Update app.py

app.py

@@ -4,14 +4,7 @@ import gradio as gr
 import time
 import heapq
 import re
-from utils import package_installer, context_identifier, query, convert_ipynb_to_html
-
-
-# package_installer('sentence_transformers')
-# package_installer('nbconvert')
-# package_installer('inflect')
-
-
+from utils import context_identifier, query, convert_ipynb_to_html
 from sentence_transformers import SentenceTransformer, util
 import nbconvert
 import nbformat
@@ -81,77 +74,62 @@ with gr.Blocks() as demo:
                     indexed_cells_list.append(f'# {i+1}th cell\n' + cell_addresses + itxt)
                     index_comments.append(f'# {i+1}th cell\n' + cell_addresses)
 
-            # print(indexed_cells_list)
+            emb_cells = embedding_model.encode(index_comments,convert_to_tensor=True)
+            emb_msg = embedding_model.encode(message,convert_to_tensor=True)
+            cosine_sim_0 = util.cos_sim(emb_msg,emb_cells)
+    
+            top_5_cells_scores = heapq.nlargest(5,cosine_sim_0[0])
+            top_5_cells = [indexed_cells_list[index] for index in sorted(list(cosine_sim_0[0]).index(score) for score in top_5_cells_scores)]
+    
+            top_2_chats = None
+            if hist_cop:
+                chat_history = [user_prompt_template(item[0]) + assistant_response_template(item[1]) for item in hist_cop]
+                # emb_formatted_user_msg = embedding_model.encode(formatted_user_msg,convert_to_tensor=True)
+                emb_chat_history = embedding_model.encode(chat_history,convert_to_tensor=True)
+                cosine_similarity_scores = util.cos_sim(emb_msg,emb_chat_history)
+                top_2_scores = heapq.nlargest(2,cosine_similarity_scores[0])
+                top_2_chats = [chat_history[i] for i in sorted(list(cosine_similarity_scores[0]).index(val) for val in top_2_scores)]
+                
+            similar_chat_history = ''
+            if top_2_chats:
+                for chats in top_2_chats:
+                    
+                    similar_chat_history += chats
+    
+            top_5_cells_string = '\n'.join(top_5_cells)
+    
+            if context == 'notebook_cell_context':
+                prompt = f"""
+                You are a coding assistant who clarifies queries based on python. You will be given two types of context. One type of context
+                consists of previous chat messages and the other consists of code from a jupyter notebook. Your task is to answer/explanation user 
+                query by picking relevant information from both context and coming up with the answer which explains the query. The user
+                query is delimited by ####.
+    
+                previous_chat_context: 
+                {similar_chat_history} 
+    
+                notebook_cell_context:
+                {top_5_cells_string}
+    
+                ####
+                {message}
+                ####
+                """
+            elif context == 'previous_cell_context':
+                prompt = f"""
+                You are a coding assistant who clarifies queries based on python. You will be given a context which consists of previous chat messages. 
+                Your task is to answer/explain user query based on the context.The user query is delimited by ####.
+    
+                previous_chat_context: 
+                {similar_chat_history} 
+    
+                ####
+                {message}
+                ####
+                """
 
         except:
-            pass
-
-        emb_cells = embedding_model.encode(index_comments,convert_to_tensor=True)
-        emb_msg = embedding_model.encode(message,convert_to_tensor=True)
-        cosine_sim_0 = util.cos_sim(emb_msg,emb_cells)
-
-        top_5_cells_scores = heapq.nlargest(5,cosine_sim_0[0])
-        top_5_cells = [indexed_cells_list[index] for index in sorted(list(cosine_sim_0[0]).index(score) for score in top_5_cells_scores)]
-
-        top_2_chats = None
-        if hist_cop:
-            chat_history = [user_prompt_template(item[0]) + assistant_response_template(item[1]) for item in hist_cop]
-            # emb_formatted_user_msg = embedding_model.encode(formatted_user_msg,convert_to_tensor=True)
-            emb_chat_history = embedding_model.encode(chat_history,convert_to_tensor=True)
-            cosine_similarity_scores = util.cos_sim(emb_msg,emb_chat_history)
-            top_2_scores = heapq.nlargest(2,cosine_similarity_scores[0])
-            top_2_chats = [chat_history[i] for i in sorted(list(cosine_similarity_scores[0]).index(val) for val in top_2_scores)]
-            
-        similar_chat_history = ''
-        if top_2_chats:
-            for chats in top_2_chats:
-                # formatted_assistant_msg = chats[1].replace(chats[0],'').strip().removesuffix('<|end|>')
-                similar_chat_history += chats
-
-        top_5_cells_string = '\n'.join(top_5_cells)
-        # print(top_5_cells_string)
-
-        if context == 'notebook_cell_context':
-            prompt = f"""
-            You are a coding assistant who clarifies queries based on python. You will be given two types of context. One type of context
-            consists of previous chat messages and the other consists of code from a jupyter notebook. Your task is to answer/explanation user 
-            query by picking relevant information from both context and coming up with the answer which explains the query. The user
-            query is delimited by ####.
-
-            previous_chat_context: 
-            {similar_chat_history} 
-
-            notebook_cell_context:
-            {top_5_cells_string}
-
-            ####
-            {message}
-            ####
-            """
-        elif context == 'previous_cell_context':
-            prompt = f"""
-            You are a coding assistant who clarifies queries based on python. You will be given a context which consists of previous chat messages. 
-            Your task is to answer/explain user query based on the context.The user query is delimited by ####.
-
-            previous_chat_context: 
-            {similar_chat_history} 
-
-            ####
-            {message}
-            ####
-            """
-        # elif context == 'notebook_cell_context':
-        #     prompt = f"""
-        #     You are a coding assistant who clarifies queries based on python. You will be given a context which consists of code from jupyter notebook
-        #     cells. Your task is to answer/explain user query based on the context.The user query is delimited by ####.
-
-        #     notebook_cell_context: 
-        #     {top_5_cells_string} 
-
-        #     ####
-        #     {message}
-        #     ####
-        #     """
+            prompt = message
 
         user_input = user_prompt_template(prompt)
         
@@ -173,8 +151,8 @@ with gr.Blocks() as demo:
             output += char
             time.sleep(0.05)
             yield output
-    # mkd = gr.Markdown('# ChatBot for Jupyter Notebook\n For more details: **link**')
-    file = gr.File(interactive=True,height=35,container=False)
+    
+    file = gr.File(interactive=True,container=False)
     chatbot = gr.ChatInterface(fn=chat,fill_height=False,additional_inputs=[file],stop_btn='Stop Generation',
                                description="[Read the Instructions here!](https://huggingface.co/spaces/noorulamean444/clma_try/blob/main/README.md)")