langchain-ChatGLM是一个基于本地知识库的LLM对话库。其基于text2vec-large-Chinese为Embedding模型，ChatGLM-6B为对话大模型。原项目地址：https://github.com/chatchat-space/langchain-ChatGLM

对于如何本地部署ChatGLM模型，可以参考我之前的文章http://t.csdn.cn/16STJ

在本项目中，我们编写了langchai-ChatGLM API调用的客户端代码。经过测试虽然客户端可以正常调用服务器的API，但是对于删除知识库的指令服务器无法正常执行

1 langchain-ChatGLM API服务器端程序
下面程序段为langchain-ChatGLM项目中的api.py文件

import argparse
import json
import os
import shutil
from typing import List, Optional
import urllib

import nltk
import pydantic
import uvicorn
from fastapi import Body, FastAPI, File, Form, Query, UploadFile, WebSocket
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
from typing_extensions import Annotated
from starlette.responses import RedirectResponse

from chains.local_doc_qa import LocalDocQA
from configs.model_config import (KB_ROOT_PATH, EMBEDDING_DEVICE,
                                  EMBEDDING_MODEL, NLTK_DATA_PATH,
                                  VECTOR_SEARCH_TOP_K, LLM_HISTORY_LEN, OPEN_CROSS_DOMAIN)
import models.shared as shared
from models.loader.args import parser
from models.loader import LoaderCheckPoint

nltk.data.path = [NLTK_DATA_PATH] + nltk.data.path


class BaseResponse(BaseModel):
    code: int = pydantic.Field(200, description="HTTP status code")
    msg: str = pydantic.Field("success", description="HTTP status message")

    class Config:
        schema_extra = {
   
            "example": {
   
                "code": 200,
                "msg": "success",
            }
        }


class ListDocsResponse(BaseResponse):
    data: List[str] = pydantic.Field(..., description="List of document names")

    class Config:
        schema_extra = {
   
            "example": {
   
                "code": 200,
                "msg": "success",
                "data": ["doc1.docx", "doc2.pdf", "doc3.txt"],
            }
        }


class ChatMessage(BaseModel):
    question: str = pydantic.Field(..., description="Question text")
    response: str = pydantic.Field(..., description="Response text")
    history: List[List[str]] = pydantic.Field(..., description="History text")
    source_documents: List[str] = pydantic.Field(
        ..., description="List of source documents and their scores"
    )

    class Config:
        schema_extra = {
   
            "example": {
   
                "question": "工伤保险如何办理？",
                "response": "根据已知信息，可以总结如下：\n\n1. 参保单位为员工缴纳工伤保险费，以保障员工在发生工伤时能够获得相应的待遇。\n2. 不同地区的工伤保险缴费规定可能有所不同，需要向当地社保部门咨询以了解具体的缴费标准和规定。\n3. 工伤从业人员及其近亲属需要申请工伤认定，确认享受的待遇资格，并按时缴纳工伤保险费。\n4. 工伤保险待遇包括工伤医疗、康复、辅助器具配置费用、伤残待遇、工亡待遇、一次性工亡补助金等。\n5. 工伤保险待遇领取资格认证包括长期待遇领取人员认证和一次性待遇领取人员认证。\n6. 工伤保险基金支付的待遇项目包括工伤医疗待遇、康复待遇、辅助器具配置费用、一次性工亡补助金、丧葬补助金等。",
                "history": [
                    [
                        "工伤保险是什么？",
                        "工伤保险是指用人单位按照国家规定，为本单位的职工和用人单位的其他人员，缴纳工伤保险费，由保险机构按照国家规定的标准，给予工伤保险待遇的社会保险制度。",
                    ]
                ],
                "source_documents": [
                    "出处 [1] 广州市单位从业的特定人员参加工伤保险办事指引.docx：\n\n\t( 一)  从业单位  (组织)  按“自愿参保”原则，  为未建 立劳动关系的特定从业人员单项参加工伤保险 、缴纳工伤保 险费。",
                    "出处 [2] ...",
                    "出处 [3] ...",
                ],
            }
        }


def get_folder_path(local_doc_id: str):
    return os.path.join(KB_ROOT_PATH, local_doc_id, "content")


def get_vs_path(local_doc_id: str):
    return os.path.join(KB_ROOT_PATH, local_doc_id, "vector_store")


def get_file_path(local_doc_id: str, doc_name: str):
    return os.path.join(KB_ROOT_PATH, local_doc_id, "content", doc_name)


async def upload_file(
        file: UploadFile = File(description="A single binary file"),
        knowledge_base_id: str = Form(..., description="Knowledge Base Name", example="kb1"),
):
    saved_path = get_folder_path(knowledge_base_id)
    if not os.path.exists(saved_path):
        os.makedirs(saved_path)

    file_content = await file.read()  # 读取上传文件的内容

    file_path = os.path.join(saved_path, file.filename)
    if os.path.exists(file_path) and os.path.getsize(file_path) == len(file_content):
        file_status = f"文件 {
     file.filename} 已存在。"
        return BaseResponse(code=200, msg=file_status)

    with open(file_path, "wb") as f:
        f.write(file_content)

    vs_path = get_vs_path(knowledge_base_id)
    vs_path, loaded_files = local_doc_qa.init_knowledge_vector_store([file_path], vs_path)
    if len(loaded_files) > 0:
        file_status = f"文件 {
     file.filename} 已上传至新的知识库，并已加载知识库，请开始提问。"
        return BaseResponse(code=200, msg=file_status)
    else:
        file_status = "文件上传失败，请重新上传"
        return BaseResponse(code=500, msg=file_status)


async def upload_files(
        files: Annotated[
            List[UploadFile], File(description="Multiple files as UploadFile")
        ],
        knowledge_base_id: str = Form(..., description="Knowledge Base Name", example="kb1"),
):
    saved_path = get_folder_path(knowledge_base_id)
    if not os.path.exists(saved_path):
        os.makedirs(saved_path)
    filelist = []
    for file in files:
        file_content = ''
        file_path = os.path.join(saved_path, file.filename)
        file_content = file.file.read()
        if os.path.exists(file_path) and os.path.getsize(file_path) == len(file_content):
            continue
        with open(file_path, "ab+") as f:
            f.write(file_content)
        filelist.append(file_path)
    if filelist:
        vs_path, loaded_files