├── LICENSE ├── README.md ├── faiss ├── camp │ ├── index.faiss │ └── index.pkl └── logistics │ ├── index.faiss │ └── index.pkl ├── get_vector.py ├── main.py ├── model.py └── 物流信息.txt /LICENSE: -------------------------------------------------------------------------------- 1 | MIT License 2 | 3 | Copyright (c) 2025 Haohao-end 4 | 5 | Permission is hereby granted, free of charge, to any person obtaining a copy 6 | of this software and associated documentation files (the "Software"), to deal 7 | in the Software without restriction, including without limitation the rights 8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 9 | copies of the Software, and to permit persons to whom the Software is 10 | furnished to do so, subject to the following conditions: 11 | 12 | The above copyright notice and this permission notice shall be included in all 13 | copies or substantial portions of the Software. 14 | 15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 21 | SOFTWARE. 22 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # 物流知识智能问答系统 - LogisticsQA 2 | 3 | **让物流信息查询像聊天一样简单！** 4 | 基于大语言模型 + 向量检索的智能问答解决方案，快速定位物流信息中的关键内容 � 5 | 6 | --- 7 | 8 | ## 项目特色 9 | 10 | - ** 极速响应**：FAISS 向量库实现毫秒级信息检索 11 | - ** 语义理解**：M3E 中文嵌入模型精准捕捉上下文 12 | - ** 智能对话**：ChatGLM2-6B 大模型生成自然流畅回答 13 | - ** 开箱即用**：模块化设计，五分钟快速部署 14 | 15 | --- 16 | 17 | ## 功能演示 18 | 19 | 20 | 21 | **用户提问**： 22 | `我买的商品来自于哪个仓库，从哪出发的，预计什么时候到达？` 23 | 24 | **系统回答**： 25 | 您的包裹正在飞奔而来！ 26 | 出发地：广州仓库 27 | 启程时间：2023年1月20日 28 | 预计到达：1月23日（3天运输时长） 29 | 目的地：重庆 30 | 运输方式：陆运 31 | 32 | --- 33 | 34 | ## 技术栈 35 | 36 | ```bash 37 | - LangChain 38 | - FAISS 39 | - HuggingFace Transformers 40 | - ChatGLM2-6B 41 | - M3E 中文嵌入模型 🇨🇳 42 | ``` 43 | 44 | --- 45 | 46 | ## 快速开始 47 | 48 | ### 1. 安装依赖 49 | ```bash 50 | pip install -r requirements.txt 51 | pip install langchain-community unstructured faiss-cpu 52 | # 处理文档需要 53 | pip install python-magic python-magic-bin 54 | ``` 55 | 56 | ### 2. 运行准备 57 | ```bash 58 | # 下载模型文件 59 | git lfs install 60 | git clone https://huggingface.co/THUDM/chatglm2-6b ./model_files/ 61 | 62 | # 准备物流文档 63 | echo "广州仓库 发货时间 2023-01-20..." > 物流信息.txt 64 | ``` 65 | 66 | ### 3. 启动系统 67 | ```python 68 | # 生成向量库 69 | python get_vector.py 70 | 71 | # 运行问答系统 72 | python Knowledge_QA/main.py 73 | ``` 74 | 75 | --- 76 | 77 | ## 核心逻辑 78 | 79 | ```mermaid 80 | graph TD 81 | A[用户提问] --> B(语义向量编码) 82 | B --> C[FAISS 向量库检索] 83 | C --> D{Top1 匹配结果} 84 | D --> E[构建Prompt模板] 85 | E --> F[ChatGLM2生成回答] 86 | F --> G[返回结构化结果] 87 | ``` 88 | 89 | --- 90 | 91 | ## 使用示例 92 | 93 | ```python 94 | from Knowledge_QA.main import qa 95 | 96 | # 像朋友一样提问！ 97 | response = qa("我的包裹现在到哪了？") 98 | print(f" 物流小助手：{response}") 99 | ``` 100 | 101 | --- 102 | 103 | ## 重要提示 104 | 105 | 1. **模型路径配置** 106 | 修改 `ChatGLM2.load_model()` 为本地模型路径： 107 | `llm.load_model("/path/to/chatglm2-6b")` 108 | 109 | 2. **文本分割优化** 110 | 根据文档特性调整 `chunk_size`： 111 | ```python 112 | # 推荐值：100-500字符 113 | TextSplitter(chunk_size=200) 114 | ``` 115 | 116 | 3. **性能优化建议** 117 | ```python 118 | # 在模型加载时添加参数 119 | llm.load_model(..., device_map="auto", torch_dtype=torch.float16) 120 | ``` 121 | 122 | --- 123 | 124 | ## 开源协议 125 | 126 | 本项目采用 [MIT License](LICENSE)，拥抱开源，共同进步！ 127 | 128 | --- 129 | 130 | **让物流信息流动起来** 131 | 有任何建议欢迎提交 Issue 或 Star 支持我们！ 132 | -------------------------------------------------------------------------------- /faiss/camp/index.faiss: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haohao-end/LangChain-ChatGLM-6B-RAG/ec7e35c94140951dba12418eaf8fa055e9f7d263/faiss/camp/index.faiss -------------------------------------------------------------------------------- /faiss/camp/index.pkl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haohao-end/LangChain-ChatGLM-6B-RAG/ec7e35c94140951dba12418eaf8fa055e9f7d263/faiss/camp/index.pkl -------------------------------------------------------------------------------- /faiss/logistics/index.faiss: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haohao-end/LangChain-ChatGLM-6B-RAG/ec7e35c94140951dba12418eaf8fa055e9f7d263/faiss/logistics/index.faiss -------------------------------------------------------------------------------- /faiss/logistics/index.pkl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Haohao-end/LangChain-ChatGLM-6B-RAG/ec7e35c94140951dba12418eaf8fa055e9f7d263/faiss/logistics/index.pkl -------------------------------------------------------------------------------- /get_vector.py: -------------------------------------------------------------------------------- 1 | from langchain_community.document_loaders import UnstructuredFileLoader 2 | from langchain.text_splitter import RecursiveCharacterTextSplitter 3 | from langchain_community.embeddings import HuggingFaceEmbeddings 4 | from langchain_community.vectorstores import FAISS # 向量数据库 5 | # from langchain.document_loaders import UnstructuredFileLoader 6 | # from langchain.text_splitter import RecursiveCharacterTextSplitter 7 | # from langchain.embeddings import HuggingFaceEmbeddings 8 | # from langchain.vectorstores import FAISS # 向量数据库 9 | 10 | def main(): 11 | # 定义向量模型路径 12 | EMBEDDING_MODEL = './m3e-base' 13 | 14 | # 第一步：加载文档： 15 | loader = UnstructuredFileLoader('物流信息.txt') 16 | data = loader.load() 17 | # print(f'data-->{data}') 18 | # 第二步：切分文档： 19 | text_split = RecursiveCharacterTextSplitter(chunk_size=128, 20 | chunk_overlap=4) 21 | split_data = text_split.split_documents(data) 22 | # print(f'split_data-->{split_data}') 23 | 24 | # 第三步：初始化huggingface模型embedding 25 | embeddings = HuggingFaceEmbeddings(model_name=EMBEDDING_MODEL) 26 | 27 | # 第四步：将切分后的文档进行向量化，并且存储下来 28 | db = FAISS.from_documents(split_data, embeddings) 29 | db.save_local('./faiss/camp') 30 | 31 | return split_data 32 | 33 | 34 | if __name__ == '__main__': 35 | split_data = main() 36 | print(f'split_data-->{split_data}') 37 | 38 | -------------------------------------------------------------------------------- /main.py: -------------------------------------------------------------------------------- 1 | # coding:utf-8 2 | # 导入必备的工具包 3 | from langchain.prompts import PromptTemplate 4 | from get_vector import * 5 | from model import ChatGLM2 6 | # 加载FAISS向量库 7 | EMBEDDING_MODEL = './m3e-base' 8 | embeddings = HuggingFaceEmbeddings(model_name=EMBEDDING_MODEL) 9 | db = FAISS.load_local('./faiss/camp', embeddings) 10 | 11 | 12 | def get_related_content(related_docs): 13 | related_content = [] 14 | for doc in related_docs: 15 | related_content.append(doc.page_content.replace('\n\n', '\n')) 16 | return '\n'.join(related_content) 17 | 18 | def define_prompt(): 19 | question = '我买的商品来自于哪个仓库，从哪出发的，预计什么到达' 20 | docs = db.similarity_search(question, k=1) 21 | # print(f'docs-->{docs}') 22 | related_docs = get_related_content(docs) 23 | 24 | # 构建模板 25 | PROMPT_TEMPLATE = """ 26 | 基于以下已知信息，简洁和专业的来回答用户的问题。不允许在答案中添加编造成分。 27 | 已知内容: 28 | {context} 29 | 问题: 30 | {question}""" 31 | prompt = PromptTemplate(input_variables=["context", "question"], 32 | template=PROMPT_TEMPLATE) 33 | 34 | my_prompt = prompt.format(context=related_docs, 35 | question=question) 36 | return my_prompt 37 | 38 | def qa(): 39 | llm = ChatGLM2() 40 | llm.load_model('/Users/ligang/PycharmProjects/llm/ChatGLM-6B/THUDM/chatglm-6b') 41 | my_prompt = define_prompt() 42 | result = llm(my_prompt) 43 | return result 44 | 45 | if __name__ == '__main__': 46 | result = qa() 47 | print(f'result-->{result}') 48 | -------------------------------------------------------------------------------- /model.py: -------------------------------------------------------------------------------- 1 | from langchain.llms.base import LLM 2 | from langchain.llms.utils import enforce_stop_tokens 3 | from transformers import AutoTokenizer, AutoModel 4 | from typing import List, Optional, Any 5 | 6 | 7 | # 自定义GLM类 8 | class ChatGLM2(LLM): 9 | max_token: int = 4096 10 | temperature: float = 0.8 11 | top_p = 0.9 12 | tokenizer: object = None 13 | model: object = None 14 | history = [] 15 | 16 | def __init__(self): 17 | super().__init__() 18 | 19 | @property 20 | def _llm_type(self) -> str: 21 | return "custom_chatglm2" 22 | 23 | # 定义load_model的方法 24 | def load_model(self, model_path=None): 25 | # 加载分词器 26 | self.tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True) 27 | # 加载模型 28 | self.model = AutoModel.from_pretrained(model_path, trust_remote_code=True).float() 29 | 30 | 31 | # 定义_call方法：进行模型的推理 32 | def _call(self,prompt: str, stop: Optional[List[str]] = None) -> str: 33 | response, _ = self.model.chat(self.tokenizer, 34 | prompt, 35 | history=self.history, 36 | temperature=self.temperature, 37 | top_p=self.top_p) 38 | 39 | if stop is not None: 40 | response = enforce_stop_tokens(response, stop) 41 | 42 | self.history = self.history + [[None, response]] 43 | return response 44 | 45 | if __name__ == '__main__': 46 | llm = ChatGLM2() 47 | llm.load_model(model_path='/Users/ligang/PycharmProjects/llm/ChatGLM-6B/THUDM/chatglm-6b-int4') 48 | print(f'llm--->{llm}') 49 | print(llm("1+1等于几？")) 50 | 51 | 52 | -------------------------------------------------------------------------------- /物流信息.txt: -------------------------------------------------------------------------------- 1 | 物流公司：速达物流 2 | 公司总部：北京市 3 | 业务范围：国际快递、仓储管理 4 | 5 | 货物追踪： 6 | - 货物编号：ABC123456 7 | - 发货日期：2023-01-15 8 | - 当前位置：上海分拨中心 9 | - 预计到达日期：2023-01-20 10 | 11 | 运输方式： 12 | - 运输公司：快运通 13 | - 运输方式：陆运 14 | - 出发地：广州 15 | - 目的地：重庆 16 | - 预计运输时间：3天 17 | 18 | 仓储信息： 19 | - 仓库名称：东方仓储中心 20 | - 仓库位置：深圳市 21 | - 存储货物类型：电子产品 22 | - 存储条件：常温仓储 23 | - 当前库存量：1000件 --------------------------------------------------------------------------------