File size: 42,521 Bytes
0061c9d |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 |
import os
import tempfile
from datetime import datetime
import webbrowser
from tkinter import Toplevel
import warnings
import faiss,logging
import numpy as np
import wandb
from typing import List, Dict, Any, Optional, Union
from git import Repo
import plotly.graph_objects as go
import numpy as np
from sklearn.decomposition import PCA
import requests
from rich import print as rp
from rich.progress import Progress, TextColumn, BarColumn, TimeRemainingColumn
from dotenv import load_dotenv, find_dotenv
import speech_recognition
from TTS.api import TTS
from sklearn.decomposition import PCA
from playsound import playsound
from hugchat import hugchat
from hugchat.login import Login
from langchain_core.documents import Document
from langchain_community.llms.huggingface_endpoint import HuggingFaceEndpoint
from langchain_community.llms.huggingface_hub import HuggingFaceHub
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain.chains import create_retrieval_chain
from langchain_community.document_loaders import (
PyPDFLoader,
UnstructuredHTMLLoader,
UnstructuredWordDocumentLoader,
TextLoader,
PythonLoader
)
from langchain.retrievers import TimeWeightedVectorStoreRetriever
from langchain_community.docstore.in_memory import InMemoryDocstore
from langchain.text_splitter import RecursiveCharacterTextSplitter, Language,CharacterTextSplitter
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain.vectorstores.base import VectorStore
from langchain.retrievers import MultiQueryRetriever, ContextualCompressionRetriever
from langchain.retrievers.self_query.base import SelfQueryRetriever
from langchain.retrievers.document_compressors import LLMChainExtractor, DocumentCompressorPipeline
from langchain_community.document_transformers import EmbeddingsRedundantFilter
from langchain.retrievers.document_compressors import EmbeddingsFilter
import plotly.graph_objs as go
from langchain.chains import LLMChain
# Load environment variables
load_dotenv(find_dotenv())
warnings.filterwarnings("ignore")
os.environ['FAISS_NO_AVX2'] = '1'
os.environ["USER_AGENT"] = os.getenv("USER_AGENT")
os.environ["HUGGINGFACEHUB_API_TOKEN"] = os.getenv("HUGGINGFACEHUB_API_TOKEN")
wandb.require("core")
# Import system prompts
from system_prompts import __all__ as prompts
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline, GPT2LMHeadModel, GPT2TokenizerFast
from langchain_huggingface import HuggingFacePipeline
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
class LLMChatBot:
def __init__(self, email, password, cookie_path_dir='./cookies/', default_llm=1, default_system_prompt='default_rag_prompt'):
self.email = email
self.password = password
self.current_model = 1
self.current_system_prompt=default_system_prompt
self.cookie_path_dir = cookie_path_dir
self.cookies = self.login()
self.default_llm = default_llm
self.chatbot = hugchat.ChatBot(cookies=self.cookies.get_dict(), default_llm=default_llm,system_prompt=prompts[default_system_prompt])
self.conversation_id=None
self.check_conv_id(self.conversation_id)
rp("[self.conversation_id:{self.conversation_id}]")
def check_conv_id(self, id=None):
if not self.conversation_id and not id:
self.conversation_id = self.chatbot.new_conversation(modelIndex=self.current_model,system_prompt=self.current_system_prompt)
else:
if id:
self.conversation_id=id
self.chatbot.change_conversation(self.conversation_id)
elif not self.chatbot.get_conversation_info(self.conversation_id) == self.chatbot.get_conversation_info():
self.chatbot.change_conversation(self.conversation_id)
return self.conversation_id
def login(self):
rp("Attempting to log in...")
sign = Login(self.email, self.password)
try:
cookies = sign.login(cookie_dir_path=self.cookie_path_dir, save_cookies=True)
rp("Login successful!")
return cookies
except Exception as e:
rp(f"Login failed: {e}")
rp("Attempting manual login with requests...")
self.manual_login()
raise
def manual_login(self):
login_url = "https://huggingface.co/login"
session = requests.Session()
response = session.get(login_url)
rp("Response Cookies:", response.cookies)
rp("Response Content:", response.content.decode())
csrf_token = response.cookies.get('csrf_token')
if not csrf_token:
rp("CSRF token not found in cookies.")
return
login_data = {
'email': self.email,
'password': self.password,
'csrf_token': csrf_token
}
response = session.post(login_url, data=login_data)
if response.ok:
rp("Manual login successful!")
else:
rp("Manual login failed!")
def setup_speech_recognition(self):
self.recognizer = speech_recognition.Recognizer()
def setup_tts(self, model_name="tts_models/en/ljspeech/fast_pitch"):
self.tts = TTS(model_name=model_name)
def chat(self, message):
return self.chatbot.chat(message)
def query(self,message, web_search=False, stream=False,use_cache=True):
return self.chatbot.query(
text=message,
web_search = web_search,
temperature = 0.1,
top_p = 0.95,
repetition_penalty = 1.2,
top_k = 50,
truncate = 1000,
watermark = False,
max_new_tokens = 1024,
stop = ["</s>"],
return_full_text = False,
stream = stream,
_stream_yield_all = False,
use_cache = False,
is_retry = False,
retry_count = 5,
conversation = None
)
def stream_response(self, message):
for resp in self.query(message, stream=True):
rp(resp)
def web_search(self, query):
query_result = self.query(query, web_search=True)
results = []
for source in query_result.web_search_sources:
results.append({
'link': source.link,
'title': source.title,
'hostname': source.hostname
})
return results
def create_new_conversation(self, switch_to=True):
return self.chatbot.new_conversation(switch_to=switch_to, modelIndex=self.current_model, system_prompt=self.current_system_prompt)
def get_remote_conversations(self):
return self.chatbot.get_remote_conversations(replace_conversation_list=True)
def get_local_conversations(self):
return self.chatbot.get_conversation_list()
def get_available_models(self):
return self.chatbot.get_available_llm_models()
def switch_model(self, index):
self.chatbot.switch_llm(index)
def switch_conversation(self, id):
self.conv_id = id
self.chatbot.change_conversation(self.conv_id)
def get_assistants(self):
return self.chatbot.get_assistant_list_by_page(1)
def switch_role(self, system_prompt, model_id=1):
self.chatbot.delete_all_conversations()
self.check_conv_id = self.chatbot.new_conversation(switch_to=True, system_prompt=system_prompt, modelIndex=model_id)
return self.check_conv_id
def __run__(self, message):
if not self.conversation_id:
self.conversation_id = self.chatbot.new_conversation(modelIndex=self.current_model,
system_prompt=self.current_system_prompt,
switch_to=True)
return self.query(message)
def __call__(self, message):
if not self.conversation_id:
self.conversation_id = self.chatbot.new_conversation(modelIndex=self.current_model,
system_prompt=self.current_system_prompt,
switch_to=True)
return self.chat(message)
class AdvancedVectorStore:
def __init__(self,
embedding_model: str = "all-MiniLM-L6-v2",
email: str = None,
password: str = None,
chunk_size=384,
chunk_overlap=0,
device='cpu',
normalize_embeddings=True,
log_level=logging.INFO,
log_file='AdvancedVectorStore.log',
logs_dir='./logs',
test_input='./test_input',
test_output='./test_output',
storage_dir='./vectorstore',
knowledge_dir='./knowledge',
repos_dir='./repos'
):
self.chunk_size = chunk_size
self.chunk_overlap = chunk_overlap
self.device = device
self.basic_splitter= RecursiveCharacterTextSplitter(chunk_size=self.chunk_size, chunk_overlap=self.chunk_overlap)
self.storage_dir=storage_dir
self.test_input=test_input
self.test_output=test_output
self.repos_dir=repos_dir
self.knowledge_dir=knowledge_dir
self.logs_dir=logs_dir
self.log_file=log_file
self.doc_ids = []
self.documents: List[Document] = []
self.embeddings = HuggingFaceEmbeddings(
model_name=embedding_model,
model_kwargs={'device': self.device},
encode_kwargs={'normalize_embeddings': normalize_embeddings}
)
self.qwen_llm = HuggingFaceHub(repo_id="Qwen/Qwen2-0.5B-Instruct", model_kwargs={"temperature": 0.5, "max_length": 512})
self.llm = HuggingFaceHub(repo_id="google-t5/t5-small", model_kwargs={"temperature": 0.5, "max_length": 512})
self.alpaca_llm = HuggingFaceHub(repo_id="reasonwang/google-flan-t5-small-alpaca", model_kwargs={"temperature": 0.1, "max_length": 512})
self.chatbot_llm = LLMChatBot(email, password, default_system_prompt= 'copilot_prompt') if email and password else None
rp("create_indexed_vectorstore:")
print(self.alpaca_llm("What is Deep Learning?"))
self.vectorstore, self.docstore, self.index = self.create_indexed_vectorstore(self.chunk_size)
self.document_count = 0
self.chunk_count = 0
self.setup_folders()
self.setup_logging(log_level,os.path.join(self.logs_dir,self.log_file))
self.logger.info("Initializing AdvancedVectorStore")
self.set_bot_role()
def setup_logging(self,level,file):
self.logger = logging.getLogger(__name__)
self.logger.setLevel(level)
# Create console handler and set level
ch = logging.StreamHandler()
ch.setLevel(level)
# Create file handler and set level
fh = logging.FileHandler(file)
fh.setLevel(level)
# Create formatter
formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
# Add formatter to console handler
ch.setFormatter(formatter)
# Add formatter to file handler
fh.setFormatter(formatter)
# Add handlers to logger
self.logger.addHandler(ch)
self.logger.addHandler(fh)
self.logger.info("Done settingload_documents_folder up logger for {__name__} [AdvancedVectorStore]")
def setup_folders(self):
self.dirs = [
self.test_input,
self.test_output,
self.logs_dir,
self.storage_dir,
self.knowledge_dir,
self.repos_dir
]
for d in self.dirs:
os.makedirs(d, exist_ok=True)
def set_bot_role(self,prompt='default_rag_prompt',context="",history=""):
self.chatbot_llm.current_system_prompt = prompts[prompt].replace("<<VSCONTEXT>>",context).replace("<<WSCONTEXT>>",history)
self.current_conversation_id=self.chatbot_llm.chatbot.new_conversation(system_prompt=self.chatbot_llm.current_system_prompt,
modelIndex=self.chatbot_llm.current_model,
switch_to=True)
#self.logger.info(f"Setting Bot Role!\n[{prompt}]")
""" result=self.chatbot_llm("Confirm you understand the TASK.")
self.logger.info(f"Test results chatbot role set:{result}") """
#rp(f"[Result:{result}]")
def load_documents(self, directory: str) -> None:
"""Load documents from a directory with specific loaders for each file type."""
loaders = {
".py": (PythonLoader, {}),
".txt": (TextLoader, {}),
".pdf": (PyPDFLoader, {}),
".html": (UnstructuredHTMLLoader, {}),
".docx": (UnstructuredWordDocumentLoader, {})
}
for root, _, files in os.walk(directory):
for file in files:
file_path = os.path.join(root, file)
file_extension = os.path.splitext(file)[1].lower()
if file_extension in loaders:
# Check if the file can be read as UTF-8
try:
with open(file_path, 'r', encoding='utf-8') as f:
f.read()
except (UnicodeDecodeError, IOError):
rp(f"Skipping non-UTF-8 or unreadable file: {file_path}")
continue
loader_class, loader_args = loaders[file_extension]
loader = loader_class(file_path, **loader_args)
self.documents.extend(loader.load())
def split_documents(self) -> None:
"""Split documents using appropriate splitters for each file type."""
splitters = {
".py": RecursiveCharacterTextSplitter.from_language(language=Language.PYTHON, chunk_size=self.chunk_size, chunk_overlap=self.chunk_overlap),
".txt": RecursiveCharacterTextSplitter(chunk_size=self.chunk_size, chunk_overlap=self.chunk_overlap),
".pdf": RecursiveCharacterTextSplitter(chunk_size=self.chunk_size, chunk_overlap=self.chunk_overlap),
".html": RecursiveCharacterTextSplitter.from_language(language=Language.HTML, chunk_size=self.chunk_size, chunk_overlap=self.chunk_overlap),
".docx": RecursiveCharacterTextSplitter(chunk_size=self.chunk_size, chunk_overlap=self.chunk_overlap)
}
split_docs = []
for doc in self.documents:
file_extension = os.path.splitext(doc.metadata.get("source", ""))[1].lower()
splitter = splitters.get(file_extension, RecursiveCharacterTextSplitter(chunk_size=self.chunk_size, chunk_overlap=self.chunk_overlap))
split_docs.extend(splitter.split_documents([doc]))
self.documents = split_docs
def create_vectorstore(self, store_type: str = "FAISS") -> None:
"""Create a vectorstore of the specified type."""
if store_type == "FAISS":
self.vectorstore = FAISS.from_documents(self.documents, self.embeddings)
else:
raise ValueError(f"Unsupported vectorstore type: {store_type}")
def create_indexed_vectorstore(self,embedding_size):
rp("Creating indexed vectorstore...")
#embedding_size = 384 # Size for all-MiniLM-L6-v2 embeddings
index = faiss.IndexFlatL2(embedding_size)
docstore = InMemoryDocstore({})
vectorstore = FAISS(
self.embeddings.embed_query,
index,
docstore,
{}
)
rp("Indexed vectorstore created.")
return vectorstore, docstore, index
def get_self_query_retriever(self, k: int = 4) -> SelfQueryRetriever:
"""Get a SelfQueryRetriever."""
if not self.vectorstore:
raise ValueError("Vectorstore not initialized. Call create_vectorstore() first.")
return SelfQueryRetriever.from_llm(
self.chatbot_llm.chatbot,
self.vectorstore,
document_contents="Document about various topics.",
metadata_field_info=[],
search_kwargs={"k": k}
)
def get_contextual_t5_compression_retriever(self, k: int = 4, similarity_threshold=0.78) -> ContextualCompressionRetriever:
"""Get a ContextualCompressionRetriever."""
base_compressor = LLMChainExtractor.from_llm(self.llm)
redundant_filter = EmbeddingsRedundantFilter(embeddings=self.embeddings, similarity_threshold=similarity_threshold)
relevant_filter = EmbeddingsFilter(embeddings=self.embeddings, similarity_threshold=similarity_threshold)
return ContextualCompressionRetriever(
name="CompressedRetriever",
base_compressor=DocumentCompressorPipeline(transformers=[self.basic_splitter, base_compressor, redundant_filter, relevant_filter]),
base_retriever=self.get_basic_retriever(k=k)
)
def get_contextual_qwen_compression_retriever(self, k=4, similarity_threshold=0.78):
# Initialize the components for the compressor pipeline
base_compressor = LLMChainExtractor.from_llm(self.qwen_llm)
redundant_filter = EmbeddingsRedundantFilter(embeddings=self.embeddings, similarity_threshold=similarity_threshold)
relevant_filter = EmbeddingsFilter(embeddings=self.embeddings, similarity_threshold=similarity_threshold)
# Create the ContextualCompressionRetriever
return ContextualCompressionRetriever(
name="CompressedRetriever",
base_compressor= DocumentCompressorPipeline(transformers=[self.basic_splitter, base_compressor, redundant_filter, relevant_filter]),
base_retriever=self.get_basic_retriever(k=k)
)
def get_contextual_compression_retriever(self, k: int = 4,similarity_threshold=0.78) -> ContextualCompressionRetriever:
"""Get a ContextualCompressionRetriever."""
base_compressor = LLMChainExtractor.from_llm(self.alpaca_llm)
redundant_filter = EmbeddingsRedundantFilter(embeddings=self.embeddings, similarity_threshold=similarity_threshold)
relevant_filter = EmbeddingsFilter(embeddings=self.embeddings, similarity_threshold=similarity_threshold)
return ContextualCompressionRetriever(
name="CompressedRetriever",
base_compressor=DocumentCompressorPipeline(transformers=[self.basic_splitter, base_compressor, redundant_filter, relevant_filter]),
base_retriever=self.get_basic_retriever(k=k)
)
def get_basic_retriever(self, k: int = 4) -> VectorStore:
"""Get a basic retriever from the vectorstore."""
if not self.vectorstore:
raise ValueError("Vectorstore not initialized. Call create_vectorstore() first.")
return self.vectorstore.as_retriever(search_kwargs={"k": k})
def get_multi_query_retriever(self, k: int = 4) -> MultiQueryRetriever:
"""Get a MultiQueryRetriever."""
if not self.vectorstore:
raise ValueError("Vectorstore not initialized. Call create_vectorstore() first.")
return MultiQueryRetriever.from_llm(
retriever=self.vectorstore.as_retriever(search_kwargs={"k": k}),
llm=self.chatbot_llm
)
def get_timed_retriever(self, k=1, decay_rate=0.0000000000000000000000001):
return TimeWeightedVectorStoreRetriever(
vectorstore=self.vectorstore, decay_rate=decay_rate, k=k
)
def set_current_retriever(self,mode='basic',k=4,sim_rate=0.78):
if mode == 'compressed':
retriever = self.get_contextual_compression_retriever(k, sim_rate)
elif mode == 'qwen_compressed':
retriever = self.get_contextual_qwen_compression_retriever(k, sim_rate)
elif mode == 't5_compressed':
retriever = self.get_contextual_t5_compression_retriever(k, sim_rate)
elif mode == 'self_query':
retriever = self.get_self_query_retriever(k)
elif mode == 'multi_query':
retriever = self.get_multi_query_retriever(k)
elif mode == 'time':
retriever = self.get_timed_retriever(k=1)
else:
retriever = self.get_basic_retriever(k)
#rp(retriever.get_prompts)
return retriever
def search(self, query: str, mode='basic', retriever: Optional[Any] = None, k: int = 4, sim_rate: float = 0.78) -> List[Document]:
"""Search the vectorstore using the specified retriever."""
if not retriever:
retriever = self.set_current_retriever(mode=mode, k=k, sim_rate=sim_rate)
return retriever.get_relevant_documents(query)
def add_documents(self, documents: List[Document]) -> None:
import uuid
"""Add new documents to the existing vectorstore."""
with Progress(
TextColumn("[progress.description]{task.description}"),
BarColumn(),
TextColumn("[green][progress.percentage]{task.percentage:>3.0f}%"),
TimeRemainingColumn()
) as progress:
task = progress.add_task("[cyan]Adding documents to vectorstore...", total=len(documents))
for id, doc in enumerate(documents):
#self.vectorstore.add_documents([doc])
metadata = doc.metadata
if not metadata:
metadata = {}
metadata["last_accessed_at"] = datetime.now()
new_doc = Document(page_content=doc.page_content, metadata=metadata)
nr=id
id = str(uuid.uuid4())
self.vectorstore.docstore.add({id: new_doc})
self.doc_ids.append(id)
self.set_current_retriever(mode='time', k=1).add_documents([new_doc])
total = self.index.ntotal
#self.logger.info(f"Added doc to vectorstore {new_doc.metadata['last_accessed_at']} with {total} id's so far.")
self.logger.info(f"Added doc to docstore[{nr}/{len(self.documents)}] with Id:{id} Path:{new_doc.metadata['source']}")
progress.update(task, advance=1)
rp(f"Added {len(documents)} documents to the vectorstore with index in doc_ids.")
def delete_documents(self, document_ids: List[str]) -> None:
"""Delete documents from the vectorstore by their IDs."""
for id in document_ids:
#self.logger.info(f"[Deleting DocumenId{id}...]")
self.vectorstore.delete(document_ids)
#self.logger.info(f"[Done! Saving Faiss...{id}]")
def save_vectorstore(self, path: str) -> None:
"""Save the vectorstore to disk."""
if not self.vectorstore:
raise ValueError("Vectorstore not initialized. Call create_vectorstore() first.")
#self.logger.info("[Saving Faiss...]")
self.vectorstore.save_local(path)
#self.logger.info(f"[Done! Saving Faiss to:{path}]")
def load_vectorstore(self, path: str) -> None:
"""Load the vectorstore from disk."""
#self.logger.info("Loading Faiss...")
self.vectorstore = FAISS.load_local(folder_path=path,
embeddings=self.embeddings,
allow_dangerous_deserialization=True)
#self.logger.info(f"[Done! Loading Faiss from:{path}]")
def create_retrieval_chain(self, prompt: str = "default_rag_prompt", retriever: Optional[Any] = None) -> Any:
"""Create a retrieval chain using the specified prompt and retriever."""
if not retriever:
retriever = self.get_basic_retriever()
combine_docs_chain = create_stuff_documents_chain(self.chatbot_llm, prompt=prompts[prompt])
return create_retrieval_chain(retriever, combine_docs_chain)
def run_retrieval_chain(self, chain: Any, query: str) -> Dict[str, Any]:
"""Run a retrieval chain with the given query."""
return chain.invoke({"input": query})
def generate_3d_scatterplot(self, num_points=1000):
"""
Generate a 3D scatter plot of the vector store content and log it to wandb.
:param num_points: Maximum number of points to plot (default: 1000)
:return: None (logs the plot to wandb)
"""
all_docs = self.get_all_documents()
if not all_docs:
raise ValueError("No documents found in the vector store.")
# Extract vectors and metadata from documents
vectors = []
doc_ids = []
for doc in all_docs:
if hasattr(doc, 'embedding') and doc.embedding is not None:
vectors.append(doc.embedding)
else:
vectors.append(self.embeddings.embed_query(doc.page_content))
doc_ids.append(doc.metadata.get('id', 'Unknown'))
vectors = np.array(vectors)
# If we have more vectors than requested points, sample randomly
if len(vectors) > num_points:
indices = np.random.choice(len(vectors), num_points, replace=False)
vectors = vectors[indices]
doc_ids = [doc_ids[i] for i in indices]
# Perform PCA to reduce to 3 dimensions
pca = PCA(n_components=3)
vectors_3d = pca.fit_transform(vectors)
# Initialize wandb run
wandb.init(project="vector_store_visualization")
# Create the Plotly figure
fig = go.Figure(data=[go.Scatter3d(
x=vectors_3d[:, 0],
y=vectors_3d[:, 1],
z=vectors_3d[:, 2],
mode="markers",
marker=dict(
size=[28.666666666666668, 20.666666666666668, 15.333333333333334, 17.666666666666668, 19.0, 17.666666666666668, 26.0, 21.0, 21.666666666666668, 27.0, 21.666666666666668, 16.666666666666668, 27.0, 14.0, 29.666666666666668, 22.0, 16.0, 28.0, 27.0, 25.333333333333332],
color=[28.666666666666668, 20.666666666666668, 15.333333333333334, 17.666666666666668, 19.0, 17.666666666666668, 26.0, 21.0, 21.666666666666668, 27.0, 21.666666666666668, 16.666666666666668, 27.0, 14.0, 29.666666666666668, 22.0, 16.0, 28.0, 27.0, 25.333333333333332],
colorscale='Viridis',
showscale=True,
colorbar=dict(x=0),
cmin=14,
cmax=20
),
text=self.doc_ids,
hoverinfo="text",
name="Document Vectors",
)])
# Update layout
fig.update_layout(
showlegend=True,
scene=dict(
xaxis_title="X Axis",
yaxis_title="Y Axis",
zaxis_title="Z Axis"
),
width=1200,
height=1000,
)
# Log the plot to wandb
wandb.log({"3D Scatter Plot": fig})
# Finish the wandb run
wandb.finish()
def load_documents_folder(self, folder_path):
rp("[Loading documents from cloned repository]")
self.load_documents(folder_path)
self.document_count=len(self.documents)
rp(f"Splitting {self.document_count} documents")
self.split_documents()
self.chunk_count=len(self.documents)
rp(f"Adding {self.chunk_count} document chunks to vectorstore")
self.add_documents(self.documents)
def load_github_repo(self, repo_url: str) -> None:
"""
Clone a GitHub repository to a temporary folder, load documents, and remove the folder.
"""
split=repo_url.split('/')
repo_name = split.pop()
author_name = split.pop()
new_repo_path=os.path.join(self.repos_dir,f"{author_name}_{repo_name}")
if not os.path.exists(new_repo_path):
rp(f'Cloning repository {repo_url} to {new_repo_path}')
Repo.clone_from(repo_url, new_repo_path)
rp("Loading documents from cloned repository")
self.load_documents(new_repo_path)
rp(f"Splitting {len(self.documents)} documents into chunks")
self.split_documents()
rp(f"Adding {len(self.documents)} documents to vectorstore")
self.add_documents(self.documents)
self.save_vectorstore(self.storage_dir)
self.load_vectorstore(self.storage_dir)
rp("Temporary folder removed")
else:
rp(f"Repository {repo_url} already exists in {new_repo_path}")
self.load_vectorstore(self.storage_dir)
def get_all_documents(self):
"""
Fetch all documents from the document store.
"""
all_docs = []
# Number of vectors in the index
num_vectors = self.index.ntotal
# Assuming 'd' is the dimensionality of the vectors
d = self.index.d
#rp(f"D:{d}")
# Retrieve all vectors (this part is straightforward if you have access to the original vectors)
retrieved_vectors = np.empty((num_vectors, d), dtype='float32')
for i in range(num_vectors):
retrieved_vectors[i] = self.index.reconstruct(i)
# Assuming you have a way to get the document IDs
# In a real scenario, you would maintain a mapping of FAISS index positions to document IDs
# Example: you might have an attribute like 'self.doc_ids' which is a list of IDs
retrieved_ids = self.doc_ids[:num_vectors] # Ensure you have this attribute properly maintained
#rp(f"Retrieved ids{retrieved_ids}")
# Fetch documents using the retrieved IDs
retrieved_docs = [self.docstore.search(doc_id) for doc_id in retrieved_ids]
# Collect all documents
all_docs.extend(retrieved_docs)
#for doc_id, doc in zip(retrieved_ids, retrieved_docs):
#rp(f"ID: {doc_id}, Document.page_content: {doc.page_content}, Document.metadata: {doc.metadata}")
return all_docs
def test_chat(self,text,context='This is a chat with a nice Senior programmer.',history='Your Birth as fresh outof the box agent.'):
self.set_bot_role(context=context,history=history)
return self.chatbot_llm(text)
def chat(self, message: str) -> str:
"""
Send a message to the HugChat bot and get a response.
:param message: The message to send to the bot
:return: The bot's response
"""
if not self.chatbot_llm:
raise ValueError("HugChat bot not initialized. Provide email and password when creating AdvancedVectorStore.")
return self.chatbot_llm.chat(message)
def setup_speech_recognition(self):
"""Set up speech recognition for the HugChat bot."""
if not self.chatbot_llm:
raise ValueError("HugChat bot not initialized. Provide email and password when creating AdvancedVectorStore.")
self.chatbot_llm.setup_speech_recognition()
def setup_tts(self, model_name="tts_models/en/ljspeech/fast_pitch"):
"""Set up text-to-speech for the HugChat bot."""
if not self.chatbot_llm:
raise ValueError("HugChat bot not initialized. Provide email and password when creating AdvancedVectorStore.")
self.chatbot_llm.setup_tts(model_name)
def voice_chat(self):
"""
Initiate a voice chat session with the HugChat bot.
"""
if not self.chatbot_llm or not hasattr(self.chatbot_llm, 'recognizer') or not hasattr(self.chatbot_llm, 'tts'):
raise ValueError("Speech recognition and TTS not set up. Call setup_speech_recognition() and setup_tts() first.")
rp("Voice chat initiated. Speak your message (or say 'exit' to end the chat).")
while True:
with speech_recognition.Microphone() as source:
rp("Listening...")
audio = self.chatbot_llm.recognizer.listen(source)
try:
user_input = self.chatbot_llm.recognizer.recognize_google(audio)
rp(f"You said: {user_input}")
if user_input.lower() == 'exit':
rp("Ending voice chat.")
break
response = self.chat(user_input)
rp(f"Bot: {response}")
# Generate speech from the bot's response
speech_file = "bot_response.wav"
self.chatbot_llm.tts.tts_to_file(text=response, file_path=speech_file)
playsound(speech_file)
os.remove(speech_file) # Clean up the temporary audio file
except speech_recognition.UnknownValueError:
rp("Sorry, I couldn't understand that. Please try again.")
except speech_recognition.RequestError as e:
rp(f"Could not request results from the speech recognition service; {e}")
def rag_chat(self, query: str, prompt: str = "default_rag_prompt") -> str:
"""
Perform a RAG (Retrieval-Augmented Generation) chat using the vectorstore and HugChat bot.
:param query: The user's query
:param prompt: The prompt to use for the retrieval chain (default: "default_rag_prompt")
:return: The bot's response
"""
if not self.vectorstore:
raise ValueError("Vectorstore not initialized. Call create_vectorstore() first.")
retriever = self.get_basic_retriever()
chain = self.create_retrieval_chain(prompt, retriever)
result = self.run_retrieval_chain(chain, query)
return result['answer']
def search_web(self):
search_query = input("Enter your web search query: ")
future_date = "July 12, 2024"
search_url = f"https://www.google.com/search?q={search_query}+before:{future_date}"
webbrowser.open(search_url)
rp(f"Search results for '{search_query}' on {future_date}:")
rp("=" * 50)
rp(search_url)
rp("=" * 50)
def advanced_rag_chatbot(self):
rp("Welcome to the Advanced RAG Chatbot!")
rp("This chatbot uses a compressed retriever and integrates all components of the vector store.")
rp("Type 'exit' to end the conversation.")
# Ensure the vectorstore is initialized
if self.vectorstore is None:
rp("Initializing vector store...")
self.vectorstore, self.docstore, self.index = self.create_indexed_vectorstore(self.chunk_size)
# Create a compressed retriever
# compressed_retriever = self.get_contextual_compression_retriever(k=5, similarity_threshold=0.75)
mode='basic'
k=5
similarity_threshold=0.75
retriever = self.set_current_retriever(mode=mode, k=k, sim_rate=similarity_threshold)
#basic_retriever = self.get_basic_retriever(k=4)
# Initialize conversation history
conversation_history = []
while True:
user_input = input("\nYou: ").strip()
if user_input.lower() == 'exit':
rp("Thank you for using the Advanced RAG Chatbot. Goodbye!")
break
rp("# Step 1: Retrieve relevant documents")
retrieved_docs = self.get_basic_retriever(k=4).get_relevant_documents(user_input)
rp("# Step 2: Prepare context from retrieved documents")
context = "\n".join([doc.page_content for doc in retrieved_docs])
rp("# Step 3: Prepare the prompt")
#prompt = prompts['default_rag_prompt']
self.set_bot_role(context=context, history=' '.join(conversation_history[-5:]))
rp("# Step 4: Generate response using the chatbot")
response = self.chatbot_llm(f"User:{user_input}\n")
rp(f"Chatbot: {response}")
# Update conversation history
conversation_history.append(f"User: {user_input}")
conversation_history.append(f"Chatbot: {response}")
# Step 5: Demonstrate use of individual components
rp("\nAdditional Information:")
rp(f'- Number of documents in docstore: {len(self.docstore.search("* *"))}')
rp(f"- Number of vectors in index: {self.index.ntotal}")
# Demonstrate direct use of vectorstore for similarity search
similar_docs = self.vectorstore.similarity_search(user_input, k=1)
similar_docs = self.vectorstore.similarity_search_with_relevance_scores(user_input,k=1)
if similar_docs:
rp(type(similar_docs))
rp(f"-[Most similar document: [{similar_docs[0].metadata.get('source', 'Unknown')}]]-")
# Generate a 3D scatter plot of the vectorstore content
#avs.generate_3d_scatterplot_wandb()
avs.generate_3d_scatterplot()
# Optional: Add user feedback loop
feedback = input("Was this response helpful? (yes/no): ").strip().lower()
if feedback == 'no':
rp("I'm sorry the response wasn't helpful. Let me try to improve it.")
# Here you could implement logic to refine the response or adjust the retrieval process
with open(file="./feedback_NO.txt",mode="a+")as f:
f.write(f"chat_feedback_NO\nChatHistory--->{' '.join(conversation_history[-10:])}")
# Example usage:
if __name__ == "__main__":
email = os.getenv("EMAIL")
password = os.getenv("PASSWD")
github_token = os.getenv("GITHUB_TOKEN")
# Initialize AdvancedVectorStore with HugChat bot
avs = AdvancedVectorStore(email=email, password=password)
# Create the indexed vectorstore
#avs.create_indexed_vectorstore()
# Clone a GitHub repository and load its contents
# avs.load_documents_folder("/nr_ywo/coding/voice_chat_rag_web/venv/lib/python3.10/site-packages/huggingface_hub/inference")
avs.load_documents_folder("/nr_ywo/coding/voice_chat_rag_web/venv/lib/python3.10/site-packages/hugchat")
avs.load_documents_folder("/nr_ywo/coding/voice_chat_rag_web/venv/lib/python3.10/site-packages/langchain/agents")
avs.load_documents_folder("/nr_ywo/coding/voice_chat_rag_web/venv/lib/python3.10/site-packages/langchain_experimental/autonomous_agents")
#avs.chatbot_llm.load_documents("/nr_ywo/coding/voice_chat_rag_web/test_input")
# avs.load_github_repo("https://github.com/bxck75/voice_chat_rag_web")
avs.save_vectorstore(path=avs.storage_dir)
avs.load_vectorstore(path=avs.storage_dir)
# rp document and chunk counts
#rp(f"Total documents: {avs.chunk_count / avs.chunk_size}")
#rp(f"Total chunks: {avs.chunk_count}")
#avs.logger.info(avs.chatbot_llm.current_model)
#avs.logger.info(avs.chatbot_llm.current_system_prompt)
retriever=avs.set_current_retriever(mode='basic',k=4)
comptriever=avs.set_current_retriever(mode='compression',k=4,sim_rate=0.87)
timetriever=avs.set_current_retriever(mode='time',k=1)
q="Demonstrate your knowledge of developing advanced AI scripts in OOP python. try to come up with cutting edge ideas"
rel_docs=retriever.invoke(input=q)
#okrp(f"[Raw Knowledge Retrieved:{rel_docs}]")
# Start the advanced RAG chatbot
avs.advanced_rag_chatbot()
# Perform a RAG chat
#rag_response = avs.rag_chat(query="Explain the concept of neural networks.")
#rp("RAG chat response:", rag_response)
# Set up speech recognition and TTS for voice chat
#avs.setup_speech_recognition()
#avs.setup_tts()
# Start a voice chat session
#avs.voice_chat()
"""
# Using different retrievers
multi_query_retriever = avs.get_multi_query_retriever()
results = avs.search("What is deep learning?", mode="multi_query")
rp("Multi-query retriever results:", results)
self_query_retriever = avs.get_self_query_retriever()
results = avs.search("Find documents about reinforcement learning", self_query_retriever)
rp("Self-query retriever results:", results)
contextual_compression_retriever = avs.get_contextual_compression_retriever()
results = avs.search("Explain the difference between supervised and unsupervised learning", contextual_compression_retriever)
rp("Contextual compression retriever results:", results)
"""
""" # Perform a basic search
k = 4
similarity_threshold = 0.78
q = "What is machine learning?"
basic_results = avs.search(q, mode='basic', k=k)
rp("Basic search results:", basic_results)
rp("self_query search results:", self_query_results)
rp("multi_query search results:", multi_results)
rp("Compressed search results:", commpressed_results)
"""
""" This advanced example demonstrates:
Use of the compressed retriever for efficient document retrieval.
Integration of conversation history for context-aware responses.
Direct use of the vectorstore for similarity search.
Access to the docstore and index for additional information.
A feedback loop to potentially improve responses (though the improvement logic is not implemented in this example).
This chatbot loop showcases how all components of the system can work together to provide informative responses based on the loaded documents. It also demonstrates how you can access and use individual components (docstore, index, vectorstore) for additional functionality or information.
To further optimize this system, you could consider:
Implementing caching mechanisms to speed up repeated queries.
Adding more sophisticated feedback handling to improve retrieval and response generation over time.
Implementing dynamic index updates if new information becomes available during the chat session.
Adding options for users to see the sources of information or request more details on specific topics.
This example provides a solid foundation that you can further customize and expand based on your specific needs and use cases. """ |