rerank model

pages/Multimodal_Conversational_Search.py

@@ -13,49 +13,39 @@ import botocore.session
 import json
 import random
 import string
-#import rag_DocumentLoader 
+import rag_DocumentLoader 
 import rag_DocumentSearcher
+#import colpali
 import pandas as pd
 from PIL import Image 
 import shutil
 import base64
 import time
 import botocore
+#from langchain.callbacks.base import BaseCallbackHandler
+import streamlit_nested_layout
+#from IPython.display import clear_output, display, display_markdown, Markdown
 from requests_aws4auth import AWS4Auth
 import colpali
 from requests.auth import HTTPBasicAuth
-import warnings
 
 
-warnings.filterwarnings("ignore", category=DeprecationWarning)
 
 st.set_page_config(
     #page_title="Semantic Search using OpenSearch",
     layout="wide",
     page_icon="images/opensearch_mark_default.png"
 )
-
 parent_dirname = "/".join((os.path.dirname(__file__)).split("/")[0:-1])
 USER_ICON = "images/user.png"
 AI_ICON = "images/opensearch-twitter-card.png"
 REGENERATE_ICON = "images/regenerate.png"
 s3_bucket_ = "pdf-repo-uploads"
             #"pdf-repo-uploads"
-
-# @st.cache_resource
-# def get_polly_client():
-#     return boto3.client('polly',
-#         aws_access_key_id=st.secrets['user_access_key'],
-#         aws_secret_access_key=st.secrets['user_secret_key'],
-#         region_name='us-east-1'
-#     )
-
-# polly_client = get_polly_client()
-
+polly_client = boto3.Session(
+            region_name='us-east-1').client('polly')
 
 # Check if the user ID is already stored in the session state
-
-
 if 'user_id' in st.session_state:
     user_id = st.session_state['user_id']
     #print(f"User ID: {user_id}")
@@ -80,19 +70,23 @@ if "chats" not in st.session_state:
 
 if "questions_" not in st.session_state:
     st.session_state.questions_ = []
-
+    
+    
 if "show_columns" not in st.session_state:
     st.session_state.show_columns = False
+    
+if "answer_ready" not in st.session_state:
+    st.session_state.answer_ready = False
 
 if "answers_" not in st.session_state:
     st.session_state.answers_ = []
 
 if "input_index" not in st.session_state:
-    st.session_state.input_index = "globalwarming"#"hpijan2024hometrack"#"#"hpijan2024hometrack_no_img_no_table"
+    st.session_state.input_index = "hpijan2024hometrack"#"globalwarmingnew"#"hpijan2024hometrack_no_img_no_table"
     
 if "input_is_rerank" not in st.session_state:
     st.session_state.input_is_rerank = True
-
+    
 if "input_is_colpali" not in st.session_state:
     st.session_state.input_is_colpali = False
 
@@ -103,12 +97,25 @@ if "input_table_with_sql" not in st.session_state:
     st.session_state.input_table_with_sql = False
     
 if "input_query" not in st.session_state:
-    st.session_state.input_query="What is the projected energy percentage from renewable sources in future?"#"which city has the highest average housing price in UK ?"#"Which city in United Kingdom has the highest average housing price ?"#"How many aged above 85 years died due to covid ?"# What is the projected energy from renewable sources ?"
+    if(st.session_state.input_index == "globalwarming"):
+        st.session_state.input_query="What is the projected energy percentage from renewable sources in future ?"#"What is the projected energy percentage from renewable sources in future?"#"Which city in United Kingdom has the highest average housing price ?"#"How many aged above 85 years died due to covid ?"# What is the projected energy from renewable sources ?"
+    if(st.session_state.input_index == "hpijan2024hometrack"):
+        st.session_state.input_query="which city has the highest average housing price in UK ?"#"What is the projected energy percentage from renewable sources in future?"#"Which city in United Kingdom has the highest average housing price ?"#"How many aged above 85 years died due to covid ?"# What is the projected energy from renewable sources ?"
+    if(st.session_state.input_index == "covid19ie"):
+        st.session_state.input_query="How many aged above 85 years died due to covid ?"#"What is the projected energy percentage from renewable sources in future?"#"Which city in United Kingdom has the highest average housing price ?"#"How many aged above 85 years died due to covid ?"# What is the projected energy from renewable sources ?"
+
 
 if "input_rag_searchType" not in st.session_state:
-    st.session_state.input_rag_searchType = ["Vector Search"]  
+    st.session_state.input_rag_searchType = ["Vector Search"]
+    
 
 
+        
+region = 'us-east-1'
+bedrock_runtime_client = boto3.client('bedrock-runtime',region_name=region)
+output = []
+service = 'es'
+
 st.markdown("""
     <style>
     [data-testid=column]:nth-of-type(2) [data-testid=stVerticalBlock]{
@@ -120,10 +127,48 @@ st.markdown("""
     </style>
     """,unsafe_allow_html=True)
 
+################ OpenSearch Py client #####################
+    
+# credentials = boto3.Session().get_credentials()
+# awsauth = AWSV4SignerAuth(credentials, region, service)
+
+# ospy_client = OpenSearch(
+#     hosts = [{'host': 'search-opensearchservi-75ucark0bqob-bzk6r6h2t33dlnpgx2pdeg22gi.us-east-1.es.amazonaws.com', 'port': 443}],
+#     http_auth = awsauth,
+#     use_ssl = True,
+#     verify_certs = True,
+#     connection_class = RequestsHttpConnection,
+#     pool_maxsize = 20
+# )
+
+################# using boto3 credentials ###################
+
+
 credentials = boto3.Session().get_credentials()
-awsauth  = HTTPBasicAuth('master',st.secrets['ml_search_demo_api_access'])
+awsauth = AWS4Auth(credentials.access_key, credentials.secret_key, 'us-west-2', service, session_token=credentials.token)
 service = 'es'
 
+
+################# using boto3 credentials ####################
+
+
+
+# if "input_searchType" not in st.session_state:
+#     st.session_state.input_searchType = "Conversational Search (RAG)"
+
+# if "input_temperature" not in st.session_state:
+#     st.session_state.input_temperature = "0.001"
+
+# if "input_topK" not in st.session_state:
+#     st.session_state.input_topK = 200
+
+# if "input_topP" not in st.session_state:
+#     st.session_state.input_topP = 0.95
+
+# if "input_maxTokens" not in st.session_state:
+#     st.session_state.input_maxTokens = 1024
+
+
 def write_logo():
     col1, col2, col3 = st.columns([5, 1, 5])
     with col2:
@@ -151,9 +196,16 @@ if clear:
     st.session_state.questions_ = []
     st.session_state.answers_ = []
     st.session_state.input_query=""
+    # st.session_state.input_searchType="Conversational Search (RAG)"
+    # st.session_state.input_temperature = "0.001"
+    # st.session_state.input_topK = 200
+    # st.session_state.input_topP = 0.95
+    # st.session_state.input_maxTokens = 1024
 
 
 def handle_input():
+    # st.session_state.answer_ready = True
+    # st.session_state.show_columns = False  # reset column display
     print("Question: "+st.session_state.input_query)
     print("-----------")
     print("\n\n")
@@ -164,6 +216,11 @@ def handle_input():
         if key.startswith('input_'):
             inputs[key.removeprefix('input_')] = st.session_state[key]
     st.session_state.inputs_ = inputs
+    
+    #######
+    
+    
+    #st.write(inputs) 
     question_with_id = {
         'question': inputs["query"],
         'id': len(st.session_state.questions_)
@@ -171,7 +228,7 @@ def handle_input():
     st.session_state.questions_.append(question_with_id)
     if(st.session_state.input_is_colpali):
         out_ = colpali.colpali_search_rerank(st.session_state.input_query)
-        
+        #print(out_)
     else:
         out_ = rag_DocumentSearcher.query_(awsauth, inputs, st.session_state['session_id'],st.session_state.input_rag_searchType)
     st.session_state.answers_.append({
@@ -182,67 +239,122 @@ def handle_input():
         'table':out_['table']
     })
     st.session_state.input_query=""
+    
+
+    
+# search_type = st.selectbox('Select the Search type',
+#     ('Conversational Search (RAG)',
+#     'OpenSearch vector search', 
+#     'LLM Text Generation'
+#     ),
+   
+#     key = 'input_searchType',
+#     help = "Select the type of retriever\n1. Conversational Search (Recommended) - This will include both the OpenSearch and LLM in the retrieval pipeline \n (note: This will put opensearch response as context to LLM to answer) \n2. OpenSearch vector search - This will put only OpenSearch's vector search in the pipeline, \n(Warning: this will lead to unformatted results )\n3. LLM Text Generation - This will include only LLM in the pipeline, \n(Warning: This will give hallucinated and out of context answers_)"
+#     )
 
+# col1, col2, col3, col4 = st.columns(4)
+    
+# with col1:
+#     st.text_input('Temperature', value = "0.001", placeholder='LLM Temperature', key = 'input_temperature',help = "Set the temperature of the Large Language model. \n Note: 1. Set this to values lower to 1 in the order of 0.001, 0.0001, such low values reduces hallucination and creativity in the LLM response; 2. This applies only when LLM is a part of the retriever pipeline")
+# with col2:
+#     st.number_input('Top K', value = 200, placeholder='Top K', key = 'input_topK', step = 50, help = "This limits the LLM's predictions to the top k most probable tokens at each step of generation, this applies only when LLM is a prt of the retriever pipeline")
+# with col3:
+#     st.number_input('Top P', value = 0.95, placeholder='Top P', key = 'input_topP', step = 0.05, help = "This sets a threshold probability and selects the top tokens whose cumulative probability exceeds the threshold while the tokens are generated by the LLM")
+# with col4:
+#     st.number_input('Max Output Tokens', value = 500, placeholder='Max Output Tokens', key = 'input_maxTokens', step = 100, help = "This decides the total number of tokens generated as the final response. Note: Values greater than 1000 takes longer response time")
 
-def write_user_message(placeholder,md):
-    placeholder.empty()
-    with placeholder.container():
-        col1, col2 = st.columns([3,97])
+# st.markdown('---')
+
+
+def write_user_message(md):
+    col1, col2 = st.columns([3,97])
+    
+    with col1:
+        st.image(USER_ICON, use_column_width='always')
+    with col2:
+        #st.warning(md['question'])
+
+        st.markdown("<div style='color:#e28743';font-size:18px;padding:3px 7px 3px 7px;borderWidth: 0px;borderColor: red;borderStyle: solid;width: fit-content;height: fit-content;border-radius: 10px;font-style: italic;'>"+md['question']+"</div>", unsafe_allow_html = True)
+       
+
+
+def render_answer(question,answer,index,res_img):
+    
+    
+    col1, col2, col_3 = st.columns([4,74,22])
+    with col1:
+        st.image(AI_ICON, use_column_width='always')
+    with col2:
+        ans_ = answer['answer']
+        st.write(ans_)
+       
+        
         
-        with col1:
-            st.image(USER_ICON, use_column_width='always')
-        with col2:
-            #st.warning(md['question'])
-
-            st.markdown("<div style='color:#e28743';font-size:18px;padding:3px 7px 3px 7px;borderWidth: 0px;borderColor: red;borderStyle: solid;width: fit-content;height: fit-content;border-radius: 10px;font-style: italic;'>"+md['question']+"</div>", unsafe_allow_html = True)
-    return placeholder
-
-
-def render_answer(placeholder,question,answer,index,res_img):
-    placeholder.empty()
-    with placeholder.container():
-        col1, col2, col_3 = st.columns([4,74,22])
-        with col1:
-            st.image(AI_ICON, use_column_width='always')
-        with col2:
-            ans_ = answer['answer']
-            st.write(ans_)
-
-            # polly_response = polly_client.synthesize_speech(VoiceId='Joanna',
-            #                 OutputFormat='ogg_vorbis', 
-            #                 Text = ans_,
-            #                 Engine = 'neural')
-
-            # audio_col1, audio_col2 = st.columns([50,50])
-            # with audio_col1:
-            #     st.audio(polly_response['AudioStream'].read(), format="audio/ogg")
-            rdn_key_1 = ''.join([random.choice(string.ascii_letters)
-                                for _ in range(10)])
-            def show_maxsim(placeholder,dummy):
-                st.session_state.show_columns = True
-                st.session_state.input_query = ""
-                st.session_state.maxSimImages = colpali.img_highlight(st.session_state.top_img, st.session_state.query_token_vectors, st.session_state.query_tokens)
-                #handle_input()
-                # placeholder = st.empty()
-                # with placeholder.container():
-                placeholder.empty()
-                render_all(placeholder)    
-            if(st.session_state.input_is_colpali):
-                st.button("Show similarity map",key=rdn_key_1,on_click = show_maxsim,args=(placeholder,"default_img"))
+        # def stream_():
+        #     #use for streaming response on the client side
+        #     for word in ans_.split(" "):
+        #         yield word + " "
+        #         time.sleep(0.04)
+        #     #use for streaming response from Llm directly
+        #     if(isinstance(ans_,botocore.eventstream.EventStream)):
+        #         for event in ans_:
+        #             chunk = event.get('chunk')
                     
-        colu1,colu2,colu3 = st.columns([4,82,20])
-        with colu2:
-            @st.cache_data
-            def load_table_from_file(filepath):
-                df = pd.read_csv(filepath, skipinitialspace=True, on_bad_lines='skip', delimiter='`')
-                df.fillna(method='pad', inplace=True)
-                return df
-
-            with st.expander("Relevant Sources:"):
-                with st.container():
-                    if(len(res_img)>0):
-                        #with st.expander("Images:"):
+        #             if chunk:
+                        
+        #                 chunk_obj = json.loads(chunk.get('bytes').decode())
+                        
+        #                 if('content_block' in chunk_obj or ('delta' in chunk_obj and 'text' in chunk_obj['delta'])):
+        #                     key_ = list(chunk_obj.keys())[2]
+        #                     text = chunk_obj[key_]['text']
                             
+        #                     clear_output(wait=True)
+        #                     output.append(text)
+        #                     yield text
+        #                     time.sleep(0.04)
+            
+                
+        
+        # if(index == len(st.session_state.questions_)):
+        #     st.write_stream(stream_)
+        #     if(isinstance(st.session_state.answers_[index-1]['answer'],botocore.eventstream.EventStream)):
+        #         st.session_state.answers_[index-1]['answer'] = "".join(output)
+        # else:
+        #     st.write(ans_)
+        
+
+        polly_response = polly_client.synthesize_speech(VoiceId='Joanna',
+                        OutputFormat='ogg_vorbis', 
+                        Text = ans_,
+                        Engine = 'neural')
+
+        audio_col1, audio_col2 = st.columns([50,50])
+        with audio_col1:
+            st.audio(polly_response['AudioStream'].read(), format="audio/ogg")
+            
+        rdn_key_1 = ''.join([random.choice(string.ascii_letters)
+                              for _ in range(10)])
+        def show_maxsim():
+            st.session_state.show_columns = True
+            st.session_state.maxSimImages = colpali.img_highlight(st.session_state.top_img, st.session_state.query_token_vectors, st.session_state.query_tokens)
+            handle_input()
+            with placeholder.container():
+                render_all()    
+        if(st.session_state.input_is_colpali):
+            st.button("Show similarity map",key=rdn_key_1,on_click = show_maxsim)
+                
+        
+        
+        #st.markdown("<div style='font-size:18px;padding:3px 7px 3px 7px;borderWidth: 0px;borderColor: red;borderStyle: solid;border-radius: 10px;'>"+ans_+"</div>", unsafe_allow_html = True)
+    #st.markdown("<div style='color:#e28743';padding:3px 7px 3px 7px;borderWidth: 0px;borderColor: red;borderStyle: solid;width: fit-content;height: fit-content;border-radius: 10px;'><b>Relevant images from the document :</b></div>", unsafe_allow_html = True)
+    #st.write("")
+    colu1,colu2,colu3 = st.columns([4,82,20])
+    with colu2:
+        with st.expander("Relevant Sources:"):
+            with st.container():
+                if(len(res_img)>0):
+                    with st.expander("Images:"):
+                        
                         idx = 0
                         print(res_img)
                         for i in range(0,len(res_img)):
@@ -263,8 +375,13 @@ def render_answer(placeholder,question,answer,index,res_img):
                                         col3_,col4_,col5_ = st.columns([33,33,33])
                                         with col3_:
                                             st.image(res_img[i]['file'])
+                                    
+                                    
+
+                                
+                                
                             else:
-                                if(res_img[i]['file'].lower()!='none' and idx < 1):
+                                if(res_img[i]['file'].lower()!='none' and idx < 2):
                                     col3,col4,col5 = st.columns([33,33,33])
                                     cols = [col3,col4]
                                     img = res_img[i]['file'].split(".")[0]
@@ -273,86 +390,120 @@ def render_answer(placeholder,question,answer,index,res_img):
                                     with cols[idx]:
                                         
                                         st.image(parent_dirname+"/figures/"+st.session_state.input_index+"/"+img+".jpg")
+                                #st.write(caption)
                                 idx = idx+1
                         if(st.session_state.show_columns == True):
                             st.session_state.show_columns = False
-                    if(len(answer["table"] )>0):
-                        #with st.expander("Table:"):
-                        df = load_table_from_file(answer["table"][0]['name'])
+                #st.markdown("<div style='color:#e28743';padding:3px 7px 3px 7px;borderWidth: 0px;borderColor: red;borderStyle: solid;width: fit-content;height: fit-content;border-radius: 10px;'><b>Sources from the document:</b></div>", unsafe_allow_html = True)
+                if(len(answer["table"] )>0):
+                    with st.expander("Table:"):
+                        df = pd.read_csv(answer["table"][0]['name'],skipinitialspace = True, on_bad_lines='skip',delimiter='`')
+                        df.fillna(method='pad', inplace=True)
                         st.table(df)
-                    #with st.expander("Raw sources:"):
+                with st.expander("Raw sources:"):
                     st.write(answer["source"])
-               
+                
+        
+        
+    with col_3:
+        
+        #st.markdown("<div style='color:#e28743;borderWidth: 0px;borderColor: red;borderStyle: solid;width: fit-content;height: fit-content;border-radius: 5px;'><b>"+",".join(st.session_state.input_rag_searchType)+"</b></div>", unsafe_allow_html = True)
         
-    # with col_3:
-    #     if(index == len(st.session_state.questions_)):
-
-    #         rdn_key = ''.join([random.choice(string.ascii_letters)
-    #                           for _ in range(10)])
-    #         currentValue = ''.join(st.session_state.input_rag_searchType)+str(st.session_state.input_is_rerank)+str(st.session_state.input_table_with_sql)+st.session_state.input_index
-    #         oldValue = ''.join(st.session_state.inputs_["rag_searchType"])+str(st.session_state.inputs_["is_rerank"])+str(st.session_state.inputs_["table_with_sql"])+str(st.session_state.inputs_["index"])
-    #         def on_button_click():
-    #             if(currentValue!=oldValue or 1==1): 
-    #                 st.session_state.input_query = st.session_state.questions_[-1]["question"]
-    #                 st.session_state.answers_.pop()
-    #                 st.session_state.questions_.pop()
+       
+        
+        if(index == len(st.session_state.questions_)):
+
+            rdn_key = ''.join([random.choice(string.ascii_letters)
+                              for _ in range(10)])
+            # rdn_key_1 = ''.join([random.choice(string.ascii_letters)
+            #                   for _ in range(10)])
+            currentValue = ''.join(st.session_state.input_rag_searchType)+str(st.session_state.input_is_rerank)+str(st.session_state.input_table_with_sql)+st.session_state.input_index
+            oldValue = ''.join(st.session_state.inputs_["rag_searchType"])+str(st.session_state.inputs_["is_rerank"])+str(st.session_state.inputs_["table_with_sql"])+str(st.session_state.inputs_["index"])
+            #print("changing values-----------------")
+            def on_button_click():
+                if(currentValue!=oldValue or 1==1): 
+                    st.session_state.input_query = st.session_state.questions_[-1]["question"]
+                    st.session_state.answers_.pop()
+                    st.session_state.questions_.pop()
                     
-    #                 handle_input()
-    #                 with placeholder.container():
-    #                     render_all()
-    #         if("currentValue"  in st.session_state):
-    #             del st.session_state["currentValue"]
-
-    #         try:
-    #             del regenerate
-    #         except:
-    #             pass  
-    #         placeholder__ = st.empty()
-    #         placeholder__.button("🔄",key=rdn_key,on_click=on_button_click)
-            
+                    
+            # def show_maxsim():
+            #     st.session_state.show_columns = True
+            #     st.session_state.maxSimImages = colpali.img_highlight(st.session_state.top_img, st.session_state.query_token_vectors, st.session_state.query_tokens)
+            #     handle_input()
+            #     with placeholder.container():
+            #         render_all()
+                
+
+            if("currentValue"  in st.session_state):
+                del st.session_state["currentValue"]
+
+            try:
+                del regenerate
+            except:
+                pass  
+
+            placeholder__ = st.empty()
+            placeholder__.button("🔄",key=rdn_key,on_click=on_button_click)
+            #placeholder__.button("Show similarity map",key=rdn_key_1,on_click = show_maxsim)
         
+        
+     
 #Each answer will have context of the question asked in order to associate the provided feedback with the respective question
-def write_chat_message(placeholder, md, q,index):
+def write_chat_message(md, q,index):
     if(st.session_state.show_columns):
         res_img = st.session_state.maxSimImages
     else:
         res_img = md['image']
-    render_answer(placeholder,q,md,index,res_img)
+    #st.session_state['session_id'] = res['session_id']   to be added in memory
+    chat = st.container()
+    with chat:
+        #print("st.session_state.input_index------------------")
+        #print(st.session_state.input_index)
+        render_answer(q,md,index,res_img)
     
-def render_all(placeholder):  
+def render_all():  
     index = 0
     for (q, a) in zip(st.session_state.questions_, st.session_state.answers_):
         index = index +1
         
-        placeholder = write_user_message(placeholder,q)
-        write_chat_message(placeholder, a, q,index)
+        write_user_message(q)
+        write_chat_message(a, q,index)
 
 placeholder = st.empty()
-render_all(placeholder)
-        
+with placeholder.container():
+  render_all()
 
-
-        
 st.markdown("")
-col_2, col_3 = st.columns([75, 20])
+col_2, col_3 = st.columns([75,20])
+#col_1, col_2, col_3 = st.columns([7.5,71.5,22])
+# with col_1:
+#     st.markdown("<p style='padding:0px 0px 0px 0px; color:#FF9900;font-size:120%'><b>Ask:</b></p>",unsafe_allow_html=True, help = 'Enter the questions and click on "GO"')
+    
 with col_2:
-    input = st.text_input("Ask here", key="input_query", label_visibility="collapsed")
+    #st.markdown("")
+    input = st.text_input( "Ask here",label_visibility = "collapsed",key="input_query")
 with col_3:
-    play = st.button("Go",on_click=handle_input, key="play")
-
-
-##### Sidebar #####
+    #hidden = st.button("RUN",disabled=True,key = "hidden")
+    play = st.button("Go",on_click=handle_input,key = "play")
 with st.sidebar:
     st.page_link("app.py", label=":orange[Home]", icon="🏠")
     st.subheader(":blue[Sample Data]")
     coln_1,coln_2 = st.columns([70,30])
+    # index_select = st.radio("Choose one index",["UK Housing","Covid19 impacts on Ireland","Environmental Global Warming","BEIR Research"],
+    #                         captions = ['[preview](https://github.com/aws-samples/AI-search-with-amazon-opensearch-service/blob/b559f82c07dfcca973f457c0a15d6444752553ab/rag/sample_pdfs/HPI-Jan-2024-Hometrack.pdf)',
+    #                                     '[preview](https://github.com/aws-samples/AI-search-with-amazon-opensearch-service/blob/b559f82c07dfcca973f457c0a15d6444752553ab/rag/sample_pdfs/covid19_ie.pdf)',
+    #                                     '[preview](https://github.com/aws-samples/AI-search-with-amazon-opensearch-service/blob/b559f82c07dfcca973f457c0a15d6444752553ab/rag/sample_pdfs/global_warming.pdf)',
+    #                                     '[preview](https://github.com/aws-samples/AI-search-with-amazon-opensearch-service/blob/b559f82c07dfcca973f457c0a15d6444752553ab/rag/sample_pdfs/BEIR.pdf)'],
+    #                         key="input_rad_index")
     with coln_1:
-        index_select = st.radio("Choose one index",["Global Warming stats","UK Housing","Covid19 impacts on Ireland"],key="input_rad_index")
+        index_select = st.radio("Choose one index",["UK Housing","Global Warming stats","Covid19 impacts on Ireland"],key="input_rad_index")
     with coln_2:
         st.markdown("<p style='font-size:15px'>Preview file</p>",unsafe_allow_html=True)
-        st.write("[:eyes:](https://github.com/aws-samples/AI-search-with-amazon-opensearch-service/blob/b559f82c07dfcca973f457c0a15d6444752553ab/rag/sample_pdfs/global_warming.pdf)")
         st.write("[:eyes:](https://github.com/aws-samples/AI-search-with-amazon-opensearch-service/blob/b559f82c07dfcca973f457c0a15d6444752553ab/rag/sample_pdfs/HPI-Jan-2024-Hometrack.pdf)")
+        st.write("[:eyes:](https://github.com/aws-samples/AI-search-with-amazon-opensearch-service/blob/b559f82c07dfcca973f457c0a15d6444752553ab/rag/sample_pdfs/global_warming.pdf)")
         st.write("[:eyes:](https://github.com/aws-samples/AI-search-with-amazon-opensearch-service/blob/b559f82c07dfcca973f457c0a15d6444752553ab/rag/sample_pdfs/covid19_ie.pdf)")
+        #st.write("[:eyes:](https://github.com/aws-samples/AI-search-with-amazon-opensearch-service/blob/b559f82c07dfcca973f457c0a15d6444752553ab/rag/sample_pdfs/BEIR.pdf)")
     st.markdown("""
     <style>
     [data-testid=column]:nth-of-type(2) [data-testid=stVerticalBlock]{
@@ -364,23 +515,92 @@ with st.sidebar:
     </style>
     """,unsafe_allow_html=True)   
     with st.expander("Sample questions:"):
-        st.markdown("<span style = 'color:#FF9900;'>Global Warming stats</span> - What is the projected energy percentage from renewable sources in future?",unsafe_allow_html=True)
         st.markdown("<span style = 'color:#FF9900;'>UK Housing</span> - which city has the highest average housing price in UK ?",unsafe_allow_html=True)
+        st.markdown("<span style = 'color:#FF9900;'>Global Warming stats</span> - What is the projected energy percentage from renewable sources in future?",unsafe_allow_html=True)
         st.markdown("<span style = 'color:#FF9900;'>Covid19 impacts</span> - How many aged above 85 years died due to covid ?",unsafe_allow_html=True)
         
- 
+    # Initialize boto3 to use the S3 client.
+    s3_client = boto3.resource('s3')
+    bucket=s3_client.Bucket(s3_bucket_)
+
+    objects = bucket.objects.filter(Prefix="sample_pdfs/")
+    urls = []
+
+    client = boto3.client('s3')
+
+    for obj in objects:
+        if obj.key.endswith('.pdf'): 
+
+            # Generate the S3 presigned URL
+            s3_presigned_url = client.generate_presigned_url(
+                ClientMethod='get_object',
+                Params={
+                    'Bucket': s3_bucket_,
+                    'Key': obj.key
+                },
+                ExpiresIn=3600
+            )
+
+            # Print the created S3 presigned URL
+            print(s3_presigned_url)
+            urls.append(s3_presigned_url)
+            #st.write("["+obj.key.split('/')[1]+"]("+s3_presigned_url+")")
+            st.link_button(obj.key.split('/')[1], s3_presigned_url)
+    
+    
+    # st.subheader(":blue[Your multi-modal documents]")
+    # pdf_doc_ = st.file_uploader(
+    #     "Upload your PDFs here and click on 'Process'", accept_multiple_files=False)
+                    
+                
+    # pdf_docs = [pdf_doc_]
+    # if st.button("Process"):
+    #     with st.spinner("Processing"):
+    #         if os.path.isdir(parent_dirname+"/pdfs") == False:
+    #             os.mkdir(parent_dirname+"/pdfs")
+            
+    #         for pdf_doc in pdf_docs:
+    #             print(type(pdf_doc))
+    #             pdf_doc_name = (pdf_doc.name).replace(" ","_")
+    #             with open(os.path.join(parent_dirname+"/pdfs",pdf_doc_name),"wb") as f: 
+    #                 f.write(pdf_doc.getbuffer())  
+                    
+    #             request_ = { "bucket": s3_bucket_,"key": pdf_doc_name}
+    #             # if(st.session_state.input_copali_rerank):
+    #             #     copali.process_doc(request_)
+    #             # else:
+    #             rag_DocumentLoader.load_docs(request_)
+    #             print('lambda done')
+    #     st.success('you can start searching on your PDF')
+        
+    ############## haystach demo temporary addition ############    
+    # st.subheader(":blue[Multimodality]")
+    # colu1,colu2 = st.columns([50,50])
+    # with colu1:
+    #     in_images = st.toggle('Images', key = 'in_images', disabled = False)
+    # with colu2:
+    #     in_tables = st.toggle('Tables', key = 'in_tables', disabled = False)   
+    # if(in_tables):
+    #     st.session_state.input_table_with_sql = True
+    # else:
+    #     st.session_state.input_table_with_sql = False
         
      ############## haystach demo temporary addition ############       
     #if(pdf_doc_ is None or pdf_doc_ == ""):
     if(index_select == "Global Warming stats"):
-        st.session_state.input_index = "globalwarming"
+        st.session_state.input_index = "globalwarmingnew"
     if(index_select == "Covid19 impacts on Ireland"):
         st.session_state.input_index = "covid19ie"#"choosetheknnalgorithmforyourbillionscaleusecasewithopensearchawsbigdatablog"
     if(index_select == "BEIR"):
         st.session_state.input_index = "2104"
     if(index_select == "UK Housing"):
         st.session_state.input_index = "hpijan2024hometrack"
-   
+          
+    # custom_index = st.text_input("If uploaded the file already, enter the original file name", value = "")
+    # if(custom_index!=""):
+    #     st.session_state.input_index = re.sub('[^A-Za-z0-9]+', '', (custom_index.lower().replace(".pdf","").split("/")[-1].split(".")[0]).lower())
+    
+    
     
     st.subheader(":blue[Retriever]")
     search_type = st.multiselect('Select the Retriever(s)',
@@ -403,6 +623,7 @@ with st.sidebar:
     
     st.subheader(":blue[Multi-vector retrieval]")
     
+    #st.write("Dataset indexed: https://huggingface.co/datasets/vespa-engine/gpfg-QA")
     colpali_search_rerank =  st.checkbox('Try Colpali multi-vector retrieval on the [sample dataset](https://huggingface.co/datasets/vespa-engine/gpfg-QA)', key = 'input_colpali', disabled = False, value = False, help = "Checking this box will use colpali as the embedding model and retrieval is performed using multi-vectors followed by re-ranking using MaxSim")
         
     if(colpali_search_rerank):
@@ -413,6 +634,16 @@ with st.sidebar:
     
     with st.expander("Sample questions for Colpali retriever:"):
         st.write("1. Proportion of female new hires 2021-2023? \n\n 2. First-half 2021 return on unlisted real estate investments? \n\n 3. Trend of the fund's expected absolute volatility between January 2014 and January 2016? \n\n 4. Fund return percentage in 2017? \n\n 5. Annualized gross return of the fund from 1997 to 2008?")
-
     
-           
+        
+    # copali_rerank = st.checkbox("Search and Re-rank with Token level vectors",key = 'copali_rerank',help = "Enabling this option uses 'Copali' model's page level image embeddings to retrieve documents and MaxSim to re-rank the pages.\n\n Hugging Face Model: https://huggingface.co/vidore/colpali")
+        
+    # if(copali_rerank):
+    #     st.session_state.input_copali_rerank = True
+    # else:
+    #     st.session_state.input_copali_rerank = False
+    
+    
+                                       
+        
+