits, suggest avoiding negative thoughts, and guide through steps to manage the health issue. Discuss what's on your mind, or ask me for a quick health tips.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-
-    examples = [ 
-        ["I feel stressed."],
-        ["Can you guide me through a quick health tips?"],
-        ["How do I stop worrying about things I can't control?"]
-    ],
-    title = 'Diagnify 🕊️ '
-)
+with demo:
+    gr.Markdown("# Health Mate 🕊️ (RAG-based)")
 
+    chatbot = gr.ChatInterface(
+        respond,
+        additional_inputs=[
+            gr.Textbox(
+                value=(
+                    "You are going to act like a medical practitioner. Hear the symptoms, "
+                    "diagnose the disease, mention the disease name as heading, and suggest tips "
+                    "to overcome the issue. Base your answers on the provided document. limit the response to 3-4 sentences. list out the response point by point" 
+                ),
+                label="System message",
+            ),
+            gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
+            gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+            gr.Slider(
+                minimum=0.1,
+                maximum=1.0,
+                value=0.95,
+                step=0.05,
+                label="Top-p (nucleus sampling)",
+            ),
+        ],
+        examples=[
+            ["I feel stressed."],
+            ["Can you guide me through quick health tips?"],
+            ["How do I stop worrying about things I can't control?"],
+        ],
+        title="Health Mate 🕊️",
+    )
 
 if __name__ == "__main__":
+    preload_pdf()
     demo.launch()