File size: 4,970 Bytes
232abe6 |
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 |
import os
import logging
from llama_index import download_loader
from llama_index import (
Document,
LLMPredictor,
PromptHelper,
QuestionAnswerPrompt,
RefinePrompt,
)
import colorama
import PyPDF2
from tqdm import tqdm
from modules.presets import *
from modules.utils import *
def get_index_name(file_src):
file_paths = [x.name for x in file_src]
file_paths.sort(key=lambda x: os.path.basename(x))
md5_hash = hashlib.md5()
for file_path in file_paths:
with open(file_path, "rb") as f:
while chunk := f.read(8192):
md5_hash.update(chunk)
return md5_hash.hexdigest()
def block_split(text):
blocks = []
while len(text) > 0:
blocks.append(Document(text[:1000]))
text = text[1000:]
return blocks
def get_documents(file_src):
documents = []
logging.debug("Loading documents...")
logging.debug(f"file_src: {file_src}")
for file in file_src:
filepath = file.name
filename = os.path.basename(filepath)
file_type = os.path.splitext(filepath)[1]
logging.info(f"loading file: {filename}")
if file_type == ".pdf":
logging.debug("Loading PDF...")
try:
from modules.pdf_func import parse_pdf
from modules.config import advance_docs
two_column = advance_docs["pdf"].get("two_column", False)
pdftext = parse_pdf(filepath, two_column).text
except:
pdftext = ""
with open(filepath, 'rb') as pdfFileObj:
pdfReader = PyPDF2.PdfReader(pdfFileObj)
for page in tqdm(pdfReader.pages):
pdftext += page.extract_text()
text_raw = pdftext
elif file_type == ".docx":
logging.debug("Loading Word...")
DocxReader = download_loader("DocxReader")
loader = DocxReader()
text_raw = loader.load_data(file=filepath)[0].text
elif file_type == ".epub":
logging.debug("Loading EPUB...")
EpubReader = download_loader("EpubReader")
loader = EpubReader()
text_raw = loader.load_data(file=filepath)[0].text
elif file_type == ".xlsx":
logging.debug("Loading Excel...")
text_raw = excel_to_string(filepath)
else:
logging.debug("Loading text file...")
with open(filepath, "r", encoding="utf-8") as f:
text_raw = f.read()
text = add_space(text_raw)
# text = block_split(text)
# documents += text
documents += [Document(text)]
logging.debug("Documents loaded.")
return documents
def construct_index(
api_key,
file_src,
max_input_size=4096,
num_outputs=5,
max_chunk_overlap=20,
chunk_size_limit=600,
embedding_limit=None,
separator=" "
):
from langchain.chat_models import ChatOpenAI
from llama_index import GPTSimpleVectorIndex, ServiceContext
os.environ["OPENAI_API_KEY"] = api_key
chunk_size_limit = None if chunk_size_limit == 0 else chunk_size_limit
embedding_limit = None if embedding_limit == 0 else embedding_limit
separator = " " if separator == "" else separator
llm_predictor = LLMPredictor(
llm=ChatOpenAI(model_name="gpt-3.5-turbo-0301", openai_api_key=api_key)
)
prompt_helper = PromptHelper(max_input_size = max_input_size, num_output = num_outputs, max_chunk_overlap = max_chunk_overlap, embedding_limit=embedding_limit, chunk_size_limit=600, separator=separator)
index_name = get_index_name(file_src)
if os.path.exists(f"./index/{index_name}.json"):
logging.info("找到了缓存的索引文件，加载中……")
return GPTSimpleVectorIndex.load_from_disk(f"./index/{index_name}.json")
else:
try:
documents = get_documents(file_src)
logging.info("构建索引中……")
with retrieve_proxy():
service_context = ServiceContext.from_defaults(llm_predictor=llm_predictor, prompt_helper=prompt_helper, chunk_size_limit=chunk_size_limit)
index = GPTSimpleVectorIndex.from_documents(
documents, service_context=service_context
)
logging.debug("索引构建完成！")
os.makedirs("./index", exist_ok=True)
index.save_to_disk(f"./index/{index_name}.json")
logging.debug("索引已保存至本地!")
return index
except Exception as e:
logging.error("索引构建失败！", e)
print(e)
return None
def add_space(text):
punctuations = {"，": "， ", "。": "。 ", "？": "？ ", "！": "！ ", "：": "： ", "；": "； "}
for cn_punc, en_punc in punctuations.items():
text = text.replace(cn_punc, en_punc)
return text
|