change ksize in RAG

RAG/rag_DocumentSearcher.py

@@ -304,7 +304,7 @@ def query_(awsauth,inputs, session_id,search_types):
     images_2 = []
     is_table_in_result = False
     df = []
-    for hit in hits[0:5]:
+    for id,hit in enumerate(hits[0:5]):
         
         
         if(hit["_source"]["raw_element_type"] == 'table'):
@@ -312,15 +312,15 @@ def query_(awsauth,inputs, session_id,search_types):
             is_table_in_result = True
             #table_res = invoke_models.read_from_table(hit["_source"]["table"],question) # use for complex analytical dataframe questions (uses panda at the background))
             df.append({'name':hit["_source"]["table"],'text':hit["_source"]["processed_element"]})
-            context_tables.append(hit["_source"]["processed_element"])#table_res+"\n\n"+
+            context_tables.append(id + " : Reference from a table :" + hit["_source"]["processed_element"])#table_res+"\n\n"+
             
         else:
             if(hit["_source"]["image"]!="None"):
                 with open(parent_dirname+'/figures/'+st.session_state.input_index+"/"+hit["_source"]["raw_element_type"].split("_")[1].replace(".jpg","")+"-resized.jpg", "rb") as read_img:
                     input_encoded = base64.b64encode(read_img.read()).decode("utf8")
-                context.append(invoke_models.generate_image_captions_llm(input_encoded,question))
+                context.append(id + " : Reference from a image :" + invoke_models.generate_image_captions_llm(input_encoded,question))
             else:
-                context.append(hit["_source"]["processed_element"])
+                context.append(id + " : Reference from a text chunk :" + hit["_source"]["processed_element"])
             
         if(hit["_source"]["image"]!="None"):
             images_2.append({'file':hit["_source"]["image"],'caption':hit["_source"]["processed_element"]})
@@ -340,7 +340,7 @@ def query_(awsauth,inputs, session_id,search_types):
     total_context = context_tables + context
     
 
-    llm_prompt = prompt_template.format(context=total_context[0],question=question)
+    llm_prompt = prompt_template.format(context="\n".join(total_context[0:3]),question=question)
     output = invoke_models.invoke_llm_model( "\n\nHuman: {input}\n\nAssistant:".format(input=llm_prompt) ,False)
     if(len(images_2)==0):
         images_2 = images