File size: 14,317 Bytes
eb62139 60929fd eb62139 60929fd eb62139 ae33464 eb62139 60929fd eb62139 60929fd eb62139 60929fd eb62139 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 |
import chainlit.data as cl_data
import asyncio
from typing import Any, Dict, no_type_check
import chainlit as cl
from modules.chat.llm_tutor import LLMTutor
from modules.chat.helpers import (
get_sources,
get_history_setup_llm,
)
import copy
import time
from langchain_community.callbacks import get_openai_callback
from config.config_manager import config_manager
USER_TIMEOUT = 60_000
SYSTEM = "System"
LLM = "AI Tutor"
AGENT = "Agent"
YOU = "User"
ERROR = "Error"
config = config_manager.get_config().dict()
class Chatbot:
def __init__(self, config):
"""
Initialize the Chatbot class.
"""
self.config = config
@no_type_check
async def setup_llm(self):
"""
Set up the LLM with the provided settings. Update the configuration and initialize the LLM tutor.
#TODO: Clean this up.
"""
start_time = time.time()
llm_settings = cl.user_session.get("llm_settings", {})
(
chat_profile,
retriever_method,
memory_window,
llm_style,
generate_follow_up,
chunking_mode,
) = (
llm_settings.get("chat_model"),
llm_settings.get("retriever_method"),
llm_settings.get("memory_window"),
llm_settings.get("llm_style"),
llm_settings.get("follow_up_questions"),
llm_settings.get("chunking_mode"),
)
chain = cl.user_session.get("chain")
memory_list = cl.user_session.get(
"memory",
(
list(chain.store.values())[0].messages
if len(chain.store.values()) > 0
else []
),
)
conversation_list = get_history_setup_llm(memory_list)
old_config = copy.deepcopy(self.config)
self.config["vectorstore"]["db_option"] = retriever_method
self.config["llm_params"]["memory_window"] = memory_window
self.config["llm_params"]["llm_style"] = llm_style
self.config["llm_params"]["llm_loader"] = chat_profile
self.config["llm_params"]["generate_follow_up"] = generate_follow_up
self.config["splitter_options"]["chunking_mode"] = chunking_mode
self.llm_tutor.update_llm(
old_config, self.config
) # update only llm attributes that are changed
self.chain = self.llm_tutor.qa_bot(
memory=conversation_list,
)
cl.user_session.set("chain", self.chain)
cl.user_session.set("llm_tutor", self.llm_tutor)
print("Time taken to setup LLM: ", time.time() - start_time)
@no_type_check
async def update_llm(self, new_settings: Dict[str, Any]):
"""
Update the LLM settings and reinitialize the LLM with the new settings.
Args:
new_settings (Dict[str, Any]): The new settings to update.
"""
cl.user_session.set("llm_settings", new_settings)
await self.inform_llm_settings()
await self.setup_llm()
async def make_llm_settings_widgets(self, config=None):
"""
Create and send the widgets for LLM settings configuration.
Args:
config: The configuration to use for setting up the widgets.
"""
config = config or self.config
await cl.ChatSettings(
[
cl.input_widget.Select(
id="chat_model",
label="Model Name (Default GPT-3)",
values=["local_llm", "gpt-3.5-turbo-1106", "gpt-4", "gpt-4o-mini"],
initial_index=[
"local_llm",
"gpt-3.5-turbo-1106",
"gpt-4",
"gpt-4o-mini",
].index(config["llm_params"]["llm_loader"]),
),
cl.input_widget.Select(
id="retriever_method",
label="Retriever (Default FAISS)",
values=["FAISS", "Chroma", "RAGatouille", "RAPTOR"],
initial_index=["FAISS", "Chroma", "RAGatouille", "RAPTOR"].index(
config["vectorstore"]["db_option"]
),
),
cl.input_widget.Slider(
id="memory_window",
label="Memory Window (Default 3)",
initial=3,
min=0,
max=10,
step=1,
),
cl.input_widget.Switch(
id="view_sources", label="View Sources", initial=False
),
cl.input_widget.Switch(
id="stream_response",
label="Stream response",
initial=config["llm_params"]["stream"],
),
cl.input_widget.Select(
id="chunking_mode",
label="Chunking mode",
values=["fixed", "semantic"],
initial_index=1,
),
cl.input_widget.Switch(
id="follow_up_questions",
label="Generate follow up questions",
initial=False,
),
cl.input_widget.Select(
id="llm_style",
label="Type of Conversation (Default Normal)",
values=["Normal", "ELI5"],
initial_index=0,
),
]
).send()
@no_type_check
async def inform_llm_settings(self):
"""
Inform the user about the updated LLM settings and display them as a message.
"""
llm_settings: Dict[str, Any] = cl.user_session.get("llm_settings", {})
llm_tutor = cl.user_session.get("llm_tutor")
settings_dict = {
"model": llm_settings.get("chat_model"),
"retriever": llm_settings.get("retriever_method"),
"memory_window": llm_settings.get("memory_window"),
"num_docs_in_db": (
len(llm_tutor.vector_db)
if llm_tutor and hasattr(llm_tutor, "vector_db")
else 0
),
"view_sources": llm_settings.get("view_sources"),
"follow_up_questions": llm_settings.get("follow_up_questions"),
}
print("Settings Dict: ", settings_dict)
await cl.Message(
author=SYSTEM,
content="LLM settings have been updated. You can continue with your Query!",
# elements=[
# cl.Text(
# name="settings",
# display="side",
# content=json.dumps(settings_dict, indent=4),
# language="json",
# ),
# ],
).send()
async def set_starters(self):
"""
Set starter messages for the chatbot.
"""
return [
cl.Starter(
label="recording on CNNs?",
message="Where can I find the recording for the lecture on Transformers?",
icon="/public/adv-screen-recorder-svgrepo-com.svg",
),
cl.Starter(
label="where's the slides?",
message="When are the lectures? I can't find the schedule.",
icon="/public/alarmy-svgrepo-com.svg",
),
cl.Starter(
label="Due Date?",
message="When is the final project due?",
icon="/public/calendar-samsung-17-svgrepo-com.svg",
),
cl.Starter(
label="Explain backprop.",
message="I didn't understand the math behind backprop, could you explain it?",
icon="/public/acastusphoton-svgrepo-com.svg",
),
]
def rename(self, orig_author: str):
"""
Rename the original author to a more user-friendly name.
Args:
orig_author (str): The original author's name.
Returns:
str: The renamed author.
"""
rename_dict = {"Chatbot": LLM}
return rename_dict.get(orig_author, orig_author)
async def start(self):
"""
Start the chatbot, initialize settings widgets,
and display and load previous conversation if chat logging is enabled.
"""
start_time = time.time()
await self.make_llm_settings_widgets(self.config) # Reload the settings widgets
# TODO: remove self.user with cl.user_session.get("user")
self.user = {
"user_id": "guest",
"session_id": cl.context.session.thread_id,
}
memory = cl.user_session.get("memory", [])
self.llm_tutor = LLMTutor(self.config, user=self.user)
self.chain = self.llm_tutor.qa_bot(
memory=memory,
)
self.question_generator = self.llm_tutor.question_generator
cl.user_session.set("llm_tutor", self.llm_tutor)
cl.user_session.set("chain", self.chain)
print("Time taken to start LLM: ", time.time() - start_time)
async def stream_response(self, response):
"""
Stream the response from the LLM.
Args:
response: The response from the LLM.
"""
msg = cl.Message(content="")
await msg.send()
output = {}
for chunk in response:
if "answer" in chunk:
await msg.stream_token(chunk["answer"])
for key in chunk:
if key not in output:
output[key] = chunk[key]
else:
output[key] += chunk[key]
return output
async def main(self, message):
"""
Process and Display the Conversation.
Args:
message: The incoming chat message.
"""
start_time = time.time()
chain = cl.user_session.get("chain")
token_count = 0 # initialize token count
if not chain:
await self.start() # start the chatbot if the chain is not present
chain = cl.user_session.get("chain")
# update user info with last message time
llm_settings = cl.user_session.get("llm_settings", {})
view_sources = llm_settings.get("view_sources", False)
stream = llm_settings.get("stream_response", False)
stream = False # Fix streaming
user_query_dict = {"input": message.content}
# Define the base configuration
cb = cl.AsyncLangchainCallbackHandler()
chain_config = {
"configurable": {
"user_id": self.user["user_id"],
"conversation_id": self.user["session_id"],
"memory_window": self.config["llm_params"]["memory_window"],
},
"callbacks": (
[cb]
if cl_data._data_layer and self.config["chat_logging"]["callbacks"]
else None
),
}
with get_openai_callback() as token_count_cb:
if stream:
res = chain.stream(user_query=user_query_dict, config=chain_config)
res = await self.stream_response(res)
else:
res = await chain.invoke(
user_query=user_query_dict,
config=chain_config,
)
token_count += token_count_cb.total_tokens
answer = res.get("answer", res.get("result"))
answer_with_sources, source_elements, sources_dict = get_sources(
res, answer, stream=stream, view_sources=view_sources
)
answer_with_sources = answer_with_sources.replace("$$", "$")
print("Time taken to process the message: ", time.time() - start_time)
actions = []
if self.config["llm_params"]["generate_follow_up"]:
start_time = time.time()
cb_follow_up = cl.AsyncLangchainCallbackHandler()
config = {
"callbacks": (
[cb_follow_up]
if cl_data._data_layer and self.config["chat_logging"]["callbacks"]
else None
)
}
with get_openai_callback() as token_count_cb:
list_of_questions = await self.question_generator.generate_questions(
query=user_query_dict["input"],
response=answer,
chat_history=res.get("chat_history"),
context=res.get("context"),
config=config,
)
token_count += token_count_cb.total_tokens
for question in list_of_questions:
actions.append(
cl.Action(
name="follow up question",
value="example_value",
description=question,
label=question,
)
)
print("Time taken to generate questions: ", time.time() - start_time)
print("Total Tokens Used: ", token_count)
await cl.Message(
content=answer_with_sources,
elements=source_elements,
author=LLM,
actions=actions,
).send()
async def on_follow_up(self, action: cl.Action):
user = cl.user_session.get("user")
message = await cl.Message(
content=action.description,
type="user_message",
author=user.identifier,
).send()
async with cl.Step(
name="on_follow_up", type="run", parent_id=message.id
) as step:
await self.main(message)
step.output = message.content
chatbot = Chatbot(config=config)
async def start_app():
cl.set_starters(chatbot.set_starters)
cl.author_rename(chatbot.rename)
cl.on_chat_start(chatbot.start)
cl.on_message(chatbot.main)
cl.on_settings_update(chatbot.update_llm)
cl.action_callback("follow up question")(chatbot.on_follow_up)
loop = asyncio.get_event_loop()
if loop.is_running():
asyncio.ensure_future(start_app())
else:
asyncio.run(start_app())
|