""" File: lollms_web_ui.py Author: ParisNeo Description: Singleton class for the LoLLMS web UI. This class provides a singleton instance of the LoLLMS web UI, allowing access to its functionality and data across multiple endpoints. """ from lollms.server.elf_server import LOLLMSElfServer from flask import request from datetime import datetime from api.db import DiscussionsDB, Discussion from pathlib import Path from lollms.config import InstallOption from lollms.types import MSG_TYPE, SENDER_TYPES from lollms.extension import LOLLMSExtension, ExtensionBuilder from lollms.personality import AIPersonality, PersonalityBuilder from lollms.binding import LOLLMSConfig, BindingBuilder, LLMBinding, ModelBuilder, BindingType from lollms.paths import LollmsPaths from lollms.helpers import ASCIIColors, trace_exception from lollms.com import NotificationType, NotificationDisplayType, LoLLMsCom from lollms.app import LollmsApplication from lollms.utilities import File64BitsManager, PromptReshaper, PackageManager, find_first_available_file_index, run_async, is_asyncio_loop_running from lollms.generation import RECPTION_MANAGER, ROLE_CHANGE_DECISION, ROLE_CHANGE_OURTPUT import git import asyncio import os try: from lollms.media import WebcamImageSender, AudioRecorder Media_on=True except: ASCIIColors.warning("Couldn't load media library.\nYou will not be able to perform any of the media linked operations. please verify the logs and install any required installations") Media_on=False from safe_store import TextVectorizer, VectorizationMethod, VisualizationMethod import threading from tqdm import tqdm import traceback import sys import gc import ctypes from functools import partial import json import shutil import re import string import requests from datetime import datetime from typing import List, Tuple import time import numpy as np from lollms.utilities import find_first_available_file_index, convert_language_name if not PackageManager.check_package_installed("requests"): PackageManager.install_package("requests") if not PackageManager.check_package_installed("bs4"): PackageManager.install_package("beautifulsoup4") import requests from bs4 import BeautifulSoup def terminate_thread(thread): if thread: if not thread.is_alive(): ASCIIColors.yellow("Thread not alive") return thread_id = thread.ident exc = ctypes.py_object(SystemExit) res = ctypes.pythonapi.PyThreadState_SetAsyncExc(thread_id, exc) if res > 1: ctypes.pythonapi.PyThreadState_SetAsyncExc(thread_id, None) del thread gc.collect() raise SystemError("Failed to terminate the thread.") else: ASCIIColors.yellow("Canceled successfully")# The current version of the webui lollms_webui_version="9.0" class LOLLMSWebUI(LOLLMSElfServer): __instance = None @staticmethod def build_instance( config: LOLLMSConfig, lollms_paths: LollmsPaths, load_binding=True, load_model=True, load_voice_service=True, load_sd_service=True, try_select_binding=False, try_select_model=False, callback=None, args=None, sio = None ): if LOLLMSWebUI.__instance is None: LOLLMSWebUI( config, lollms_paths, load_binding=load_binding, load_model=load_model, load_sd_service=load_sd_service, load_voice_service=load_voice_service, try_select_binding=try_select_binding, try_select_model=try_select_model, callback=callback, args=args, sio=sio ) return LOLLMSWebUI.__instance def __init__( self, config: LOLLMSConfig, lollms_paths: LollmsPaths, load_binding=True, load_model=True, load_voice_service=True, load_sd_service=True, try_select_binding=False, try_select_model=False, callback=None, args=None, sio=None ) -> None: super().__init__( config, lollms_paths, load_binding=load_binding, load_model=load_model, load_sd_service=load_sd_service, load_voice_service=load_voice_service, try_select_binding=try_select_binding, try_select_model=try_select_model, callback=callback, sio=sio ) self.app_name:str = "LOLLMSWebUI" self.version:str = lollms_webui_version self.args = args self.busy = False self.nb_received_tokens = 0 self.config_file_path = config.file_path self.cancel_gen = False if self.config.auto_update: if self.check_update_(): ASCIIColors.info("New version found. Updating!") self.run_update_script() # Keeping track of current discussion and message self._current_user_message_id = 0 self._current_ai_message_id = 0 self._message_id = 0 self.db_path = config["db_path"] if Path(self.db_path).is_absolute(): # Create database object self.db = DiscussionsDB(self.db_path) else: # Create database object self.db = DiscussionsDB(self.lollms_paths.personal_databases_path/self.db_path) # If the database is empty, populate it with tables ASCIIColors.info("Checking discussions database... ",end="") self.db.create_tables() self.db.add_missing_columns() ASCIIColors.success("ok") # prepare vectorization if self.config.data_vectorization_activate and self.config.use_discussions_history: try: ASCIIColors.yellow("Loading long term memory") folder = self.lollms_paths.personal_databases_path/"vectorized_dbs" folder.mkdir(parents=True, exist_ok=True) self.build_long_term_skills_memory() ASCIIColors.yellow("Ready") except Exception as ex: trace_exception(ex) self.long_term_memory = None else: self.long_term_memory = None # This is used to keep track of messages self.download_infos={} self.connections = { 0:{ "current_discussion":None, "generated_text":"", "cancel_generation": False, "generation_thread": None, "processing":False, "schedule_for_deletion":False, "continuing": False, "first_chunk": True, "reception_manager": RECPTION_MANAGER() } } if Media_on: try: self.webcam = WebcamImageSender(sio,lollmsCom=self) except: self.webcam = None try: self.rec_output_folder = lollms_paths.personal_outputs_path/"audio_rec" self.rec_output_folder.mkdir(exist_ok=True, parents=True) self.summoned = False self.audio_cap = AudioRecorder(sio,self.rec_output_folder/"rt.wav", callback=self.audio_callback,lollmsCom=self) except: self.audio_cap = None self.rec_output_folder = None else: self.webcam = None self.rec_output_folder = None # Define a WebSocket event handler @sio.event async def connect(sid, environ): #Create a new connection information self.connections[sid] = { "current_discussion":self.db.load_last_discussion(), "generated_text":"", "continuing": False, "first_chunk": True, "cancel_generation": False, "generation_thread": None, "processing":False, "schedule_for_deletion":False, "reception_manager":RECPTION_MANAGER() } await self.sio.emit('connected', to=sid) ASCIIColors.success(f'Client {sid} connected') @sio.event def disconnect(sid): try: if self.connections[sid]["processing"]: self.connections[sid]["schedule_for_deletion"]=True # else: # del self.connections[sid] except Exception as ex: pass ASCIIColors.error(f'Client {sid} disconnected') # generation status self.generating=False ASCIIColors.blue(f"Your personal data is stored here :",end="") ASCIIColors.green(f"{self.lollms_paths.personal_path}") # Other methods and properties of the LoLLMSWebUI singleton class def check_module_update_(self, repo_path, branch_name="main"): try: # Open the repository ASCIIColors.yellow(f"Checking for updates from {repo_path}") repo = git.Repo(repo_path) # Fetch updates from the remote for the specified branch repo.remotes.origin.fetch(refspec=f"refs/heads/{branch_name}:refs/remotes/origin/{branch_name}") # Compare the local and remote commit IDs for the specified branch local_commit = repo.head.commit remote_commit = repo.remotes.origin.refs[branch_name].commit # Check if the local branch is behind the remote branch is_behind = repo.is_ancestor(local_commit, remote_commit) and local_commit!= remote_commit ASCIIColors.yellow(f"update availability: {is_behind}") # Return True if the local branch is behind the remote branch return is_behind except Exception as e: # Handle any errors that may occur during the fetch process # trace_exception(e) return False def check_update_(self, branch_name="main"): try: # Open the repository repo_path = str(Path(__file__).parent) if self.check_module_update_(repo_path, branch_name): return True repo_path = str(Path(__file__).parent/"lollms_core") if self.check_module_update_(repo_path, branch_name): return True repo_path = str(Path(__file__).parent/"utilities/safe_store") if self.check_module_update_(repo_path, branch_name): return True return False except Exception as e: # Handle any errors that may occur during the fetch process # trace_exception(e) return False def run_update_script(self, args=None): update_script = Path(__file__).parent/"update_script.py" # Convert Namespace object to a dictionary if args: args_dict = vars(args) else: args_dict = {} # Filter out any key-value pairs where the value is None valid_args = {key: value for key, value in args_dict.items() if value is not None} # Save the arguments to a temporary file temp_file = Path(__file__).parent/"temp_args.txt" with open(temp_file, "w") as file: # Convert the valid_args dictionary to a string in the format "key1 value1 key2 value2 ..." arg_string = " ".join([f"--{key} {value}" for key, value in valid_args.items()]) file.write(arg_string) os.system(f"python {update_script}") sys.exit(0) def run_restart_script(self, args): restart_script = Path(__file__).parent/"restart_script.py" # Convert Namespace object to a dictionary args_dict = vars(args) # Filter out any key-value pairs where the value is None valid_args = {key: value for key, value in args_dict.items() if value is not None} # Save the arguments to a temporary file temp_file = Path(__file__).parent/"temp_args.txt" with open(temp_file, "w") as file: # Convert the valid_args dictionary to a string in the format "key1 value1 key2 value2 ..." arg_string = " ".join([f"--{key} {value}" for key, value in valid_args.items()]) file.write(arg_string) os.system(f"python {restart_script}") sys.exit(0) def audio_callback(self, text): if self.summoned: client_id = 0 self.cancel_gen = False self.connections[client_id]["generated_text"]="" self.connections[client_id]["cancel_generation"]=False self.connections[client_id]["continuing"]=False self.connections[client_id]["first_chunk"]=True if not self.model: ASCIIColors.error("Model not selected. Please select a model") self.error("Model not selected. Please select a model", client_id=client_id) return if not self.busy: if self.connections[client_id]["current_discussion"] is None: if self.db.does_last_discussion_have_messages(): self.connections[client_id]["current_discussion"] = self.db.create_discussion() else: self.connections[client_id]["current_discussion"] = self.db.load_last_discussion() prompt = text ump = self.config.discussion_prompt_separator +self.config.user_name.strip() if self.config.use_user_name_in_discussions else self.personality.user_message_prefix message = self.connections[client_id]["current_discussion"].add_message( message_type = MSG_TYPE.MSG_TYPE_FULL.value, sender_type = SENDER_TYPES.SENDER_TYPES_USER.value, sender = ump.replace(self.config.discussion_prompt_separator,"").replace(":",""), content=prompt, metadata=None, parent_message_id=self.message_id ) ASCIIColors.green("Starting message generation by "+self.personality.name) self.connections[client_id]['generation_thread'] = threading.Thread(target=self.start_message_generation, args=(message, message.id, client_id)) self.connections[client_id]['generation_thread'].start() self.sio.sleep(0.01) ASCIIColors.info("Started generation task") self.busy=True #tpe = threading.Thread(target=self.start_message_generation, args=(message, message_id, client_id)) #tpe.start() else: self.error("I am busy. Come back later.", client_id=client_id) else: if output["text"].lower()=="lollms": self.summoned = True def scrape_and_save(self, url, file_path): # Send a GET request to the URL response = requests.get(url) # Parse the HTML content using BeautifulSoup soup = BeautifulSoup(response.content, 'html.parser') # Find all the text content in the webpage text_content = soup.get_text() # Remove extra returns and spaces text_content = ' '.join(text_content.split()) # Save the text content as a text file with open(file_path, 'w', encoding="utf-8") as file: file.write(text_content) self.info(f"Webpage content saved to {file_path}") def rebuild_personalities(self, reload_all=False): if reload_all: self.mounted_personalities=[] loaded = self.mounted_personalities loaded_names = [f"{p.category}/{p.personality_folder_name}:{p.selected_language}" if p.selected_language else f"{p.category}/{p.personality_folder_name}" for p in loaded] mounted_personalities=[] ASCIIColors.success(f" ╔══════════════════════════════════════════════════╗ ") ASCIIColors.success(f" ║ Building mounted Personalities ║ ") ASCIIColors.success(f" ╚══════════════════════════════════════════════════╝ ") to_remove=[] for i,personality in enumerate(self.config['personalities']): if i==self.config["active_personality_id"]: ASCIIColors.red("*", end="") ASCIIColors.green(f" {personality}") else: ASCIIColors.yellow(f" {personality}") if personality in loaded_names: mounted_personalities.append(loaded[loaded_names.index(personality)]) else: personality_path = f"{personality}" if not ":" in personality else f"{personality.split(':')[0]}" try: personality = AIPersonality(personality_path, self.lollms_paths, self.config, model=self.model, app=self, selected_language=personality.split(":")[1] if ":" in personality else None, run_scripts=True) mounted_personalities.append(personality) if self.config.enable_voice_service and self.config.auto_read and len(personality.audio_samples)>0: try: from lollms.services.xtts.lollms_xtts import LollmsXTTS if self.tts is None: self.tts = LollmsXTTS(self, voice_samples_path=Path(__file__).parent.parent/"voices", xtts_base_url= self.config.xtts_base_url) except: self.warning(f"Personality {personality.name} request using custom voice but couldn't load XTTS") except Exception as ex: ASCIIColors.error(f"Personality file not found or is corrupted ({personality_path}).\nReturned the following exception:{ex}\nPlease verify that the personality you have selected exists or select another personality. Some updates may lead to change in personality name or category, so check the personality selection in settings to be sure.") ASCIIColors.info("Trying to force reinstall") if self.config["debug"]: print(ex) try: personality = AIPersonality( personality_path, self.lollms_paths, self.config, self.model, app = self, run_scripts=True, selected_language=personality.split(":")[1] if ":" in personality else None, installation_option=InstallOption.FORCE_INSTALL) mounted_personalities.append(personality) if personality.processor: personality.processor.mounted() except Exception as ex: ASCIIColors.error(f"Couldn't load personality at {personality_path}") trace_exception(ex) ASCIIColors.info(f"Unmounting personality") to_remove.append(i) personality = AIPersonality(None, self.lollms_paths, self.config, self.model, app=self, run_scripts=True, installation_option=InstallOption.FORCE_INSTALL) mounted_personalities.append(personality) if personality.processor: personality.processor.mounted() ASCIIColors.info("Reverted to default personality") if self.config["active_personality_id"]>=0 and self.config["active_personality_id"]=len(self.config["personalities"]): self.config["active_personality_id"]=0 return mounted_personalities def rebuild_extensions(self, reload_all=False): if reload_all: self.mounted_extensions=[] loaded = self.mounted_extensions loaded_names = [f"{p.category}/{p.extension_folder_name}" for p in loaded] mounted_extensions=[] ASCIIColors.success(f" ╔══════════════════════════════════════════════════╗ ") ASCIIColors.success(f" ║ Building mounted Extensions ║ ") ASCIIColors.success(f" ╚══════════════════════════════════════════════════╝ ") to_remove=[] for i,extension in enumerate(self.config['extensions']): ASCIIColors.yellow(f" {extension}") if extension in loaded_names: mounted_extensions.append(loaded[loaded_names.index(extension)]) else: extension_path = self.lollms_paths.extensions_zoo_path/f"{extension}" try: extension = ExtensionBuilder().build_extension(extension_path,self.lollms_paths, self) mounted_extensions.append(extension) except Exception as ex: ASCIIColors.error(f"Extension file not found or is corrupted ({extension_path}).\nReturned the following exception:{ex}\nPlease verify that the personality you have selected exists or select another personality. Some updates may lead to change in personality name or category, so check the personality selection in settings to be sure.") trace_exception(ex) ASCIIColors.info("Trying to force reinstall") if self.config["debug"]: print(ex) ASCIIColors.success(f" ╔══════════════════════════════════════════════════╗ ") ASCIIColors.success(f" ║ Done ║ ") ASCIIColors.success(f" ╚══════════════════════════════════════════════════╝ ") # Sort the indices in descending order to ensure correct removal to_remove.sort(reverse=True) # Remove elements from the list based on the indices for index in to_remove: if 0 <= index < len(mounted_extensions): mounted_extensions.pop(index) self.config["extensions"].pop(index) ASCIIColors.info(f"removed personality {extension_path}") return mounted_extensions # ================================== LOLLMSApp #properties @property def message_id(self): return self._message_id @message_id.setter def message_id(self, id): self._message_id=id @property def current_user_message_id(self): return self._current_user_message_id @current_user_message_id.setter def current_user_message_id(self, id): self._current_user_message_id=id self._message_id = id @property def current_ai_message_id(self): return self._current_ai_message_id @current_ai_message_id.setter def current_ai_message_id(self, id): self._current_ai_message_id=id self._message_id = id def download_file(self, url, installation_path, callback=None): """ Downloads a file from a URL, reports the download progress using a callback function, and displays a progress bar. Args: url (str): The URL of the file to download. installation_path (str): The path where the file should be saved. callback (function, optional): A callback function to be called during the download with the progress percentage as an argument. Defaults to None. """ try: response = requests.get(url, stream=True) # Get the file size from the response headers total_size = int(response.headers.get('content-length', 0)) with open(installation_path, 'wb') as file: downloaded_size = 0 with tqdm(total=total_size, unit='B', unit_scale=True, ncols=80) as progress_bar: for chunk in response.iter_content(chunk_size=8192): if chunk: file.write(chunk) downloaded_size += len(chunk) if callback is not None: callback(downloaded_size, total_size) progress_bar.update(len(chunk)) if callback is not None: callback(total_size, total_size) print("File downloaded successfully") except Exception as e: print("Couldn't download file:", str(e)) def clean_string(self, input_string): # Remove extra spaces by replacing multiple spaces with a single space #cleaned_string = re.sub(r'\s+', ' ', input_string) # Remove extra line breaks by replacing multiple consecutive line breaks with a single line break cleaned_string = re.sub(r'\n\s*\n', '\n', input_string) # Create a string containing all punctuation characters punctuation_chars = string.punctuation # Define a regular expression pattern to match and remove non-alphanumeric characters #pattern = f'[^a-zA-Z0-9\s{re.escape(punctuation_chars)}]' # This pattern matches any character that is not a letter, digit, space, or punctuation pattern = f'[^a-zA-Z0-9\u00C0-\u017F\s{re.escape(punctuation_chars)}]' # Use re.sub to replace the matched characters with an empty string cleaned_string = re.sub(pattern, '', cleaned_string) return cleaned_string def make_discussion_title(self, discussion, client_id=None): """ Builds a title for a discussion """ # Get the list of messages messages = discussion.get_messages() discussion_messages = "!@>instruction: Create a short title to this discussion\n" discussion_title = "\n!@>Discussion title:" available_space = self.config.ctx_size - 150 - len(self.model.tokenize(discussion_messages))- len(self.model.tokenize(discussion_title)) # Initialize a list to store the full messages full_message_list = [] # Accumulate messages until the cumulative number of tokens exceeds available_space tokens_accumulated = 0 # Accumulate messages starting from message_index for message in messages: # Check if the message content is not empty and visible to the AI if message.content != '' and ( message.message_type <= MSG_TYPE.MSG_TYPE_FULL_INVISIBLE_TO_USER.value and message.message_type != MSG_TYPE.MSG_TYPE_FULL_INVISIBLE_TO_AI.value): # Tokenize the message content message_tokenized = self.model.tokenize( "\n" + self.config.discussion_prompt_separator + message.sender + ": " + message.content.strip()) # Check if adding the message will exceed the available space if tokens_accumulated + len(message_tokenized) > available_space: break # Add the tokenized message to the full_message_list full_message_list.insert(0, message_tokenized) # Update the cumulative number of tokens tokens_accumulated += len(message_tokenized) # Build the final discussion messages by detokenizing the full_message_list for message_tokens in full_message_list: discussion_messages += self.model.detokenize(message_tokens) discussion_messages += discussion_title title = [""] def receive( chunk:str, message_type:MSG_TYPE ): if chunk: title[0] += chunk antiprompt = self.personality.detect_antiprompt(title[0]) if antiprompt: ASCIIColors.warning(f"\nDetected hallucination with antiprompt: {antiprompt}") title[0] = self.remove_text_from_string(title[0],antiprompt) return False else: return True self._generate(discussion_messages, 150, client_id, receive) ASCIIColors.info(title[0]) return title[0] def prepare_reception(self, client_id): if not self.connections[client_id]["continuing"]: self.connections[client_id]["generated_text"] = "" self.connections[client_id]["first_chunk"]=True self.nb_received_tokens = 0 self.start_time = datetime.now() def recover_discussion(self,client_id, message_index=-1): messages = self.connections[client_id]["current_discussion"].get_messages() discussion="" for msg in messages: if message_index!=-1 and msg>message_index: break discussion += "\n" + self.config.discussion_prompt_separator + msg.sender + ": " + msg.content.strip() return discussion def prepare_query(self, client_id: str, message_id: int = -1, is_continue: bool = False, n_tokens: int = 0, generation_type = None) -> Tuple[str, str, List[str]]: """ Prepares the query for the model. Args: client_id (str): The client ID. message_id (int): The message ID. Default is -1. is_continue (bool): Whether the query is a continuation. Default is False. n_tokens (int): The number of tokens. Default is 0. Returns: Tuple[str, str, List[str]]: The prepared query, original message content, and tokenized query. """ # Get the list of messages messages = self.connections[client_id]["current_discussion"].get_messages() # Find the index of the message with the specified message_id message_index = -1 for i, message in enumerate(messages): if message.id == message_id: message_index = i break # Define current message current_message = messages[message_index] # Build the conditionning text block conditionning = self.personality.personality_conditioning # Check if there are document files to add to the prompt documentation = "" knowledge = "" # boosting information if self.config.positive_boost: positive_boost="\n!@>important information: "+self.config.positive_boost+"\n" n_positive_boost = len(self.model.tokenize(positive_boost)) else: positive_boost="" n_positive_boost = 0 if self.config.negative_boost: negative_boost="\n!@>important information: "+self.config.negative_boost+"\n" n_negative_boost = len(self.model.tokenize(negative_boost)) else: negative_boost="" n_negative_boost = 0 if self.config.force_output_language_to_be: force_language="\n!@>important information: Answer the user in this language :"+self.config.force_output_language_to_be+"\n" n_force_language = len(self.model.tokenize(force_language)) else: force_language="" n_force_language = 0 if self.config.fun_mode: fun_mode="\n!@>important information: Fun mode activated. In this mode you must answer in a funny playful way. Do not be serious in your answers. Each answer needs to make the user laugh.\n" n_fun_mode = len(self.model.tokenize(positive_boost)) else: fun_mode="" n_fun_mode = 0 if generation_type != "simple_question": if self.personality.persona_data_vectorizer: if documentation=="": documentation="\n!@>important information: Use the documentation data to answer the user questions. If the data is not present in the documentation, please tell the user that the information he is asking for does not exist in the documentation section. It is strictly forbidden to give the user an answer without having actual proof from the documentation.\n!@>Documentation:\n" if self.config.data_vectorization_build_keys_words: discussion = self.recover_discussion(client_id)[-512:] query = self.personality.fast_gen(f"\n!@>instruction: Read the discussion and rewrite the last prompt for someone who didn't read the entire discussion.\nDo not answer the prompt. Do not add explanations.\n!@>discussion:\n{discussion}\n!@>enhanced query: ", max_generation_size=256, show_progress=True) ASCIIColors.cyan(f"Query:{query}") else: query = current_message.content try: docs, sorted_similarities = self.personality.persona_data_vectorizer.recover_text(query, top_k=self.config.data_vectorization_nb_chunks) for doc, infos in zip(docs, sorted_similarities): documentation += f"document chunk:\n{doc}" except: self.warning("Couldn't add documentation to the context. Please verify the vector database") if len(self.personality.text_files) > 0 and self.personality.vectorizer: if documentation=="": documentation="\n!@>important information: Use the documentation data to answer the user questions. If the data is not present in the documentation, please tell the user that the information he is asking for does not exist in the documentation section. It is strictly forbidden to give the user an answer without having actual proof from the documentation.\n!@>Documentation:\n" if self.config.data_vectorization_build_keys_words: discussion = self.recover_discussion(client_id)[-512:] query = self.personality.fast_gen(f"\n!@>instruction: Read the discussion and rewrite the last prompt for someone who didn't read the entire discussion.\nDo not answer the prompt. Do not add explanations.\n!@>discussion:\n{discussion}\n!@>enhanced query: ", max_generation_size=256, show_progress=True) ASCIIColors.cyan(f"Query:{query}") else: query = current_message.content try: docs, sorted_similarities = self.personality.vectorizer.recover_text(query, top_k=self.config.data_vectorization_nb_chunks) for doc, infos in zip(docs, sorted_similarities): documentation += f"document chunk:\nchunk path: {infos[0]}\nchunk content:{doc}" documentation += "\n!@>important information: Use the documentation data to answer the user questions. If the data is not present in the documentation, please tell the user that the information he is asking for does not exist in the documentation section. It is strictly forbidden to give the user an answer without having actual proof from the documentation." except: self.warning("Couldn't add documentation to the context. Please verify the vector database") # Check if there is discussion knowledge to add to the prompt if self.config.use_discussions_history and self.long_term_memory is not None: if knowledge=="": knowledge="!@>knowledge:\n" try: docs, sorted_similarities = self.long_term_memory.recover_text(current_message.content, top_k=self.config.data_vectorization_nb_chunks) for i,(doc, infos) in enumerate(zip(docs, sorted_similarities)): knowledge += f"!@>knowledge {i}:\n!@>title:\n{infos[0]}\ncontent:\n{doc}" except: self.warning("Couldn't add long term memory information to the context. Please verify the vector database") # Add information about the user user_description="" if self.config.use_user_name_in_discussions: user_description="!@>User description:\n"+self.config.user_description+"\n" # Tokenize the conditionning text and calculate its number of tokens tokens_conditionning = self.model.tokenize(conditionning) n_cond_tk = len(tokens_conditionning) # Tokenize the documentation text and calculate its number of tokens if len(documentation)>0: tokens_documentation = self.model.tokenize(documentation) n_doc_tk = len(tokens_documentation) else: tokens_documentation = [] n_doc_tk = 0 # Tokenize the knowledge text and calculate its number of tokens if len(knowledge)>0: tokens_history = self.model.tokenize(knowledge) n_history_tk = len(tokens_history) else: tokens_history = [] n_history_tk = 0 # Tokenize user description if len(user_description)>0: tokens_user_description = self.model.tokenize(user_description) n_user_description_tk = len(tokens_user_description) else: tokens_user_description = [] n_user_description_tk = 0 # Calculate the total number of tokens between conditionning, documentation, and knowledge total_tokens = n_cond_tk + n_doc_tk + n_history_tk + n_user_description_tk + n_positive_boost + n_negative_boost + n_force_language + n_fun_mode # Calculate the available space for the messages available_space = self.config.ctx_size - n_tokens - total_tokens if self.config.debug: self.info(f"Tokens summary:\nConditionning:{n_cond_tk}\ndoc:{n_doc_tk}\nhistory:{n_history_tk}\nuser description:{n_user_description_tk}\nAvailable space:{available_space}",10) # Raise an error if the available space is 0 or less if available_space<1: self.error("Not enough space in context!!") raise Exception("Not enough space in context!!") # Accumulate messages until the cumulative number of tokens exceeds available_space tokens_accumulated = 0 # Initialize a list to store the full messages full_message_list = [] # If this is not a continue request, we add the AI prompt if not is_continue: message_tokenized = self.model.tokenize( "\n" +self.personality.ai_message_prefix.strip() ) full_message_list.append(message_tokenized) # Update the cumulative number of tokens tokens_accumulated += len(message_tokenized) if generation_type != "simple_question": # Accumulate messages starting from message_index for i in range(message_index, -1, -1): message = messages[i] # Check if the message content is not empty and visible to the AI if message.content != '' and ( message.message_type <= MSG_TYPE.MSG_TYPE_FULL_INVISIBLE_TO_USER.value and message.message_type != MSG_TYPE.MSG_TYPE_FULL_INVISIBLE_TO_AI.value): # Tokenize the message content message_tokenized = self.model.tokenize( "\n" + self.config.discussion_prompt_separator + message.sender + ": " + message.content.strip()) # Check if adding the message will exceed the available space if tokens_accumulated + len(message_tokenized) > available_space: break # Add the tokenized message to the full_message_list full_message_list.insert(0, message_tokenized) # Update the cumulative number of tokens tokens_accumulated += len(message_tokenized) else: message = messages[message_index] # Check if the message content is not empty and visible to the AI if message.content != '' and ( message.message_type <= MSG_TYPE.MSG_TYPE_FULL_INVISIBLE_TO_USER.value and message.message_type != MSG_TYPE.MSG_TYPE_FULL_INVISIBLE_TO_AI.value): # Tokenize the message content message_tokenized = self.model.tokenize( "\n" + self.config.discussion_prompt_separator + message.sender + ": " + message.content.strip()) # Add the tokenized message to the full_message_list full_message_list.insert(0, message_tokenized) # Update the cumulative number of tokens tokens_accumulated += len(message_tokenized) # Build the final discussion messages by detokenizing the full_message_list discussion_messages = "" for i in range(len(full_message_list)-1): message_tokens = full_message_list[i] discussion_messages += self.model.detokenize(message_tokens) if len(full_message_list)>0: ai_prefix = self.model.detokenize(full_message_list[-1]) else: ai_prefix = "" # Build the final prompt by concatenating the conditionning and discussion messages prompt_data = conditionning + documentation + knowledge + user_description + discussion_messages + positive_boost + negative_boost + force_language + fun_mode + ai_prefix # Tokenize the prompt data tokens = self.model.tokenize(prompt_data) # if this is a debug then show prompt construction details if self.config["debug"]: ASCIIColors.bold("CONDITIONNING") ASCIIColors.yellow(conditionning) ASCIIColors.bold("DOC") ASCIIColors.yellow(documentation) ASCIIColors.bold("HISTORY") ASCIIColors.yellow(knowledge) ASCIIColors.bold("DISCUSSION") ASCIIColors.hilight(discussion_messages,"!@>",ASCIIColors.color_yellow,ASCIIColors.color_bright_red,False) ASCIIColors.bold("Final prompt") ASCIIColors.hilight(prompt_data,"!@>",ASCIIColors.color_yellow,ASCIIColors.color_bright_red,False) ASCIIColors.info(f"prompt size:{len(tokens)} tokens") ASCIIColors.info(f"available space after doc and knowledge:{available_space} tokens") self.info(f"Tokens summary:\nPrompt size:{len(tokens)}\nTo generate:{available_space}",10) # Details context_details = { "conditionning":conditionning, "documentation":documentation, "knowledge":knowledge, "user_description":user_description, "discussion_messages":discussion_messages, "positive_boost":positive_boost, "negative_boost":negative_boost, "force_language":force_language, "fun_mode":fun_mode, "ai_prefix":ai_prefix } # Return the prepared query, original message content, and tokenized query return prompt_data, current_message.content, tokens, context_details def get_discussion_to(self, client_id, message_id=-1): messages = self.connections[client_id]["current_discussion"].get_messages() full_message_list = [] ump = self.config.discussion_prompt_separator +self.config.user_name.strip() if self.config.use_user_name_in_discussions else self.personality.user_message_prefix for message in messages: if message["id"]<= message_id or message_id==-1: if message["type"]!=MSG_TYPE.MSG_TYPE_FULL_INVISIBLE_TO_USER: if message["sender"]==self.personality.name: full_message_list.append(self.personality.ai_message_prefix+message["content"]) else: full_message_list.append(ump + message["content"]) link_text = "\n"# self.personality.link_text if len(full_message_list) > self.config["nb_messages_to_remember"]: discussion_messages = self.personality.personality_conditioning+ link_text.join(full_message_list[-self.config["nb_messages_to_remember"]:]) else: discussion_messages = self.personality.personality_conditioning+ link_text.join(full_message_list) return discussion_messages # Removes the last return def notify( self, content, notification_type:NotificationType=NotificationType.NOTIF_SUCCESS, duration:int=4, client_id=None, display_type:NotificationDisplayType=NotificationDisplayType.TOAST, verbose:bool|None=None, ): if verbose is None: verbose = self.verbose run_async(partial(self.sio.emit,'notification', { 'content': content,# self.connections[client_id]["generated_text"], 'notification_type': notification_type.value, "duration": duration, 'display_type':display_type.value }, to=client_id ) ) if verbose: if notification_type==NotificationType.NOTIF_SUCCESS: ASCIIColors.success(content) elif notification_type==NotificationType.NOTIF_INFO: ASCIIColors.info(content) elif notification_type==NotificationType.NOTIF_WARNING: ASCIIColors.warning(content) else: ASCIIColors.red(content) def new_message(self, client_id, sender=None, content="", parameters=None, metadata=None, ui=None, message_type:MSG_TYPE=MSG_TYPE.MSG_TYPE_FULL, sender_type:SENDER_TYPES=SENDER_TYPES.SENDER_TYPES_AI, open=False ): mtdt = metadata if metadata is None or type(metadata) == str else json.dumps(metadata, indent=4) if sender==None: sender= self.personality.name msg = self.connections[client_id]["current_discussion"].add_message( message_type = message_type.value, sender_type = sender_type.value, sender = sender, content = content, metadata = mtdt, ui = ui, rank = 0, parent_message_id = self.connections[client_id]["current_discussion"].current_message.id, binding = self.config["binding_name"], model = self.config["model_name"], personality = self.config["personalities"][self.config["active_personality_id"]], ) # first the content is empty, but we'll fill it at the end run_async(partial( self.sio.emit,'new_message', { "sender": sender, "message_type": message_type.value, "sender_type": SENDER_TYPES.SENDER_TYPES_AI.value, "content": content, "parameters": parameters, "metadata": metadata, "ui": ui, "id": msg.id, "parent_message_id": msg.parent_message_id, 'binding': self.config["binding_name"], 'model' : self.config["model_name"], 'personality': self.config["personalities"][self.config["active_personality_id"]], 'created_at': self.connections[client_id]["current_discussion"].current_message.created_at, 'finished_generating_at': self.connections[client_id]["current_discussion"].current_message.finished_generating_at, 'open': open }, to=client_id ) ) def update_message(self, client_id, chunk, parameters=None, metadata=[], ui=None, msg_type:MSG_TYPE=None ): self.connections[client_id]["current_discussion"].current_message.finished_generating_at=datetime.now().strftime('%Y-%m-%d %H:%M:%S') mtdt = json.dumps(metadata, indent=4) if metadata is not None and type(metadata)== list else metadata if self.nb_received_tokens==1: run_async( partial(self.sio.emit,'update_message', { "sender": self.personality.name, 'id':self.connections[client_id]["current_discussion"].current_message.id, 'content': "✍ warming up ...",# self.connections[client_id]["generated_text"], 'ui': ui, 'discussion_id':self.connections[client_id]["current_discussion"].discussion_id, 'message_type': MSG_TYPE.MSG_TYPE_STEP_END.value, 'finished_generating_at': self.connections[client_id]["current_discussion"].current_message.finished_generating_at, 'parameters':parameters, 'metadata':metadata }, to=client_id ) ) run_async( partial(self.sio.emit,'update_message', { "sender": self.personality.name, 'id':self.connections[client_id]["current_discussion"].current_message.id, 'content': chunk,# self.connections[client_id]["generated_text"], 'ui': ui, 'discussion_id':self.connections[client_id]["current_discussion"].discussion_id, 'message_type': msg_type.value if msg_type is not None else MSG_TYPE.MSG_TYPE_CHUNK.value if self.nb_received_tokens>1 else MSG_TYPE.MSG_TYPE_FULL.value, 'finished_generating_at': self.connections[client_id]["current_discussion"].current_message.finished_generating_at, 'parameters':parameters, 'metadata':metadata }, to=client_id ) ) if msg_type != MSG_TYPE.MSG_TYPE_INFO: self.connections[client_id]["current_discussion"].update_message(self.connections[client_id]["generated_text"], new_metadata=mtdt, new_ui=ui) def close_message(self, client_id): if not self.connections[client_id]["current_discussion"]: return #fix halucination self.connections[client_id]["generated_text"]=self.connections[client_id]["generated_text"].split("!@>")[0] # Send final message self.connections[client_id]["current_discussion"].current_message.finished_generating_at=datetime.now().strftime('%Y-%m-%d %H:%M:%S') run_async( partial(self.sio.emit,'close_message', { "sender": self.personality.name, "id": self.connections[client_id]["current_discussion"].current_message.id, "content":self.connections[client_id]["generated_text"], 'binding': self.config["binding_name"], 'model' : self.config["model_name"], 'personality':self.config["personalities"][self.config["active_personality_id"]], 'created_at': self.connections[client_id]["current_discussion"].current_message.created_at, 'finished_generating_at': self.connections[client_id]["current_discussion"].current_message.finished_generating_at, }, to=client_id ) ) def process_chunk( self, chunk:str, message_type:MSG_TYPE, parameters:dict=None, metadata:list=None, client_id:int=0, personality:AIPersonality=None ): """ Processes a chunk of generated text """ if chunk is None: return True if not client_id in list(self.connections.keys()): self.error("Connection lost", client_id=client_id) return if message_type == MSG_TYPE.MSG_TYPE_STEP: ASCIIColors.info("--> Step:"+chunk) if message_type == MSG_TYPE.MSG_TYPE_STEP_START: ASCIIColors.info("--> Step started:"+chunk) if message_type == MSG_TYPE.MSG_TYPE_STEP_END: if parameters['status']: ASCIIColors.success("--> Step ended:"+chunk) else: ASCIIColors.error("--> Step ended:"+chunk) if message_type == MSG_TYPE.MSG_TYPE_EXCEPTION: self.error(chunk, client_id=client_id) ASCIIColors.error("--> Exception from personality:"+chunk) if message_type == MSG_TYPE.MSG_TYPE_WARNING: self.warning(chunk,client_id=client_id) ASCIIColors.error("--> Exception from personality:"+chunk) if message_type == MSG_TYPE.MSG_TYPE_INFO: self.info(chunk, client_id=client_id) ASCIIColors.info("--> Info:"+chunk) if message_type == MSG_TYPE.MSG_TYPE_UI: self.update_message(client_id, "", parameters, metadata, chunk, MSG_TYPE.MSG_TYPE_UI) if message_type == MSG_TYPE.MSG_TYPE_NEW_MESSAGE: self.nb_received_tokens = 0 self.start_time = datetime.now() self.new_message( client_id, self.personality.name if personality is None else personality.name, chunk if parameters["type"]!=MSG_TYPE.MSG_TYPE_UI.value else '', metadata = [{ "title":chunk, "content":parameters["metadata"] } ] if parameters["type"]==MSG_TYPE.MSG_TYPE_JSON_INFOS.value else None, ui= chunk if parameters["type"]==MSG_TYPE.MSG_TYPE_UI.value else None, message_type= MSG_TYPE(parameters["type"]) ) elif message_type == MSG_TYPE.MSG_TYPE_FINISHED_MESSAGE: self.close_message(client_id) elif message_type == MSG_TYPE.MSG_TYPE_CHUNK: if self.nb_received_tokens==0: self.start_time = datetime.now() dt =(datetime.now() - self.start_time).seconds if dt==0: dt=1 spd = self.nb_received_tokens/dt ASCIIColors.green(f"Received {self.nb_received_tokens} tokens (speed: {spd:.2f}t/s) ",end="\r",flush=True) sys.stdout = sys.__stdout__ sys.stdout.flush() if chunk: self.connections[client_id]["generated_text"] += chunk antiprompt = self.personality.detect_antiprompt(self.connections[client_id]["generated_text"]) if antiprompt: ASCIIColors.warning(f"\nDetected hallucination with antiprompt: {antiprompt}") self.connections[client_id]["generated_text"] = self.remove_text_from_string(self.connections[client_id]["generated_text"],antiprompt) self.update_message(client_id, self.connections[client_id]["generated_text"], parameters, metadata, None, MSG_TYPE.MSG_TYPE_FULL) return False else: self.nb_received_tokens += 1 if self.connections[client_id]["continuing"] and self.connections[client_id]["first_chunk"]: self.update_message(client_id, self.connections[client_id]["generated_text"], parameters, metadata) else: self.update_message(client_id, chunk, parameters, metadata, msg_type=MSG_TYPE.MSG_TYPE_CHUNK) self.connections[client_id]["first_chunk"]=False # if stop generation is detected then stop if not self.cancel_gen: return True else: self.cancel_gen = False ASCIIColors.warning("Generation canceled") return False # Stream the generated text to the main process elif message_type == MSG_TYPE.MSG_TYPE_FULL: self.connections[client_id]["generated_text"] = chunk self.nb_received_tokens += 1 dt =(datetime.now() - self.start_time).seconds if dt==0: dt=1 spd = self.nb_received_tokens/dt ASCIIColors.green(f"Received {self.nb_received_tokens} tokens (speed: {spd:.2f}t/s) ",end="\r",flush=True) antiprompt = self.personality.detect_antiprompt(self.connections[client_id]["generated_text"]) if antiprompt: ASCIIColors.warning(f"\nDetected hallucination with antiprompt: {antiprompt}") self.connections[client_id]["generated_text"] = self.remove_text_from_string(self.connections[client_id]["generated_text"],antiprompt) self.update_message(client_id, self.connections[client_id]["generated_text"], parameters, metadata, None, MSG_TYPE.MSG_TYPE_FULL) return False self.update_message(client_id, chunk, parameters, metadata, ui=None, msg_type=message_type) return True # Stream the generated text to the frontend else: self.update_message(client_id, chunk, parameters, metadata, ui=None, msg_type=message_type) return True def generate(self, full_prompt, prompt, context_details, n_predict, client_id, callback=None): if self.personality.processor is not None: ASCIIColors.info("Running workflow") try: self.personality.callback = callback self.personality.processor.run_workflow( prompt, full_prompt, callback, context_details) except Exception as ex: trace_exception(ex) # Catch the exception and get the traceback as a list of strings traceback_lines = traceback.format_exception(type(ex), ex, ex.__traceback__) # Join the traceback lines into a single string traceback_text = ''.join(traceback_lines) ASCIIColors.error(f"Workflow run failed.\nError:{ex}") ASCIIColors.error(traceback_text) if callback: callback(f"Workflow run failed\nError:{ex}", MSG_TYPE.MSG_TYPE_EXCEPTION) print("Finished executing the workflow") return self._generate(full_prompt, n_predict, client_id, callback) ASCIIColors.success("\nFinished executing the generation") def _generate(self, prompt, n_predict, client_id, callback=None): self.nb_received_tokens = 0 self.start_time = datetime.now() if self.model is not None: if self.model.binding_type==BindingType.TEXT_IMAGE and len(self.personality.image_files)>0: ASCIIColors.info(f"warmup for generating up to {n_predict} tokens") if self.config["override_personality_model_parameters"]: output = self.model.generate_with_images( prompt, self.personality.image_files, callback=callback, n_predict=n_predict, temperature=self.config['temperature'], top_k=self.config['top_k'], top_p=self.config['top_p'], repeat_penalty=self.config['repeat_penalty'], repeat_last_n = self.config['repeat_last_n'], seed=self.config['seed'], n_threads=self.config['n_threads'] ) else: output = self.model.generate_with_images( prompt, self.personality.image_files, callback=callback, n_predict=min(n_predict,self.personality.model_n_predicts), temperature=self.personality.model_temperature, top_k=self.personality.model_top_k, top_p=self.personality.model_top_p, repeat_penalty=self.personality.model_repeat_penalty, repeat_last_n = self.personality.model_repeat_last_n, seed=self.config['seed'], n_threads=self.config['n_threads'] ) else: ASCIIColors.info(f"warmup for generating up to {n_predict} tokens") if self.config["override_personality_model_parameters"]: output = self.model.generate( prompt, callback=callback, n_predict=n_predict, temperature=self.config['temperature'], top_k=self.config['top_k'], top_p=self.config['top_p'], repeat_penalty=self.config['repeat_penalty'], repeat_last_n = self.config['repeat_last_n'], seed=self.config['seed'], n_threads=self.config['n_threads'] ) else: output = self.model.generate( prompt, callback=callback, n_predict=min(n_predict,self.personality.model_n_predicts), temperature=self.personality.model_temperature, top_k=self.personality.model_top_k, top_p=self.personality.model_top_p, repeat_penalty=self.personality.model_repeat_penalty, repeat_last_n = self.personality.model_repeat_last_n, seed=self.config['seed'], n_threads=self.config['n_threads'] ) else: print("No model is installed or selected. Please make sure to install a model and select it inside your configuration before attempting to communicate with the model.") print("To do this: Install the model to your models/ folder.") print("Then set your model information in your local configuration file that you can find in configs/local_config.yaml") print("You can also use the ui to set your model in the settings page.") output = "" return output def start_message_generation(self, message, message_id, client_id, is_continue=False, generation_type=None): if self.personality is None: self.warning("Select a personality") return ASCIIColors.info(f"Text generation requested by client: {client_id}") # send the message to the bot print(f"Received message : {message.content}") if self.connections[client_id]["current_discussion"]: if not self.model: self.error("No model selected. Please make sure you select a model before starting generation", client_id=client_id) return # First we need to send the new message ID to the client if is_continue: self.connections[client_id]["current_discussion"].load_message(message_id) self.connections[client_id]["generated_text"] = message.content else: self.new_message(client_id, self.personality.name, "") self.update_message(client_id, "✍ warming up ...", msg_type=MSG_TYPE.MSG_TYPE_STEP_START) # prepare query and reception self.discussion_messages, self.current_message, tokens, context_details = self.prepare_query(client_id, message_id, is_continue, n_tokens=self.config.min_n_predict, generation_type=generation_type) self.prepare_reception(client_id) self.generating = True self.connections[client_id]["processing"]=True try: self.generate( self.discussion_messages, self.current_message, context_details=context_details, n_predict = self.config.ctx_size-len(tokens)-1, client_id=client_id, callback=partial(self.process_chunk,client_id = client_id) ) if self.config.enable_voice_service and self.config.auto_read and len(self.personality.audio_samples)>0: try: self.process_chunk("Generating voice output",MSG_TYPE.MSG_TYPE_STEP_START,client_id=client_id) from lollms.services.xtts.lollms_xtts import LollmsXTTS if self.tts is None: self.tts = LollmsXTTS(self, voice_samples_path=Path(__file__).parent.parent/"voices", xtts_base_url= self.config.xtts_base_url) language = convert_language_name(self.personality.language) self.tts.set_speaker_folder(Path(self.personality.audio_samples[0]).parent) fn = self.personality.name.lower().replace(' ',"_").replace('.','') fn = f"{fn}_{message_id}.wav" url = f"audio/{fn}" self.tts.tts_to_file(self.connections[client_id]["generated_text"], Path(self.personality.audio_samples[0]).name, f"{fn}", language=language) fl = f""" """ self.process_chunk("Generating voice output", MSG_TYPE.MSG_TYPE_STEP_END, {'status':True},client_id=client_id) self.process_chunk(fl,MSG_TYPE.MSG_TYPE_UI, client_id=client_id) """ self.info("Creating audio output",10) self.personality.step_start("Creating audio output") if not PackageManager.check_package_installed("tortoise"): PackageManager.install_package("tortoise-tts") from tortoise import utils, api import sounddevice as sd if self.tts is None: self.tts = api.TextToSpeech( kv_cache=True, half=True) reference_clips = [utils.audio.load_audio(str(p), 22050) for p in self.personality.audio_samples] tk = self.model.tokenize(self.connections[client_id]["generated_text"]) if len(tk)>100: chunk_size = 100 for i in range(0, len(tk), chunk_size): chunk = self.model.detokenize(tk[i:i+chunk_size]) if i==0: pcm_audio = self.tts.tts_with_preset(chunk, voice_samples=reference_clips, preset='fast').numpy().flatten() else: pcm_audio = np.concatenate([pcm_audio, self.tts.tts_with_preset(chunk, voice_samples=reference_clips, preset='ultra_fast').numpy().flatten()]) else: pcm_audio = self.tts.tts_with_preset(self.connections[client_id]["generated_text"], voice_samples=reference_clips, preset='fast').numpy().flatten() sd.play(pcm_audio, 22050) self.personality.step_end("Creating audio output") """ except Exception as ex: ASCIIColors.error("Couldn't read") trace_exception(ex) print() ASCIIColors.success("## Done Generation ##") print() except Exception as ex: trace_exception(ex) print() ASCIIColors.error("## Generation Error ##") print() self.cancel_gen = False # Send final message self.close_message(client_id) self.connections[client_id]["processing"]=False if self.connections[client_id]["schedule_for_deletion"]: del self.connections[client_id] ASCIIColors.success(f" ╔══════════════════════════════════════════════════╗ ") ASCIIColors.success(f" ║ Done ║ ") ASCIIColors.success(f" ╚══════════════════════════════════════════════════╝ ") if self.config.auto_title: d = self.connections[client_id]["current_discussion"] ttl = d.title() if ttl is None or ttl=="" or ttl=="untitled": title = self.make_discussion_title(d, client_id=client_id) d.rename(title) asyncio.run( self.sio.emit('disucssion_renamed',{ 'status': True, 'discussion_id':d.discussion_id, 'title':title }, to=client_id) ) self.busy=False else: ump = self.config.discussion_prompt_separator +self.config.user_name.strip() if self.config.use_user_name_in_discussions else self.personality.user_message_prefix self.cancel_gen = False #No discussion available ASCIIColors.warning("No discussion selected!!!") self.error("No discussion selected!!!", client_id=client_id) print() self.busy=False return ""