# views.py from django.shortcuts import render from django.contrib.auth.decorators import login_required from .models import ChatMessage from django.contrib.auth.models import User from django.http import JsonResponse from django.views.decorators.csrf import csrf_exempt import json from datetime import timedelta,datetime,timezone from time import time,sleep from langchain.chat_models import ChatOpenAI import os,sys # from langchain_community.vectorstores import FAISS import io # import uuid from langchain_core.messages import ( BaseMessage, message_to_dict, messages_from_dict, ) from glob import glob from langchain.prompts import ChatPromptTemplate,PromptTemplate from langchain.chains.combine_documents import create_stuff_documents_chain from langchain_openai import OpenAIEmbeddings from langchain_community.vectorstores import ElasticsearchStore from langchain.chains import ConversationalRetrievalChain from langchain.memory import ConversationBufferMemory from langchain.memory import ConversationBufferWindowMemory from langchain_core.messages.human import HumanMessage from langchain_core.messages.ai import AIMessage import requests from requests.auth import HTTPBasicAuth from elasticsearch import Elasticsearch,NotFoundError from collections import OrderedDict from decouple import config from openai import OpenAI from PIL import Image from io import BytesIO from langchain_core.messages.system import SystemMessage sys.path.append(os.path.join(os.path.dirname(__file__))) # from index_attached import start_indexing from langchain_core.documents import Document import os from Loader_functions_utils import DocumentLoader document_loader = DocumentLoader() from langchain.docstore.document import Document as BaseDocument from langchain.load.dump import dumps import tiktoken import base64 import openai import LogUtils logger = LogUtils.getRootLogger() def chatbot(request): return render(request, 'index.html') # Define the search query def get_messages_by_user_and_session(username,ind_name): query = { "query": { "bool": { "must": [ {"match": {"username": username}}, ] } }, "size":10000, "sort": [ {"lastmodified_at": {"order": "asc"}} ] } # Execute the search query response = es.search(index=ind_name, body=query) # Extract and return the messages messages = [hit['_source'] for hit in response['hits']['hits']] return messages os.environ["OPENAI_API_KEY"] = config('SECRET_KEY') logger.info('openapikey %s'%config('SECRET_KEY')) embeddings = OpenAIEmbeddings() es_index_name ='allibot_v2_gpt' chatname_index='get_chatid' imtext_index='textlens' doc_index='documents_up' # logger.info(es_index_name) es= Elasticsearch('https://elastic:8oKIqy312EBsAPzWT64NUzji@scic-elasticsearch.es.us-central1.gcp.cloud.es.io:443',request_timeout=300,retry_on_timeout=True) def check_elastic_status(): basic = HTTPBasicAuth('elastic', '8oKIqy312EBsAPzWT64NUzji') response=requests.get('https://scic-elasticsearch.es.us-central1.gcp.cloud.es.io', auth=basic) if response.status_code == 200: return True else: return False try : db = ElasticsearchStore( es_connection=es, index_name=['docx_new_page','pdf_new_page','additional_commercial_forms','docx_pdf_page','botcoach_index','html_unstructured1','excel_json_ind','pdf_json_ind','iso_forms_pdf_full_pdf','iso_forms_pdf_pagewise','json_others'], #index_name=['cml_study_guide', 'docx_pdf', 'new_pdf', 'pl_pilot_24_02_21_lg','botcoach_index','mcq','pdf_test','docx_test','excel_test','csv_test','xlsx','word_docx','pdf_docs','html_unstructured1'], embedding=embeddings, strategy=ElasticsearchStore.ExactRetrievalStrategy() ) except Exception as ex: logger.exception('Exception occured due to %s' % ex) if check_elastic_status(): db = ElasticsearchStore( es_connection=es, index_name=['docx_new_page','pdf_new_page','additional_commercial_forms','docx_pdf_page','botcoach_index','html_unstructured1','excel_json_ind','pdf_json_ind','iso_forms_pdf_full_pdf','iso_forms_pdf_pagewise','json_others'], embedding=embeddings, strategy=ElasticsearchStore.ExactRetrievalStrategy() ) def unique_sources_with_pages(meta_data): unique_sources = set() for item in meta_data: source = item.get('source', '') # print('main',source) filename = os.path.basename(source) # Get the file name with extension name_without_ext = os.path.splitext(filename)[0] # print('name_without_ext',name_without_ext) unique_sources.add(name_without_ext) # Convert the set of unique sources to a comma-separated string unique_sources_str = ', '.join(unique_sources) return unique_sources_str def get_chat_id_names_by_session_id(index_name, session_id): query = { "size": 5000, "query": {"term": {"session_id.keyword": session_id}} # Match the exact session ID } response = es.search(index=index_name, body=query) chat_id_name='' for hit in response['hits']['hits']: source = hit['_source'] if 'chat_id_name' in source: chat_id_name = source.get('chat_id_name') return chat_id_name def add_user_message_to_es(customer,username, session_id, user_prompt,index,user_token,modelname): # Create a document with the required fields doc = { 'candidate_name':customer, 'modelname':modelname, 'username': username, 'session_id': session_id, 'user_prompt': user_prompt, "created_at": round(time() * 1000), "lastmodified_at": round(time() * 1000), "history": json.dumps( message_to_dict(HumanMessage(content=user_prompt)), ensure_ascii=True, ), "human token" : user_token, } # Index the document in Elasticsearch res = es.index(index=index, body=doc) return res def add_assistant_message_to_es(customer,username, session_id, answer,assistant_token,modelname): # Create a document with the required fields doc = { 'candidate_name':customer, 'modelname':modelname, 'username': username, 'session_id': session_id, 'answer': answer, "created_at": round(time() * 1000), "lastmodified_at": round(time() * 1000), "history": json.dumps( message_to_dict(AIMessage(content=answer)), ensure_ascii=True, ), "assistant token" : assistant_token, "feedback" : "null", } # Index the document in Elasticsearch res = es.index(index=es_index_name, body=doc) return res def search_by_session_id(index_name, session_id): # Define the query query = { "size": 10000, "query": { "match": { "session_id": session_id } }, "sort": [ { "lastmodified_at": { "order": "asc" } } ] } # Perform the search response = es.search(index=index_name, body=query) return response['hits']['hits'] def document_to_dict(doc): # Convert Document to a string representation if storing as text return { 'page_content': doc.page_content, 'metadata': json.dumps(doc.metadata) if hasattr(doc, 'metadata') else '', # Convert metadata to JSON string } def add_document_ai_to_es(username, session_id, docs1, es_index_name,orderid): # Convert docs1 to a list of dictionaries as strings documents_dicts = [document_to_dict(doc) for doc in docs1] # Create a document with the required fields doc = { 'username': username, 'session_id': session_id, 'user_prompt': orderid, # Store as a JSON string "created_at": round(time() * 1000), "lastmodified_at": round(time() * 1000), "history": json.dumps(documents_dicts, ensure_ascii=True), } # Index the document in Elasticsearch res = es.index(index=es_index_name, body=doc) return res def update_lastmodified_at(index_name, doc_id): # Get current timestamp in milliseconds current_timestamp = int(datetime.now().timestamp() * 1000) # Define the update body update_body = { "doc": { "lastmodified_at": current_timestamp # Update with timestamp in milliseconds } } # Perform the update es.update(index=index_name, id=doc_id, body=update_body) def update_create_at_with_sessionid(index_name,session_id): documents = search_by_session_id(index_name, session_id) # Update 'lastmodified_at' field for each matched document for doc in documents: doc_id = doc['_id'] # Get the document ID update_lastmodified_at(index_name, doc_id) def fix_syntax_error(text): # Replace all double quotes with escaped double quotes fixed_text = text.replace('"', '\\"') return fixed_text def resize_image(image_path, size=(224, 224)): with Image.open(image_path) as img: img.thumbnail(size, Image.LANCZOS) buffer = BytesIO() img.save(buffer, format="BMP") return buffer.getvalue() def encode_image(image_data): return base64.b64encode(image_data).decode("utf-8") def get_current_iuser_idso_date(): return datetime.utcnow().isoformat() def store_chat_history(es, index_name, chat_entry): es.index(index=index_name, document=chat_entry) def retrieve_chat_data(indexname,session_id): # Fetch data from Elasticsearch response = es.get(index=indexname, id=session_id) if response["found"]: chat_data = response["_source"] return chat_data["documents"], chat_data["chat_history"] else: return None, None def chat_hist_img(base64_image,user_id,message,response,session_id,index_name_img,es): # Store the initial chat history chat_entry = { "username": user_id, "session_id": session_id, "created_at": round(time() * 1000), "lastmodified_at": round(time() * 1000), #"timestamp": get_current_iso_date(), # Ensure this is a string, not a function "messages": [ img_message_to_dict(message), img_message_to_dict(response) ], "img": base64_image } store_chat_history(es, index_name_img, chat_entry) def img_message_to_dict(message): if isinstance(message, HumanMessage): content = message.content if isinstance(content, list): content = " ".join([part['text'] for part in content if part['type'] == 'text']) return {"type": "human", "data": {"content": content}} elif isinstance(message, AIMessage): content = message.content if isinstance(content, list): content = " ".join([part['text'] for part in content if part['type'] == 'text']) return {"type": "ai", "data": {"content": content}} else: raise ValueError(f"Unexpected message type: {type(message)}") def img_message_from_dict(message_dict): # Check if the input is a list of messages if isinstance(message_dict, list): return [img_message_from_dict(m) for m in message_dict] message_type = message_dict.get("type") if message_type == "human": return HumanMessage(content=message_dict["data"]["content"]) elif message_type == "ai": return AIMessage(content=message_dict["data"]["content"]) else: raise ValueError(f"Got unexpected message type: {message_type}") def retrieve_chat_history(es, index_name, session_id): res = es.search(index=index_name, body={ "query": { "match": {"session_id": session_id} }, "size": 5000 }) return res['hits']['hits'] def chat_hist_txt(user_id,session_id,message,response,es,index_name): # Store the initial chat history chat_entry = { "username": user_id, "session_id": session_id, "created_at": round(time() * 1000), "lastmodified_at": round(time() * 1000), # "timestamp": get_current_iso_date(), # Ensure this is a string, not a function "messages": [ img_message_to_dict(message), img_message_to_dict(response) ], } store_chat_history(es, index_name, chat_entry) def encode_image1(image_path): with open(image_path, "rb") as image_file: return base64.b64encode(image_file.read()).decode("utf-8") def avi_to_base64(file): with open(file, "rb") as image_file: return base64.b64encode(image_file.read()).decode("utf-8") def resize_and_convert_image(image_path): with Image.open(image_path) as img: # Resize the image if it exceeds a certain size (e.g., width or height > 1024) max_size = (1024, 1024) img.thumbnail(max_size, Image.LANCZOS) # Convert BMP images to JPEG if img.format == 'BMP': img = img.convert('RGB') # Convert to RGB first converted_path = image_path.replace('.bmp', '.jpeg') img.save(converted_path, format='JPEG') os.remove(image_path) return converted_path else: # Save the resized image to a temporary file in the same format converted_path = image_path.replace(f".{img.format.lower()}", f".{img.format.lower()}") img.save(converted_path) return converted_path def add_documents(file_name1): _, extension_file_name1 = os.path.splitext(file_name1) resp=1 docs1='' # Load the first file try: if extension_file_name1.lower() == ".xlsx": docs1 = document_loader.excel_file_loader(file_name1) elif extension_file_name1.lower() == ".csv": docs1 = document_loader.csv_file_loader(file_name1) elif extension_file_name1.lower() == ".tsv": docs1 = document_loader.tsv_file_loader(file_name1) elif extension_file_name1.lower() in [".doc", ".docx"]: docs1 = document_loader.doc_docx_file_loader(file_name1) elif extension_file_name1.lower() == ".pdf": docs1 = document_loader.pdf_file_loader(file_name1) elif extension_file_name1.lower() == ".pptx": docs1 = document_loader.pptx_file_loader(file_name1) elif extension_file_name1.lower() == ".xml": docs1 = document_loader.xml_file_loader(file_name1) elif extension_file_name1.lower() == ".json": docs1 = document_loader.json_file_loader(file_name1) elif extension_file_name1.lower() == ".txt": docs1 = document_loader.text_file_loader(file_name1) elif extension_file_name1.lower() == ".md": docs1 = document_loader.markdown_file_loader(file_name1) elif extension_file_name1.lower() == ".html": docs1 = document_loader.html_file_loader(file_name1) else: raise ValueError(f"Unsupported file type: {extension_file_name1}") except Exception as e: resp=0 return resp,docs1 def query_and_combine(past_messages, user_query): llm = ChatOpenAI( model="gpt-4o", ) # Specify the search type (e.g., 'similarity' or 'exact_match') search_type = 'similarity' # Adjust this based on your requirements # Use the search method with the search_type argument retrieved_data = db.search(user_query, search_type=search_type) # Format the retrieved data formatted_data = f"{retrieved_data}" # Combine past messages with the retrieved data combined_messages = past_messages + [ AIMessage(content=formatted_data) ] # print("combined_messages",combined_messages) # Pass the combined data to ChatOpenAI response = llm.invoke(combined_messages) return response def delete_document_content(index_name, session_id,condition_check): # Define the query to match documents with the given session_id and user_prompt as 'upload1' query = { "query": { "bool": { "must": [ {"match": {"session_id": session_id}}, {"match": {"user_prompt": condition_check}} ] } } } # Perform the delete_by_query operation delete_response = es.delete_by_query(index=index_name, body=query) # Check if any documents were deleted deleted_count = delete_response['deleted'] if deleted_count > 0: print(f"Successfully deleted {deleted_count} documents with session_id '{session_id}' and user_prompt 'upload1'.") else: print("No documents found with the provided session ID and user_prompt 'upload1'.") def num_tokens_from_string(string: str, model: str) -> int: """Returns the number of tokens in a text string.""" encoding = tiktoken.encoding_for_model(model) num_tokens = len(encoding.encode(string)) return num_tokens def org_token(input_tokens, output_tokens, org_name): # Search for the document with the given organization name org_toks = es.search( index="organization_level_token_count", body={ 'query': { 'match': { 'organization_name': org_name } } } )['hits']['hits'] if org_toks: for doc in org_toks: doc_id = doc['_id'] doc_source = doc['_source'] # Retrieve existing input/output tokens with default values inp = doc_source.get('input_overall_tokens', 0) outp = doc_source.get('output_overall_tokens', 0) # Update the token values input_overall_tokens = input_tokens + inp output_overall_tokens = output_tokens + outp # Document to be updated update_doc = { 'doc': { 'output_overall_tokens': output_overall_tokens, 'input_overall_tokens': input_overall_tokens, 'overall_tokens': output_overall_tokens + input_overall_tokens, 'organization_name': org_name } } # Update the document using its ID res = es.update(index="organization_level_token_count", id=doc_id, body=update_doc) print(f"Updated document ID: {doc_id}, Response: {res}") @csrf_exempt def allibot(request): try: if request.method=='POST': action_type = request.POST.get('operational_type') # request.session.clear() if action_type == 'action1': username =request.POST.get('username') session_id=request.POST.get('session_id') candidate =request.POST.get('candidate_name') org_name=request.POST.get('organization_name') # chat_history_1 = ElasticsearchChatMessageHistory(es_connection=es,index=es_index_name, session_id=session_id) user_prompt = request.POST.get('message') update_create_at_with_sessionid(es_index_name, session_id) input_token_cost_gpt4_0125_preview = num_tokens_from_string(user_prompt, 'gpt-4-0125-preview') add_user_message_to_es(candidate,username,session_id, user_prompt,es_index_name,input_token_cost_gpt4_0125_preview,'gpt-4-0125-preview') template = """Answer the question in your own words as truthfully as possible from the context given to you. If there is any MCQ question explain why the choice is correct. If you do not know the answer to the question, simply respond with "I don't know. Can you ask another question". Response must be in and around 200 words. it must be in paragraph manner and it must not exceed 4 paragraphs. Give a line of space between the paragraphs. If questions are asked where there is no relevant context available, simply respond with "That's a great question! I'm not sure of the answer right now. Can you ask your question a different way? I am excited to assist you further!" Context: {context} {chat_history} Human: {question} Assistant:""" prompt = PromptTemplate(input_variables=["context", "chat_history", "question"], template=template) memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True, output_key='answer') result = es.search( index=es_index_name, query={"match": {"session_id": session_id}}, # Use match instead of term sort="lastmodified_at:asc", ) if result and len(result["hits"]["hits"]) > 0: items = [ json.loads(document["_source"]["history"]) for document in result["hits"]["hits"] ] else: items = [] session_chat_hist=messages_from_dict(items) if session_chat_hist: for i in range(0,len(session_chat_hist)-1): if type(session_chat_hist[i]) == HumanMessage: memory.chat_memory.add_user_message(session_chat_hist[i].content) elif type(session_chat_hist[i]) == AIMessage: memory.chat_memory.add_ai_message(session_chat_hist[i].content) qa = ConversationalRetrievalChain.from_llm( llm=ChatOpenAI(model="gpt-4-0125-preview"),retriever=db.as_retriever(search_kwargs={"k": 4}), memory=memory, combine_docs_chain_kwargs={'prompt': prompt},return_source_documents=True) formatted_response = qa(user_prompt) answer=formatted_response['answer'] print('answer',answer) # chat_history_1.add_ai_message(answer) output_token_cost_gpt4_0125_preview = num_tokens_from_string(answer, 'gpt-4-0125-preview') add_assistant_message_to_es(candidate,username,session_id, answer,output_token_cost_gpt4_0125_preview,'gpt-4-0125-preview') org_token(input_token_cost_gpt4_0125_preview, output_token_cost_gpt4_0125_preview, org_name) source=formatted_response['source_documents'] # print('source',source) meta_data=[] for sor in source: # print('meta',sor.metadata) meta_data.append(sor.metadata) unique_entries_mod_str = unique_sources_with_pages(meta_data) try: json_str=dumps(''.join([doc._lc_kwargs['page_content']+'_lc_kwargs' for doc in formatted_response['source_documents']])) user_ques=dumps(user_prompt) answer_prompt=dumps(answer) unique_entries_str=dumps(unique_entries_mod_str) cmd=f"python3 /var/www/html/AlliBotV5_chatbot/AlliBotV50/Quesgen_answer_relevancy.py {user_ques} {answer_prompt} {json_str} {unique_entries_str}" # logger.info(cmd) os.popen(cmd) except Exception as e: logger.exception('Exception occured due to %s' % e) return JsonResponse({'response': answer,'session_id':session_id}) except Exception as e: logger.exception('Exception occured due to %s' % e) return JsonResponse({'error': str(e)}) @csrf_exempt def doccompare(request): try: if request.method=='POST': action_type = request.POST.get('operational_type') if action_type == 'action1': username = request.POST.get('username') session_id = request.POST.get('session_id') user_prompt = request.POST.get('message') candidate =request.POST.get('candidate_name') org_name=request.POST.get('organization_name') update_create_at_with_sessionid(es_index_name, session_id) input_token_cost_gpt4o = num_tokens_from_string(user_prompt, 'gpt-4o') add_user_message_to_es(candidate,username,session_id, user_prompt,es_index_name,input_token_cost_gpt4o,'gpt-4o') template = """Answer the question in your own words as truthfully as possible from the context given to you. You are a helpful assistant. Include the filename within your response where relevant. Response must be in and around 200 words. It must be in paragraph manner and it must not exceed 4 paragraphs. Give a line of space between the paragraphs. If questions are asked where there is no relevant context available, simply respond with "That's a great question! I'm not sure of the answer right now. Can you ask your question a different way? I am excited to assist you further!" Context: {context} {chat_history} Human: {question} Assistant:""" memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True, output_key='answer') result = es.search( index=es_index_name, query={"match": {"session_id": session_id}}, sort="lastmodified_at:asc", size=1000, ) if result and len(result["hits"]["hits"]) > 0: items = [ json.loads(document["_source"]["history"]) for document in result["hits"]["hits"] ] else: items = [] # print('items',items) session_chat_hist = messages_from_dict(items) # print('hist memory',session_chat_hist) session_chat_hist=messages_from_dict(items) if session_chat_hist: for i in range(0,len(session_chat_hist)): if type(session_chat_hist[i]) == HumanMessage: memory.chat_memory.add_user_message(session_chat_hist[i].content) elif type(session_chat_hist[i]) == AIMessage: memory.chat_memory.add_ai_message(session_chat_hist[i].content) result = es.search( index=doc_index, query={"match": {"session_id": session_id}}, sort="lastmodified_at:asc", ) if result and len(result["hits"]["hits"]) > 0: doc_items = [ json.loads(document["_source"]["history"]) for document in result["hits"]["hits"] ] else: doc_items = [] uploaded_document_docs = [] if doc_items: for i in doc_items: page_content = str(i[0].get('page_content')) if i[0].get('page_content') else "" # Ensure metadata is a dictionary metadata = i[0].get('metadata') if not isinstance(metadata, dict): # Handle the case where metadata is not a dict # Example: convert to dict if it's a stringified JSON try: metadata = json.loads(metadata) if isinstance(metadata, str) else {} except (json.JSONDecodeError, TypeError): metadata = {} uploaded_document_docs.append(BaseDocument(page_content=page_content, metadata=metadata)) db = ElasticsearchStore( es_connection=es, # index_name=['docx_new_page','pdf_new_page','additional_commercial_forms','docx_pdf_page','botcoach_index','html_unstructured1','excel_json_ind','pdf_json_ind','iso_forms_pdf_full_pdf','iso_forms_pdf_pagewise','json_others'], index_name=['cml_study_guide', 'docx_pdf', 'new_pdf', 'pl_pilot_24_02_21_lg','botcoach_index','mcq','pdf_test','docx_test','excel_test','csv_test','xlsx','word_docx','pdf_docs','html_unstructured1'], embedding=embeddings, strategy=ElasticsearchStore.ExactRetrievalStrategy() ) primary_docs= db.as_retriever(search_kwargs={"k": 2}).get_relevant_documents(user_prompt) primary_docs_text = [] for doc in primary_docs: # Parse the JSON content from page_content primary_docs_text.append(BaseDocument(page_content=doc.page_content)) # or handle differently combined_docs = uploaded_document_docs + primary_docs_text # print('uploaded_document_docs',uploaded_document_docs) # print('combined_docs',combined_docs) client = OpenAI() memory_variables = memory.load_memory_variables({"input": ""}) # Use an empty input since we only need chat history chat_history = memory_variables['chat_history'] print('chat_history',chat_history) formatted_template = template.format( context=combined_docs, chat_history=chat_history, question=user_prompt ) # Stream the response response = client.chat.completions.create( model="gpt-4o", messages=[{"role": "user", "content": formatted_template}], stream=False, ) assistant_response = response.choices[0].message.content output_token_cost_gpt4o = num_tokens_from_string(assistant_response, 'gpt-4o') add_assistant_message_to_es(candidate,username,session_id, assistant_response,output_token_cost_gpt4o,'gpt-4o') org_token(input_token_cost_gpt4o, output_token_cost_gpt4o, org_name) return JsonResponse({'response': assistant_response,'session_id':session_id}) elif action_type == 'action2': # print('entered second condition') files = request.FILES.getlist('files') conver_id=request.POST.get('session_id') username= request.POST.get('user') # print("conver_id",conver_id) if not files: return JsonResponse({'status': 'error', 'message': 'No files uploaded'}, status=400) upload_dir = os.path.join(os.path.dirname(__file__))+'/upload_files' if not os.path.exists(upload_dir): os.makedirs(upload_dir) for file in files: file_path = os.path.join(upload_dir, file.name) with open(file_path, 'wb+') as destination: for chunk in file.chunks(): destination.write(chunk) files_dir = glob(upload_dir+'/*') # ind_response=start_indexing(upload_dir,imtext_index,conver_id,username) delete_document_content(doc_index,conver_id,'upload1') # delete_document_content(es_index_name,conver_id) ind_response1,docs1=add_documents(files_dir[0]) file=files_dir[0].split('/')[-1] # print('docs1',docs1) if docs1: for doc in docs1: doc.page_content = { f"context of the first uploaded {file} document": doc.page_content } add_document_ai_to_es( username, conver_id, docs1,doc_index,'upload1') for f in files_dir: os.remove(f) if ind_response1==1: res_msg=str(len(files))+' uploads complete' return JsonResponse({'status': 'success', 'message':res_msg}) else: res_msg='unable to process upload1 documents.' return JsonResponse({'status': 'failed', 'message':res_msg}) elif action_type == 'action3': # print('entered third condition') files = request.FILES.getlist('files') conver_id=request.POST.get('session_id') username= request.POST.get('user') print("conver_id",conver_id) if not files: return JsonResponse({'status': 'error', 'message': 'No files uploaded'}, status=400) upload_dir = os.path.join(os.path.dirname(__file__))+'/upload_files' if not os.path.exists(upload_dir): os.makedirs(upload_dir) for file in files: file_path = os.path.join(upload_dir, file.name) with open(file_path, 'wb+') as destination: for chunk in file.chunks(): destination.write(chunk) delete_document_content(doc_index,conver_id,'upload2') files_dir = glob(upload_dir+'/*') ind_response2,docs2=add_documents(files_dir[0]) file=files_dir[0].split('/')[-1] if docs2: for doc in docs2: doc.page_content = { f"context of the second uploaded {file} document": doc.page_content } add_document_ai_to_es( username, conver_id, docs2,doc_index,'upload2') for f in files_dir: os.remove(f) # print("response status",ind_response2) if ind_response2==1: res_msg=str(len(files))+' uploads complete' return JsonResponse({'status': 'success', 'message':res_msg}) else: res_msg='unable to process upload2 documents.' return JsonResponse({'status': 'failed', 'message':res_msg}) except Exception as e: logger.exception('Exception occured due to %s' % e) return JsonResponse({'error': str(e)}) @csrf_exempt def imgtotext(request): try: if request.method=='POST': action_type = request.POST.get('operational_type') # global session_id llm = ChatOpenAI( model="gpt-4o", ) # request.session.clear() if action_type == 'action1': # print('image to text section') username =request.POST.get('username') session_id=request.POST.get('session_id') user_prompt = request.POST.get('message') candidate =request.POST.get('candidate_name') org_name=request.POST.get('organization_name') update_create_at_with_sessionid(imtext_index, session_id) input_token_cost_gpt4_0125_preview = num_tokens_from_string(user_prompt, 'gpt-4o') add_user_message_to_es(candidate,username,session_id, user_prompt,es_index_name,input_token_cost_gpt4_0125_preview,'gpt-4o') message = HumanMessage( content=[ {"type": "text", "text": user_prompt}, ], ) # Retrieve chat history for context chat_history = retrieve_chat_history(es, imtext_index, session_id) # print('chat_history',chat_history) try: # Convert retrieved history back to messages past_messages = [] for entry in chat_history: # Check if '_source' contains 'messages' if "_source" in entry and "messages" in entry["_source"]: messages = img_message_from_dict(entry["_source"]["messages"]) # Ensure messages is a list and extend the past_messages list if isinstance(messages, list): past_messages.extend(messages) # Add the new user input to the conversation new_message = message past_messages.append(new_message) print('past_messages',past_messages) # Get AI response response = query_and_combine(past_messages,user_prompt) # response = llm.invoke(past_messages) chat_hist_txt(username,session_id,message,response,es,imtext_index) # add_assistant_message_to_es(username,session_id, response.content) output_token_cost_gpt4_0125_preview = num_tokens_from_string(response.content, 'gpt-4o') add_assistant_message_to_es(candidate,username,session_id, response.content,output_token_cost_gpt4_0125_preview,'gpt-4o') org_token(input_token_cost_gpt4_0125_preview, output_token_cost_gpt4_0125_preview, org_name) return JsonResponse({'response': response.content,'session_id':session_id}) except: new_message = message # response = llm.invoke(past_messages) response = query_and_combine(past_messages,user_prompt) chat_hist_txt(username,session_id,message,response,es,imtext_index) # add_assistant_message_to_es(username,session_id, response.content) output_token_cost_gpt4_0125_preview = num_tokens_from_string(response.content, 'gpt-4o') add_assistant_message_to_es(candidate,username,session_id, response.content,output_token_cost_gpt4_0125_preview,'gpt-4o') org_token(input_token_cost_gpt4_0125_preview, output_token_cost_gpt4_0125_preview, org_name) return JsonResponse({'response': response.content,'session_id':session_id}) elif action_type == 'action2': # print('entered second condition') files = request.FILES.getlist('files') conver_id=request.POST.get('session_id') user=request.POST.get('username') # print("conver_id",conver_id) if not files: return JsonResponse({'status': 'error', 'message': 'No files uploaded'}, status=400) upload_dir = os.path.join(os.path.dirname(__file__))+'/image_pth' if not os.path.exists(upload_dir): os.makedirs(upload_dir) for file in files: file_path = os.path.join(upload_dir, file.name) with open(file_path, 'wb+') as destination: for chunk in file.chunks(): destination.write(chunk) # ind_response=Image_indexing(upload_dir,'image_datas',conver_id) image_files = glob(upload_dir + '/*.jpg') + glob(upload_dir + '/*.jpeg') + glob(upload_dir + '/*.png') + glob(upload_dir + '/*.bmp') # Use glob to get all supported image files (jpg, jpeg, png, bmp) supported_formats = ['.jpg', '.jpeg', '.png', '.bmp'] image_files = [] for ext in supported_formats: image_files.extend(glob(os.path.join(upload_dir, f'*{ext}'))) # Iterate through each image file for image_file in image_files: # Resize and convert the image if necessary processed_image_path = resize_and_convert_image(image_file) try: for i in glob(upload_dir+'/*'): if i.lower().endswith(('.jpg', '.jpeg', '.png', '.bmp')): im64=encode_image1(i) message = HumanMessage( content=[ {"type": "text", "text": "Extract the texts in the image."}, { "type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{im64}"}, }, ], ) response = llm.invoke([message]) chat_hist_img(im64,user,message,response,conver_id,imtext_index,es) else: ind_response1,docs1=add_documents(i) file=i.split('/')[-1] message = HumanMessage( content=[ {"type": "text", "text": "context of the uploaded {file} document"}, ], ) chat_hist_txt(user,conver_id,message,docs1,es,imtext_index) ind_response=1 except Exception as e: ind_response=0 files_dir = glob(upload_dir+'/*') for f in files_dir: os.remove(f) if ind_response==1: res_msg=str(len(files))+' uploads complete' return JsonResponse({'status': 'success', 'message':res_msg,'result':response.content}) else: res_msg='Unable to process image.' return JsonResponse({'status': 'failed', 'message':res_msg,'result':''}) except Exception as e: logger.exception('Exception occured due to %s' % e) return JsonResponse({'error': str(e)}) @csrf_exempt def text_to_speech(request): if request.method == 'POST': print('entered') input_text = request.POST.get('message') if not input_text: return JsonResponse({'error': 'No input text provided'}, status=400) try: print('input_text',input_text) # Fetch TTS from OpenAI response = openai.audio.speech.create( model="tts-1", voice="alloy", input=input_text ) # Stream the audio content to memory audio_content = response.content # Convert the audio content to base64 base64_encoded_audio = base64.b64encode(audio_content).decode('utf-8') return JsonResponse({'audio': base64_encoded_audio}) except Exception as e: return JsonResponse({'error': str(e)}, status=500) return JsonResponse({'error': 'Invalid request method'}, status=405) @csrf_exempt def update_thumbs_flag(request): # Search for the document using session ID and answer field try: # session_id,index_name,answer, if request.method == 'POST': print('thumbs flag') session_id = request.POST.get('session_id') index_name =request.POST.get('indexname') answer=request.POST.get('answer') thumbs_flag=request.POST.get('thumbs_flag') fixed_answer = fix_syntax_error(answer) sleep(10) response = es.search( index=index_name, body={ "query": { "bool": { "must": [ {"term": {"session_id": session_id}}, # assuming session_id is a keyword field # { # "match": { # "history": { # "query": fixed_answer, # "operator": "AND" # Match with AND operator # } # } # } ] } }, "size": 1000 # Adjust size to the number of documents you expect, e.g., 1000 } ) # print('response length',len(response)) # Check if any documents are found total_docs = response['hits']['total']['value'] logger.info('total_docs%s'%total_docs) if total_docs > 0: for hit in response['hits']['hits']: doc_id = hit['_id'] # Get document ID if 'answer' in hit['_source'] : doc_source = hit['_source']['answer'] # Get the document source if doc_source==answer: # print('yes') es.update( index=index_name, id=doc_id, body={ "doc": { "feedback": thumbs_flag, # Update the feedback flag with the value 'good', 'bad', or 'no feedback' }, "doc_as_upsert": True # Create the document if it doesn't exist } ) return JsonResponse({'status': 'success', 'message': 'Document updated'}) else: return JsonResponse({'status': 'failure', 'message': 'No documents found'}, status=404) except NotFoundError: print(f"Error: No documents found for session_id {session_id} and answer '{answer}'.") @csrf_exempt def speech_to_text(request): if request.method == 'POST': if 'file' in request.FILES: audio_file = request.FILES['file'] try: # Convert InMemoryUploadedFile to bytes audio_bytes = audio_file.read() client = OpenAI() transcription = client.audio.transcriptions.create( model="whisper-1", file=(audio_file.name, audio_bytes), response_format="text" ) return JsonResponse({'transcription': transcription}) except Exception as e: return JsonResponse({'error': str(e)}, status=500) else: return JsonResponse({'error': 'No file provided'}, status=400) else: return JsonResponse({'error': 'Invalid request method'}, status=405) def today_messages(request): try: t=[] username = request.headers.get('X-Username') if username =='image_logs': test = get_messages_by_user_and_session(username,imtext_index) else: test = get_messages_by_user_and_session(username,es_index_name) # Get current date and time in UTC current_datetime = datetime.now(timezone.utc) # Calculate the start of today and yesterday in UTC start_of_today = current_datetime.replace(hour=0, minute=0, second=0, microsecond=0) for item in test: lastmodified_at_datetime = datetime.fromtimestamp(item['lastmodified_at'] / 1000, tz=timezone.utc) if start_of_today <= lastmodified_at_datetime <= current_datetime: t.append(item) if t: #session_ids = list(set(d["session_id"] for d in t)) if username =='image_logs': session_ids=list(OrderedDict.fromkeys(d["session_id"] for d in t)) yest_dict = {} items=[] for id_user in session_ids[::-1]: result = es.search( index=imtext_index, query={"match": {"session_id": id_user}}, # Use match instead of term sort="lastmodified_at:asc", ) tot_hist=[] chat_h={} imtex={} for document in result["hits"]["hits"]: if len(document["_source"])==6: chat_h= { 'system': document["_source"]['img']} if document["_source"]["messages"][0]['type']=='ai': imtex={'assistant': document["_source"]["messages"][0]['data']['content']} elif document["_source"]["messages"][1]['type']=='ai': imtex={'assistant': document["_source"]["messages"][1]['data']['content']} chat_h=chat_h,imtex elif len(document["_source"])==5: if document["_source"]['messages'][0]['type']=='ai': chat_h= {'user': document["_source"]['messages'][1]['data']['content']},{'assistant': document["_source"]['messages'][0]['data']['content']} else: if len(document["_source"]['messages'])==2: chat_h= {'user': document["_source"]['messages'][0]['data']['content']},{'assistant': document["_source"]['messages'][1]['data']['content']} else: chat_h= {'user': document["_source"]['messages'][0]['data']['content']} tot_hist.append(chat_h) flattened_list = [] # Iterate over each tuple in the list for item1, item2 in tot_hist: # Extend the flattened list with both dictionaries flattened_list.extend([item1, item2]) chat_id_names = get_chat_id_names_by_session_id(chatname_index, id_user) yest_dict = { 'session_id': id_user, 'chat_history':flattened_list, 'content': chat_id_names } items.append(yest_dict) else: print('second condition') session_ids=list(OrderedDict.fromkeys(d["session_id"] for d in t)) yest_dict = {} items=[] for id_user in session_ids[::-1]: result = es.search( index=es_index_name, query={"match": {"session_id": id_user}}, # Use match instead of term sort="lastmodified_at:asc", ) if result and len(result["hits"]["hits"]) > 0: session_hist = messages_from_dict(json.loads(document["_source"]["history"]) for document in result["hits"]["hits"]) tot_hist=[] chat_h={} for i in session_hist: if type(i) == HumanMessage: chat_h = { 'user': i.content} elif type(i) == AIMessage: chat_h= { 'assistant': i.content} tot_hist.append(chat_h) chat_id_names = get_chat_id_names_by_session_id(chatname_index, id_user) yest_dict = { 'session_id': id_user, 'chat_history':tot_hist, 'content': chat_id_names } items.append(yest_dict) # Returning the first message content and the session ID for simplicity formatted_yest_dict = { "messages": items } else: formatted_yest_dict = {"messages": []} # print('today',formatted_yest_dict) # print('today messge',formatted_yest_dict) return JsonResponse({'messages': formatted_yest_dict}) except Exception as e: # Log the error and return a JSON error response print(f"Error in today_messages view: {e}") return JsonResponse({"error": str(e)}, status=500) def yesterday_messages(request): try: y=[] username = request.headers.get('X-Username') print('username',username) if username =='image_logs': test = get_messages_by_user_and_session(username,imtext_index) else: test = get_messages_by_user_and_session(username,es_index_name) # Get current date and time in UTC current_datetime = datetime.now(timezone.utc) # Calculate the start of today and yesterday in UTC start_of_today = current_datetime.replace(hour=0, minute=0, second=0, microsecond=0) start_of_yesterday = start_of_today - timedelta(days=1) end_of_yesterday = start_of_today - timedelta(microseconds=1) for item in test: lastmodified_at_datetime = datetime.fromtimestamp(item['lastmodified_at'] / 1000, tz=timezone.utc) if start_of_yesterday <= lastmodified_at_datetime <= end_of_yesterday: y.append(item) if y: if username =='image_logs': print('yesterday section') session_ids=list(OrderedDict.fromkeys(d["session_id"] for d in y)) yest_dict = {} items=[] for id_user in session_ids[::-1]: result = es.search( index=imtext_index, query={"match": {"session_id": id_user}}, # Use match instead of term sort="lastmodified_at:asc", ) tot_hist=[] chat_h={} imtex={} for document in result["hits"]["hits"]: if len(document["_source"])==6: chat_h= { 'system': document["_source"]['img']} if document["_source"]["messages"][0]['type']=='ai': imtex={'assistant': document["_source"]["messages"][0]['data']['content']} elif document["_source"]["messages"][1]['type']=='ai': imtex={'assistant': document["_source"]["messages"][1]['data']['content']} chat_h=chat_h,imtex elif len(document["_source"])==5: if document["_source"]['messages'][0]['type']=='ai': chat_h= {'user': document["_source"]['messages'][1]['data']['content']},{'assistant': document["_source"]['messages'][0]['data']['content']} else: if len(document["_source"]['messages'])==2: chat_h= {'user': document["_source"]['messages'][0]['data']['content']},{'assistant': document["_source"]['messages'][1]['data']['content']} else: chat_h= {'user': document["_source"]['messages'][0]['data']['content']} tot_hist.append(chat_h) flattened_list = [] # Iterate over each tuple in the list for item1, item2 in tot_hist: # Extend the flattened list with both dictionaries flattened_list.extend([item1, item2]) chat_id_names = get_chat_id_names_by_session_id(chatname_index, id_user) yest_dict = { 'session_id': id_user, 'chat_history':flattened_list, 'content': chat_id_names } items.append(yest_dict) else: print('second condition') # session_ids = list(set(d["session_id"] for d in y)) session_ids=list(OrderedDict.fromkeys(d["session_id"] for d in y)) yest_dict = {} items=[] for id_user in session_ids[::-1]: result = es.search( index=es_index_name, query={"match": {"session_id": id_user}}, # Use match instead of term sort="lastmodified_at:asc", ) if result and len(result["hits"]["hits"]) > 0: session_hist = messages_from_dict(json.loads(document["_source"]["history"]) for document in result["hits"]["hits"]) tot_hist=[] chat_h={} for i in session_hist: if type(i) == HumanMessage: chat_h = { 'user': i.content} elif type(i) == AIMessage: chat_h= { 'assistant': i.content} tot_hist.append(chat_h) chat_id_names = get_chat_id_names_by_session_id(chatname_index, id_user) yest_dict = { 'session_id': id_user, 'chat_history':tot_hist, 'content': chat_id_names } items.append(yest_dict) # Returning the first message content and the session ID for simplicity formatted_yest_dict = { "messages": items } else: formatted_yest_dict = {"messages": []} # print('yesterday',formatted_yest_dict) return JsonResponse({'messages': formatted_yest_dict}) except Exception as e: # Log the error and return a JSON error response print(f"Error in yesterday_messages view: {e}") return JsonResponse({"error": str(e)}, status=500) def previous_7_days(request): try: s=[] username = request.headers.get('X-Username') if username =='image_logs': test = get_messages_by_user_and_session(username,imtext_index) else: test = get_messages_by_user_and_session(username,es_index_name) # Get current date and time in UTC current_datetime = datetime.now(timezone.utc) # Calculate the start of today and yesterday in UTC start_of_today = current_datetime.replace(hour=0, minute=0, second=0, microsecond=0) start_of_yesterday = start_of_today - timedelta(days=1) start_of_seven_days_ago = start_of_today - timedelta(days=7) for item in test: lastmodified_at_datetime = datetime.fromtimestamp(item['lastmodified_at'] / 1000, tz=timezone.utc) if start_of_seven_days_ago <= lastmodified_at_datetime < start_of_yesterday: s.append(item) if s: if username =='image_logs': session_ids=list(OrderedDict.fromkeys(d["session_id"] for d in s)) yest_dict = {} items=[] for id_user in session_ids[::-1]: result = es.search( index=imtext_index, query={"match": {"session_id": id_user}}, # Use match instead of term sort="lastmodified_at:asc", ) tot_hist=[] chat_h={} imtex={} for document in result["hits"]["hits"]: if len(document["_source"])==6: chat_h= { 'system': document["_source"]['img']} if document["_source"]["messages"][0]['type']=='ai': imtex={'assistant': document["_source"]["messages"][0]['data']['content']} elif document["_source"]["messages"][1]['type']=='ai': imtex={'assistant': document["_source"]["messages"][1]['data']['content']} chat_h=chat_h,imtex elif len(document["_source"])==5: if document["_source"]['messages'][0]['type']=='ai': chat_h= {'user': document["_source"]['messages'][1]['data']['content']},{'assistant': document["_source"]['messages'][0]['data']['content']} else: if len(document["_source"]['messages'])==2: chat_h= {'user': document["_source"]['messages'][0]['data']['content']},{'assistant': document["_source"]['messages'][1]['data']['content']} else: chat_h= {'user': document["_source"]['messages'][0]['data']['content']} tot_hist.append(chat_h) flattened_list = [] # Iterate over each tuple in the list for item1, item2 in tot_hist: # Extend the flattened list with both dictionaries flattened_list.extend([item1, item2]) chat_id_names = get_chat_id_names_by_session_id(chatname_index, id_user) yest_dict = { 'session_id': id_user, 'chat_history':flattened_list, 'content': chat_id_names } items.append(yest_dict) else: # session_ids = list(set(d["session_id"] for d in s)) session_ids=list(OrderedDict.fromkeys(d["session_id"] for d in s)) yest_dict = {} items=[] for id_user in session_ids[::-1]: result = es.search( index=es_index_name, query={"match": {"session_id": id_user}}, # Use match instead of term sort="lastmodified_at:asc", ) if result and len(result["hits"]["hits"]) > 0: session_hist = messages_from_dict(json.loads(document["_source"]["history"]) for document in result["hits"]["hits"]) tot_hist=[] chat_h={} for i in session_hist: if type(i) == HumanMessage: chat_h = { 'user': i.content} elif type(i) == AIMessage: chat_h= { 'assistant': i.content} tot_hist.append(chat_h) chat_id_names = get_chat_id_names_by_session_id(chatname_index, id_user) yest_dict = { 'session_id': id_user, 'chat_history':tot_hist, 'content': chat_id_names } items.append(yest_dict) # Returning the first message content and the session ID for simplicity formatted_yest_dict = { "messages": items } else: formatted_yest_dict = {"messages": []} # print('previous',formatted_yest_dict) return JsonResponse({'messages': formatted_yest_dict}) except Exception as e: # Log the error and return a JSON error response print(f"Error in previous7days_messages view: {e}") return JsonResponse({"error": str(e)}, status=500)