lollms-webui/events/lollms_generation_events.py

261 lines
10 KiB
Python
Raw Normal View History

"""
project: lollms
file: lollms_generation_events.py
author: ParisNeo
description:
This module contains a set of Socketio routes that provide information about the Lord of Large Language and Multimodal Systems (LoLLMs) Web UI
application. These routes are specific to text generation operation
"""
2024-12-19 12:48:57 +00:00
import os
import threading
from datetime import datetime
from pathlib import Path
from typing import List
import pkg_resources
2024-12-19 12:48:57 +00:00
import socketio
from ascii_colors import ASCIIColors
from fastapi import APIRouter, HTTPException, Request
from fastapi.responses import FileResponse
from lollms.binding import BindingBuilder, InstallOption
2024-08-14 20:15:45 +00:00
from lollms.personality import AIPersonality
2024-03-28 22:58:51 +00:00
from lollms.security import forbid_remote_access
2024-12-19 12:48:57 +00:00
from lollms.server.elf_server import LOLLMSElfServer
from lollms.types import MSG_OPERATION_TYPE, SENDER_TYPES
from lollms.utilities import (convert_language_name,
find_first_available_file_index, gc, load_config,
trace_exception)
from pydantic import BaseModel
from lollms_webui import LOLLMSWebUI
router = APIRouter()
lollmsElfServer = LOLLMSWebUI.get_instance()
# ----------------------------------- events -----------------------------------------
2024-12-19 12:48:57 +00:00
def add_events(sio: socketio):
2024-03-28 22:58:51 +00:00
forbid_remote_access(lollmsElfServer)
2024-12-19 12:48:57 +00:00
@sio.on("generate_msg")
def handle_generate_msg(sid, data, use_threading=True):
client_id = sid
lollmsElfServer.cancel_gen = False
client = lollmsElfServer.session.get_client(client_id)
2024-12-19 12:48:57 +00:00
client.generated_text = ""
client.cancel_generation = False
client.continuing = False
client.first_chunk = True
if not lollmsElfServer.model:
ASCIIColors.error("Model not selected. Please select a model")
2024-12-19 12:48:57 +00:00
lollmsElfServer.error(
"Model not selected. Please select a model", client_id=client_id
)
return
if not lollmsElfServer.busy:
if lollmsElfServer.session.get_client(client_id).discussion is None:
if lollmsElfServer.db.does_last_discussion_have_messages():
2024-12-19 12:48:57 +00:00
lollmsElfServer.session.get_client(client_id).discussion = (
lollmsElfServer.db.create_discussion()
)
else:
2024-12-19 12:48:57 +00:00
lollmsElfServer.session.get_client(client_id).discussion = (
lollmsElfServer.db.load_last_discussion()
)
prompt = data["prompt"]
2024-04-20 22:13:15 +00:00
try:
nb_tokens = len(lollmsElfServer.model.tokenize(prompt))
except:
nb_tokens = None
2024-12-19 12:48:57 +00:00
created_at = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
ump = (
lollmsElfServer.config.discussion_prompt_separator
+ lollmsElfServer.config.user_name.strip()
if lollmsElfServer.config.use_user_name_in_discussions
else lollmsElfServer.personality.user_message_prefix
)
message = lollmsElfServer.session.get_client(
client_id
).discussion.add_message(
message_type=MSG_OPERATION_TYPE.MSG_OPERATION_TYPE_SET_CONTENT.value,
sender_type=SENDER_TYPES.SENDER_TYPES_USER.value,
sender=ump.replace(
lollmsElfServer.config.discussion_prompt_separator, ""
).replace(":", ""),
content=prompt,
2024-08-14 20:15:45 +00:00
steps=[],
metadata=None,
2024-04-20 22:13:15 +00:00
parent_message_id=lollmsElfServer.message_id,
created_at=created_at,
2024-12-19 12:48:57 +00:00
nb_tokens=nb_tokens,
)
2024-12-19 12:48:57 +00:00
ASCIIColors.green(
"Starting message generation by " + lollmsElfServer.personality.name
)
2024-05-22 13:55:43 +00:00
if use_threading:
2024-12-19 12:48:57 +00:00
client.generation_thread = threading.Thread(
target=lollmsElfServer.start_message_generation,
args=(message, message.id, client_id),
)
2024-05-22 13:55:43 +00:00
client.generation_thread.start()
else:
lollmsElfServer.start_message_generation(message, message.id, client_id)
2024-12-19 12:48:57 +00:00
2024-01-18 07:57:00 +00:00
# lollmsElfServer.sio.sleep(0.01)
ASCIIColors.info("Started generation task")
2024-12-19 12:48:57 +00:00
lollmsElfServer.busy = True
# tpe = threading.Thread(target=lollmsElfServer.start_message_generation, args=(message, message_id, client_id))
# tpe.start()
else:
lollmsElfServer.error("I am busy. Come back later.", client_id=client_id)
2024-12-19 12:48:57 +00:00
@sio.on("generate_msg_with_internet")
def generate_msg_with_internet(sid, data):
2024-03-15 00:34:21 +00:00
client_id = sid
lollmsElfServer.cancel_gen = False
client = lollmsElfServer.session.get_client(client_id)
2024-12-19 12:48:57 +00:00
client.generated_text = ""
client.cancel_generation = False
client.continuing = False
client.first_chunk = True
2024-03-15 00:34:21 +00:00
if not lollmsElfServer.model:
ASCIIColors.error("Model not selected. Please select a model")
2024-12-19 12:48:57 +00:00
lollmsElfServer.error(
"Model not selected. Please select a model", client_id=client_id
)
2024-03-15 00:34:21 +00:00
return
if not lollmsElfServer.busy:
if lollmsElfServer.session.get_client(client_id).discussion is None:
if lollmsElfServer.db.does_last_discussion_have_messages():
2024-12-19 12:48:57 +00:00
lollmsElfServer.session.get_client(client_id).discussion = (
lollmsElfServer.db.create_discussion()
)
2024-03-15 00:34:21 +00:00
else:
2024-12-19 12:48:57 +00:00
lollmsElfServer.session.get_client(client_id).discussion = (
lollmsElfServer.db.load_last_discussion()
)
2024-03-15 00:34:21 +00:00
prompt = data["prompt"]
2024-04-20 22:13:15 +00:00
try:
nb_tokens = len(lollmsElfServer.model.tokenize(prompt))
except:
nb_tokens = None
2024-12-19 12:48:57 +00:00
created_at = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
ump = (
lollmsElfServer.config.discussion_prompt_separator
+ lollmsElfServer.config.user_name.strip()
if lollmsElfServer.config.use_user_name_in_discussions
else lollmsElfServer.personality.user_message_prefix
)
message = lollmsElfServer.session.get_client(
client_id
).discussion.add_message(
message_type=MSG_OPERATION_TYPE.MSG_OPERATION_TYPE_SET_CONTENT.value,
sender_type=SENDER_TYPES.SENDER_TYPES_USER.value,
sender=ump.replace(
lollmsElfServer.config.discussion_prompt_separator, ""
).replace(":", ""),
content=prompt,
steps=[],
metadata=None,
2024-04-20 22:13:15 +00:00
parent_message_id=lollmsElfServer.message_id,
created_at=created_at,
2024-12-19 12:48:57 +00:00
nb_tokens=nb_tokens,
2024-03-15 00:34:21 +00:00
)
2024-12-19 12:48:57 +00:00
ASCIIColors.green(
"Starting message generation by " + lollmsElfServer.personality.name
)
client.generation_thread = threading.Thread(
target=lollmsElfServer.start_message_generation,
args=(message, message.id, client_id, False, None, True),
)
2024-03-15 00:34:21 +00:00
client.generation_thread.start()
2024-12-19 12:48:57 +00:00
2024-03-15 00:34:21 +00:00
# lollmsElfServer.sio.sleep(0.01)
ASCIIColors.info("Started generation task")
2024-12-19 12:48:57 +00:00
lollmsElfServer.busy = True
# tpe = threading.Thread(target=lollmsElfServer.start_message_generation, args=(message, message_id, client_id))
# tpe.start()
2024-03-15 00:34:21 +00:00
else:
lollmsElfServer.error("I am busy. Come back later.", client_id=client_id)
2024-12-19 12:48:57 +00:00
@sio.on("generate_msg_from")
2024-01-07 14:27:14 +00:00
def handle_generate_msg_from(sid, data):
client_id = sid
client = lollmsElfServer.session.get_client(client_id)
lollmsElfServer.cancel_gen = False
2024-12-19 12:48:57 +00:00
client.continuing = False
client.first_chunk = True
if lollmsElfServer.session.get_client(client_id).discussion is None:
ASCIIColors.warning("Please select a discussion")
2024-12-19 12:48:57 +00:00
lollmsElfServer.error(
"Please select a discussion first", client_id=client_id
)
return
2024-12-19 12:48:57 +00:00
id_ = data["id"]
generation_type = data.get("msg_type", None)
if id_ == -1:
message = lollmsElfServer.session.get_client(
client_id
).discussion.current_message
else:
2024-12-19 12:48:57 +00:00
message = lollmsElfServer.session.get_client(
client_id
).discussion.load_message(id_)
if message is None:
2024-12-19 12:48:57 +00:00
return
client.generation_thread = threading.Thread(
target=lollmsElfServer.start_message_generation,
args=(message, message.id, client_id, False, generation_type),
)
client.generation_thread.start()
2024-12-19 12:48:57 +00:00
@sio.on("continue_generate_msg_from")
2024-01-07 14:27:14 +00:00
def handle_continue_generate_msg_from(sid, data):
client_id = sid
client = lollmsElfServer.session.get_client(client_id)
lollmsElfServer.cancel_gen = False
2024-12-19 12:48:57 +00:00
client.continuing = True
client.first_chunk = True
if lollmsElfServer.session.get_client(client_id).discussion is None:
ASCIIColors.yellow("Please select a discussion")
lollmsElfServer.error("Please select a discussion", client_id=client_id)
return
2024-12-19 12:48:57 +00:00
id_ = data["id"]
if id_ == -1:
message = lollmsElfServer.session.get_client(
client_id
).discussion.current_message
else:
2024-12-19 12:48:57 +00:00
message = lollmsElfServer.session.get_client(
client_id
).discussion.load_message(id_)
2024-12-19 12:48:57 +00:00
client.generated_text = message.content
client.generation_thread = threading.Thread(
target=lollmsElfServer.start_message_generation,
args=(message, message.id, client_id, True),
)
client.generation_thread.start()
2024-12-19 12:48:57 +00:00
# add functions to lollm
lollmsElfServer.handle_generate_msg = handle_generate_msg
lollmsElfServer.generate_msg_with_internet = generate_msg_with_internet
lollmsElfServer.handle_continue_generate_msg_from = (
handle_continue_generate_msg_from
)