tables = page.find_tables(...)
for tab in tables:
    pandas_df = tab.to_pandas()
    # 保存Excel，生成描述，建Document

import base64
import os
import subprocess
from io import BytesIO
from pathlib import Path
import fitz
from PIL import Image
from pptx import Presentation
from .config import RagConfig
from .llms import moonshot_llm, vllm, deepseek_llm
from .ocr import ocr_file_to_text
def ocr_file_to_text_llm(file_path) -> str:
 """
 提取⽂件中的⽂本
 :param file_path:
 :return:
 """
 client = moonshot_llm()
 file_object = client.files.create(file=Path(file_path), purpose="file
-extract")
 file_content = client.files.content(file_id=file_object.id).json()
 return file_content.get("content")
def get_b64_image_from_path(image_path):
 with open(image_path, "rb") as image_file:
 return base64.b64encode(image_file.read()).decode('utf-8')
def is_image(file_path):
 try:
 with Image.open(file_path) as img:
 img.verify() # 驗證⽂件是否是有效的圖像
 return True
 except Exception as e:
 print(f"Error: {e}")
 return False
def process_table(file):
 """
 解析表格:
 1、藉助ocr直接識別表格內容
2、藉助多模態⼤模型識別
 :param file:
 :return: 表格內容及對錶格的描述信息
 """
 content = ocr_file_to_text_llm(file)
 llm = deepseek_llm()
 response = llm.complete(f"你的職責是解釋表格。"
 f"你是將線性化表格轉換成簡單中⽂⽂本供⼤型語⾔模型（L
LMs）使⽤的專家。"
 f"請解釋以下線性化表格： {content}")
 return content, response.tex

from typing import Dict
from llama_index.llms.openai import OpenAI as DeepSeeK
from openai import OpenAI
from .config import RagConfig
from llama_index.llms.openai.utils import ALL_AVAILABLE_MODELS, CHAT_MODEL
S
DEEPSEEK_MODELS: Dict[str, int] = {
 "deepseek-chat": 64000,
}
ALL_AVAILABLE_MODELS.update(DEEPSEEK_MODELS)
CHAT_MODELS.update(DEEPSEEK_MODELS)
def moonshot_llm(**kwargs):
 llm = OpenAI(api_key=RagConfig.moonshot_api_key,
 base_url="https://api.moonshot.cn/v1",
 **kwargs)
 return llm
def deepseek_llm(**kwargs):
 llm = DeepSeeK(api_key=RagConfig.deepseek_api_key,
 model="deepseek-chat",
 api_base="https://api.deepseek.com/v1",
 temperature=0.7,
 **kwargs)
 return llm
def vllm(**kwargs):
 from openai import OpenAI
 llm = OpenAI(api_key=RagConfig.vllm_api_key,
 base_url=RagConfig.vllm_base_url,
 **kwargs)
 return llm

import os
from pydantic import BaseModel, Field
class RAGConfig(BaseModel):
 """
 RAG配置模型。
 該模型使⽤Pydantic庫來定義和驗證配置參數。每個字段都設置了默認值和描述，
 以便於理解和維護。這些配置參數對於RAG系統的運⾏⾄關重要，包括數據庫連接、
 模型維度、API密鑰等。
 """
 # Milvus URI，⽤於數據庫連接
 milvus_uri: str = Field(default=os.getenv("MILVUS_URI"), description=
"Milvus URI")
 # 嵌⼊模型維度
 embedding_model_dim: int = Field(default=512, description="Embedding m
odel dimension")
 # Moonshot API密鑰
 moonshot_api_key: str = Field(default=os.getenv("MOONSHOT_API_KEY"), d
escription="Moonshot API key")
 # DeepSeek API密鑰
 deepseek_api_key: str = Field(default=os.getenv("DEEPSEEK_API_KEY"), d
escription="DeepSeek API key")
 # Postgres數據庫連接字符串
 pg_connection_string: str = Field(default=os.getenv("PG_CONNECTION_STR
ING"), description="Postgres connection string")
 # OCR下載⽬錄
 ocr_download_dir: str = Field(default=os.getenv("OCR_DOWNLOAD_PATH"),
description="OCR download directory")
 # OCR基礎URL
 ocr_base_url: str = Field(default=os.getenv("OCR_BASE_URL"), descripti
on="OCR base URL")
 # VLLM API密鑰
 vllm_api_key: str = Field(default=os.getenv("VLLM_API_KEY"), descripti
on="VLLM API key")
 # VLLM基礎URL
 vllm_base_url: str = Field(default=os.getenv("VLLM_BASE_URL"), descrip
tion="VLLM base URL")
 # VLLM模型名稱
 vllm_model_name: str = Field(default=os.getenv("VLLM_MODEL_NAME"), des
cription="VLLM model name")
# 實例化RAGConfig對象---單例模式
RagConfig = RAGConfig()

import random
from typing import Optional, List
# 導⼊chainlit庫，⽤於構建和部署AI應⽤
import chainlit as cl
# 導⼊chainlit的數據模塊，⽤於處理數據相關的操作
import chainlit.data as cl_data
# 從chainlit.element模塊中導⼊ElementBased類，⽤於創建基於元素的組件
from chainlit.element import ElementBased
# 從chainlit.input_widget模塊中導⼊Select和Switch類，⽤於創建選擇和開關輸⼊組件
from chainlit.input_widget import Select, Switch
# 從chainlit.types模塊中導⼊ThreadDict類，⽤於定義線程字典類型
from chainlit.types import ThreadDict
# 導⼊dotenv庫，⽤於加載環境變量
from dotenv import load_dotenv
# 從llama_index.core.base.llms.types模塊中導⼊ChatMessage類，⽤於定義聊天消息
類型
from llama_index.core.base.llms.types import ChatMessage
# 從llama_index.core.chat_engine模塊中導⼊SimpleChatEngine類，⽤於實現簡單的聊
天引擎
from llama_index.core.chat_engine import SimpleChatEngine
# 從llama_index.core.chat_engine.types模塊中導⼊ChatMode類，⽤於定義聊天模式
from llama_index.core.chat_engine.types import ChatMode
# 從llama_index.core.memory模塊中導⼊ChatMemoryBuffer類，⽤於創建聊天記憶緩衝區
from llama_index.core.memory import ChatMemoryBuffer
# 導⼊⾃定義模塊以⽀持特定功能
from persistent.minio_storage_client import MinioStorageClient
from persistent.postgresql_data_layer import PostgreSQLDataLayer
from rag.base_rag import RAG
from rag.config import RagConfig
from rag.multimodal_rag import MultiModalRAG
from rag.traditional_rag import TraditionalRAG
from utils.milvus import list_collections
# 加載環境變量，⽤於配置應⽤程序
load_dotenv()
# 實例化MinIO存儲客户端，⽤於與MinIO對象存儲服務交互
storage_client = MinioStorageClient()
# 配置Chainlit數據層以使⽤PostgreSQL和MinIO進⾏數據持久化
cl_data._data_layer = PostgreSQLDataLayer(conninfo=RagConfig.pg_connectio
n_string, storage_provider=storage_client)
async def view_pdf(elements: List[ElementBased]):
 """查看PDF⽂件"""
 files = []
 contents = []
 for element in elements:
 if element.name.endswith(".pdf"):
 pdf = cl.Pdf(name=element.name, display="side", path=element.
path)
 files.append(pdf)
 contents.append(element.name)
 if len(files) == 0:
 return
 await cl.Message(content=f"查看PDF⽂件：" + "，".join(contents), elemen
ts=files).send()
@cl.set_chat_profiles
async def chat_profile(current_user: cl.User):
 if current_user.metadata["role"] != "admin":
 return None
 # 知識庫信息最後存儲在關係數據庫中:名稱，描述，圖標
 kb_list = list_collections()
 profiles = [
 cl.ChatProfile(
 name="default",
 markdown_description=f"⼤模型對話",
 icon=f"/public/kbs/4.png",
 )
 ]
 for kb_name in kb_list:
 profiles.append(
 cl.ChatProfile(
 name=kb_name,
 markdown_description=f"{kb_name} 知識庫",
 icon=f"/public/kbs/{random.randint(1, 3)}.jpg",
 )
 )
 return profiles