Update app.py

app.py

@@ -9,7 +9,7 @@ import os
 
 # Load data and FAISS index
 def load_data_and_index():
-    docs_df = pd.read_pickle("data.pkl")  # Adjust path for HF Spaces
+    docs_df = pd.read_pickle("docs_with_embeddings (1).pkl")  # Adjust path for HF Spaces
     embeddings = np.array(docs_df['embeddings'].tolist(), dtype=np.float32)
     dimension = embeddings.shape[1]
     index = faiss.IndexFlatL2(dimension)
@@ -28,7 +28,7 @@ if not GEMINI_API_KEY:
 genai.configure(api_key=GEMINI_API_KEY)
 model = genai.GenerativeModel('gemini-2.0-flash')
 
-# Preprocess text
+# Preprocess text function
 def preprocess_text(text):
     text = text.lower()
     text = text.replace('\n', ' ').replace('\t', ' ')
@@ -36,7 +36,7 @@ def preprocess_text(text):
     text = ' '.join(text.split()).strip()
     return text
 
-# Retrieve top-k documents
+# Retrieve documents
 def retrieve_docs(query, k=5):
     query_embedding = minilm.encode([query], show_progress_bar=False)[0].astype(np.float32)
     distances, indices = index.search(np.array([query_embedding]), k)
@@ -44,24 +44,36 @@ def retrieve_docs(query, k=5):
     retrieved_docs['distance'] = distances[0]
     return retrieved_docs
 
-# Generate structured response
-def respond(message, system_message, max_tokens, temperature, top_p):
-    # Preprocess and retrieve
+# RAG pipeline integrated into respond function
+def respond(
+    message,
+    history: list[tuple[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,  # Keeping top_p as an input, though Gemini doesn’t use it directly
+):
+    # Preprocess the user message
     preprocessed_query = preprocess_text(message)
+    
+    # Retrieve relevant documents
     retrieved_docs = retrieve_docs(preprocessed_query, k=5)
-
-    # Combine retrieved texts
-    context = "\n".join([f"- *{row['label']}* ({row['source']}): {row['text']}" for _, row in retrieved_docs.iterrows()])
-
-    # Build prompt
+    context = "\n".join(retrieved_docs['text'].tolist())
+    
+    # Construct the prompt with system message, history, and RAG context
     prompt = f"{system_message}\n\n"
+    for user_msg, assistant_msg in history:
+        if user_msg:
+            prompt += f"User: {user_msg}\n"
+        if assistant_msg:
+            prompt += f"Assistant: {assistant_msg}\n"
     prompt += (
         f"Query: {message}\n"
         f"Relevant Context: {context}\n"
         f"Generate a short, concise, and to-the-point response to the query based only on the provided context."
     )
-
-    # Get Gemini response
+    
+    # Generate response with Gemini
     response = model.generate_content(
         prompt,
         generation_config=genai.types.GenerationConfig(
@@ -77,41 +89,37 @@ def respond(message, system_message, max_tokens, temperature, top_p):
         else:
             answer += "."
 
-    # Format output with Markdown
+    # Format the output with Gradio markdown for better readability
     formatted_answer = f"""
-**🩺 Patient Query:**  
-{message}
-
----
-
-**📚 Retrieved Context:**  
-{context}
-
----
-
-**🧠 Diagnosis / Suggestion:**  
-{answer}
-"""
-
-    return formatted_answer.strip()
-
-# Gradio app
-demo = gr.Interface(
-    fn=respond,
-    inputs=[
-        gr.Textbox(label="Your Query", placeholder="Enter your medical question here..."),
+    <div style='background-color:#f0f0f0; padding: 10px; border-radius: 5px;'>
+        <h3 style='color:#333; font-weight:bold;'>Assistant's Response:</h3>
+        <p style='color:#555;'>{answer}</p>
+    </div>
+    """
+    # Yield the formatted response
+    yield formatted_answer
+
+# Gradio Chat Interface
+demo = gr.ChatInterface(
+    respond,
+    additional_inputs=[
         gr.Textbox(
             value="You are a medical AI assistant diagnosing patients based on their query, using relevant context from past records of other patients.",
-            label="System Message"
+            label="System message"
         ),
-        gr.Slider(minimum=1, maximum=2048, value=150, step=1, label="Max New Tokens"),
+        gr.Slider(minimum=1, maximum=2048, value=150, step=1, label="Max new tokens"),
         gr.Slider(minimum=0.1, maximum=4.0, value=0.75, step=0.1, label="Temperature"),
-        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
+        gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.95,
+            step=0.05,
+            label="Top-p (nucleus sampling)",  # Included but not used by Gemini
+        ),
     ],
-    outputs=gr.Markdown(label="Diagnosis"),
-    title="🏥 Medical Assistant",
-    description="A simple medical assistant that diagnoses patient queries using AI and past records."
+    title="🏥 Medical Chat Assistant",
+    description="A chat-based medical assistant that diagnoses patient queries using AI and past records."
 )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()