万物皆可问 — 私有部署网易有道QAnything

万物皆可问 — 私有部署网易有道QAnything_第1张图片

什么是 QAnything?

QAnything(Question and Answer based on Anything)是一个本地知识库问答系统,旨在支持多种文件格式和数据库,允许离线安装和使用。使用QAnything,您可以简单地删除本地存储的任何格式的文件,并获得准确、快速和可靠的答案。QAnything目前支持的知识库文件格式包括:PDF(pdf) ,  Word(docx) ,  PPT(pptx) ,  XLS(xlsx) ,  Markdown(md) ,  Email(eml) , TXT(txt) , Image(jpg,jpeg,png) , CSV (csv)网页链接(html)等。

主要功能

  • 数据安全,支持全程不插网线安装使用。
  • 跨语言QA支持,中英文QA自由切换,无论文档语言如何。
  • 支持海量数据QA,两阶段检索排序,解决大规模数据检索的退化问题;数据越多,性能越好。
  • 高性能生产级系统,可直接部署用于企业应用。
  • 人性化,无需繁琐配置,一键安装部署,即用即用。
  • 多知识库QA支持选择多个知识库进行问答

软件架构

万物皆可问 — 私有部署网易有道QAnything_第2张图片 软件架构

部署QAnything详细步骤

QAnything目前已经在Github开源,开源项目地址:GitHub - netease-youdao/QAnything: Question and Answer based on Anything.

安装QAnything的系统要求

万物皆可问 — 私有部署网易有道QAnything_第3张图片 系统要求

安装nVidia GPU driver

sudo apt-get update
Hit:1 http://us.archive.ubuntu.com/ubuntu jammy InRelease
Hit:2 http://packages.microsoft.com/repos/code stable InRelease
Hit:3 http://us.archive.ubuntu.com/ubuntu jammy-updates InRelease
Hit:4 http://security.ubuntu.com/ubuntu jammy-security InRelease
Hit:5 https://packages.microsoft.com/repos/vscode stable InRelease
Hit:6 http://us.archive.ubuntu.com/ubuntu jammy-backports InRelease
Reading package lists... Done

我的系统用了RTX-4090,在Ubuntu 22.04上推荐使用nvidia-driver-535版本。

ubuntu-drivers devices
== /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0 ==
modalias : pci:v000010DEd00002684sv00001043sd000088E2bc03sc00i00
vendor   : NVIDIA Corporation
driver   : nvidia-driver-535-server-open - distro non-free
driver   : nvidia-driver-535-open - distro non-free
driver   : nvidia-driver-525 - distro non-free
driver   : nvidia-driver-545 - distro non-free
driver   : nvidia-driver-535 - distro non-free recommended
driver   : nvidia-driver-545-open - distro non-free
driver   : nvidia-driver-535-server - distro non-free
driver   : nvidia-driver-525-open - distro non-free
driver   : nvidia-driver-525-server - distro non-free
driver   : xserver-xorg-video-nouveau - distro free builtin
sudo apt install nvidia-driver-535
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
The following additional packages will be installed:
  dctrl-tools dkms libatomic1:i386 libbsd0:i386 libdrm-amdgpu1:i386 libdrm-intel1:i386 libdrm-nouveau2:i386
  libdrm-radeon1:i386 libdrm2:i386 libedit2:i386 libelf1:i386 libexpat1:i386 libffi8:i386 libgl1:i386
  libgl1-mesa-dri:i386 libglapi-mesa:i386 libglvnd0:i386 libglx-mesa0:i386 libglx0:i386 libicu70:i386 libllvm15:i386
  libmd0:i386 libnvidia-cfg1-535 libnvidia-common-535 libnvidia-compute-535 libnvidia-compute-535:i386
  libnvidia-decode-535 libnvidia-decode-535:i386 libnvidia-encode-535 libnvidia-encode-535:i386 libnvidia-extra-535
  libnvidia-fbc1-535 libnvidia-fbc1-535:i386 libnvidia-gl-535 libnvidia-gl-535:i386 libpciaccess0:i386
  libsensors5:i386 libstdc++6:i386 libvdpau1 libx11-6:i386 libx11-xcb1:i386 libxau6:i386 libxcb-dri2-0:i386
  libxcb-dri3-0:i386 libxcb-glx0:i386 libxcb-present0:i386 libxcb-randr0:i386 libxcb-shm0:i386 libxcb-sync1:i386
  libxcb-xfixes0:i386 libxcb1:i386 libxdmcp6:i386 libxext6:i386 libxfixes3:i386 libxml2:i386 libxnvctrl0
  libxshmfence1:i386 libxxf86vm1:i386 mesa-vdpau-drivers nvidia-compute-utils-535 nvidia-dkms-535
  nvidia-firmware-535-535.154.05 nvidia-kernel-common-535 nvidia-kernel-source-535 nvidia-prime nvidia-settings
  nvidia-utils-535 pkg-config screen-resolution-extra vdpau-driver-all xserver-xorg-video-nvidia-535

安装完成后重启系统,然后用命令nvidia-smi验证驱动安装成功与否。

nvidia-smi

万物皆可问 — 私有部署网易有道QAnything_第4张图片

安装docker

# Add Docker's official GPG key:
sudo apt-get update
sudo apt-get install ca-certificates curl
sudo install -m 0755 -d /etc/apt/keyrings
sudo curl -fsSL https://download.docker.com/linux/ubuntu/gpg -o /etc/apt/keyrings/docker.asc
sudo chmod a+r /etc/apt/keyrings/docker.asc

# Add the repository to Apt sources:
echo \
  "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu \
  $(. /etc/os-release && echo "$VERSION_CODENAME") stable" | \
  sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
sudo apt-get update
sudo apt-get install docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose docker-compose-plugin

为了方便docker的管理,我还另外安装portainer,可以通过web界面管理docker。

首先为Portainer创建数据保存volume:

docker volume create portainer

下载并安装Portainer:

docker run -d \
-p 9000:9000 \
-p 9443:9443 \
--name portainer \
--restart=always \
-v /var/run/docker.sock:/var/run/docker.sock \
-v portainer:/data \
portainer/portainer-ce:latest

安装完成后,就可以访问http://your_server_name:9000或者https://your_server_name:9443来管理docker。

万物皆可问 — 私有部署网易有道QAnything_第5张图片

万物皆可问 — 私有部署网易有道QAnything_第6张图片

安装git-fls

sudo apt-get install git-lfs

部署QAnything

接下来正式开始部署QAnything。

下载QAnything

git clone https://github.com/netease-youdao/QAnything.git

安装并运行QAnything

sudo bash ./run.sh -c local -i 0 -b default

运行QAnything的主要脚本都在项目根目录下的run.sh文件里。该shell脚本的启动选项有如下多种方式,主要针对不同的系统配置和大语言模型。比如显卡显存较小的话,就推荐调用OpenAI API,推理在OpenAI的服务器发生。我的RTX-4090共24G现存,在这里悬着运行第二个选项,跑的是Qwen-7B-QAnything的语言模型。

万物皆可问 — 私有部署网易有道QAnything_第7张图片
第一步:检测环境

这部分的代码占了run.sh的绝大部分,主要功能是检测系统的基本配置,检测本地代码的版本,并推荐相应的网络模型和准备下载。主要脚本代码如下:

# 获取最新的远程仓库信息
git fetch origin master

# 获取本地master分支的最新提交
LOCAL=$(git rev-parse master)
# 获取远程master分支的最新提交
REMOTE=$(git rev-parse origin/master)

if [ $LOCAL != $REMOTE ]; then
    # 本地分支与远程分支不一致,需要更新
    print_important_notice
else
    echo -e "${GREEN}当前master分支已是最新,无需更新。${NC}"
fi


llm_api="local"
device_id="0"
runtime_backend="default"
model_name=""
conv_template=""
tensor_parallel=1
gpu_memory_utilization=0.81

# 解析命令行参数
while getopts ":c:i:b:m:t:p:r:h" opt; do
  case $opt in
    c) llm_api=$OPTARG ;;
    i) device_id=$OPTARG ;;
    b) runtime_backend=$OPTARG ;;
    m) model_name=$OPTARG ;;
    t) conv_template=$OPTARG ;;
    p) tensor_parallel=$OPTARG ;;
    r) gpu_memory_utilization=$OPTARG ;;
    h) usage ;;
    *) usage ;;
  esac
done

# 获取大模型B数
if [ $llm_api = 'cloud' ]; then
    model_size='0B'
elif [ $runtime_backend = 'default' ]; then
    model_size='7B'
else
    read -p "请输入您使用的大模型B数(示例:1.8B/3B/7B): " model_size
    # 检查是否合法,必须输入数字+B的形式,可以是小数
    if ! [[ $model_size =~ ^[0-9]+(\.[0-9]+)?B$ ]]; then
        echo "Invalid model size. Please enter a number like '1.8B' or '3B' or '7B'."
        exit 1
    fi
fi
echo "model_size=$model_size"
model_size_num=$(echo $model_size | grep -oP '^[0-9]+(\.[0-9]+)?')

gpu_id1=0
gpu_id2=0

# 判断命令行参数
if [[ -n "$device_id" ]]; then
    # 如果传入参数,分割成两个GPU ID
    IFS=',' read -ra gpu_ids <<< "$device_id"
    gpu_id1=${gpu_ids[0]}
    gpu_id2=${gpu_ids[1]:-$gpu_id1}  # 如果没有第二个ID,则默认使用第一个ID
fi

echo "GPUID1=${gpu_id1}, GPUID2=${gpu_id2}, device_id=${device_id}"

# 检查GPU ID是否合法
if ! [[ $gpu_id1 =~ ^[0-9]+$ ]] || ! [[ $gpu_id2 =~ ^[0-9]+$ ]]; then
    echo "Invalid GPU IDs. Please enter IDs like '0' or '0,1'."
    exit 1
fi

update_or_append_to_env "GPUID1" "$gpu_id1"
update_or_append_to_env "GPUID2" "$gpu_id2"

# 获取显卡型号
gpu_model=$(nvidia-smi --query-gpu=gpu_name --format=csv,noheader,nounits -i $gpu_id1)
# nvidia RTX 30系列或40系列
gpu_series=$(echo $gpu_model | grep -oP 'RTX\s*(30|40)')
if ! command -v jq &> /dev/null; then
    echo "Error: jq 命令不存在,请使用 sudo apt update && sudo apt-get install jq 安装,再重新启动。"
    exit 1
fi
compute_capability=$(jq -r ".[\"$gpu_model\"]" scripts/gpu_capabilities.json)
# 如果compute_capability为空,则说明显卡型号不在gpu_capabilities.json中
if [ -z "$compute_capability" ]; then
    echo "您的显卡型号 $gpu_model 不在支持列表中,请联系技术支持。"
    exit 1
fi
echo "GPU1 Model: $gpu_model"
echo "Compute Capability: $compute_capability"

if ! command -v bc &> /dev/null; then
    echo "Error: bc 命令不存在,请使用 sudo apt update && sudo apt-get install bc 安装,再重新启动。"
    exit 1
fi

if [ $(echo "$compute_capability >= 7.5" | bc) -eq 1 ]; then
    OCR_USE_GPU="True"
else
    OCR_USE_GPU="False"
fi
echo "OCR_USE_GPU=$OCR_USE_GPU because $compute_capability >= 7.5"
update_or_append_to_env "OCR_USE_GPU" "$OCR_USE_GPU"

# 使用nvidia-smi命令获取GPU的显存大小(以MiB为单位)
GPU1_MEMORY_SIZE=$(nvidia-smi --query-gpu=memory.total --format=csv,noheader,nounits -i $gpu_id1)

OFFCUT_TOKEN=0
echo "===================================================="
echo "******************** 重要提示 ********************"
echo "===================================================="
echo ""

# 使用默认后端且model_size_num不为0
if [ "$runtime_backend" = "default" ] && [ "$model_size_num" -ne 0 ]; then
    if [ -z "$gpu_series" ]; then  # 不是Nvidia 30系列或40系列
        echo "您的显卡型号 $gpu_model 部署默认后端FasterTransformer需要Nvidia RTX 30系列或40系列显卡,将自动为您切换后端:"
        # 如果显存大于等于24GB且计算力大于等于8.6,则可以使用vllm后端
        if [ "$GPU1_MEMORY_SIZE" -ge 24000 ] && [ $(echo "$compute_capability >= 8.6" | bc) -eq 1 ]; then
            echo "根据匹配算法,已自动为您切换为vllm后端(推荐)"
            runtime_backend="vllm"
        else
            # 自动切换huggingface后端
            echo "根据匹配算法,已自动为您切换为huggingface后端"
            runtime_backend="hf"
        fi
    fi
fi

if [ "$GPU1_MEMORY_SIZE" -lt 4000 ]; then # 显存小于4GB
    echo "您当前的显存为 $GPU1_MEMORY_SIZE MiB 不足以部署本项目,建议升级到GTX 1050Ti或以上级别的显卡"
    exit 1
elif [ "$model_size_num" -eq 0 ]; then  # 模型大小为0B, 表示使用openai api,4G显存就够了
    echo "您当前的显存为 $GPU1_MEMORY_SIZE MiB 可以使用在线的OpenAI API"
elif [ "$GPU1_MEMORY_SIZE" -lt 8000 ]; then  # 显存小于8GB
    # 显存小于8GB,仅推荐使用在线的OpenAI API
    echo "您当前的显存为 $GPU1_MEMORY_SIZE MiB 仅推荐使用在线的OpenAI API"
    if [ "$model_size_num" -gt 0 ]; then  # 模型大小大于0B
        echo "您的显存不足以部署 $model_size 模型,请重新选择模型大小"
        exit 1
    fi
elif [ "$GPU1_MEMORY_SIZE" -ge 8000 ] && [ "$GPU1_MEMORY_SIZE" -le 10000 ]; then  # 显存[8GB-10GB)
    # 8GB显存,推荐部署1.8B的大模型
    echo "您当前的显存为 $GPU1_MEMORY_SIZE MiB 推荐部署1.8B的大模型,包括在线的OpenAI API"
    if [ "$model_size_num" -gt 2 ]; then  # 模型大小大于2B
        echo "您的显存不足以部署 $model_size 模型,请重新选择模型大小"
        exit 1
    fi
elif [ "$GPU1_MEMORY_SIZE" -ge 10000 ] && [ "$GPU1_MEMORY_SIZE" -le 16000 ]; then  # 显存[10GB-16GB)
    # 10GB, 11GB, 12GB显存,推荐部署3B及3B以下的模型
    echo "您当前的显存为 $GPU1_MEMORY_SIZE MiB,推荐部署3B及3B以下的模型,包括在线的OpenAI API"
    if [ "$model_size_num" -gt 3 ]; then  # 模型大小大于3B
        echo "您的显存不足以部署 $model_size 模型,请重新选择模型大小"
        exit 1
    fi
elif [ "$GPU1_MEMORY_SIZE" -ge 16000 ] && [ "$GPU1_MEMORY_SIZE" -le 22000 ]; then  # 显存[16-22GB)
    # 16GB显存
    echo "您当前的显存为 $GPU1_MEMORY_SIZE MiB 推荐部署小于等于7B的大模型"
    if [ "$model_size_num" -gt 7 ]; then  # 模型大小大于7B
        echo "您的显存不足以部署 $model_size 模型,请重新选择模型大小"
        exit 1
    fi
    if [ "$runtime_backend" = "default" ]; then  # 默认使用Qwen-7B-QAnything+FasterTransformer
        if [ -n "$gpu_series" ]; then
            # Nvidia 30系列或40系列
            if [ $gpu_id1 -eq $gpu_id2 ]; then
                echo "为了防止显存溢出,tokens上限默认设置为2700"
                OFFCUT_TOKEN=1400
            else
                echo "tokens上限默认设置为4096"
                OFFCUT_TOKEN=0
            fi
        else
            echo "您的显卡型号 $gpu_model 不支持部署Qwen-7B-QAnything模型"
            exit 1
        fi
    elif [ "$runtime_backend" = "hf" ]; then  # 使用Huggingface Transformers后端
        if [ "$model_size_num" -le 7 ] && [ "$model_size_num" -gt 3 ]; then  # 模型大小大于3B,小于等于7B
            if [ $gpu_id1 -eq $gpu_id2 ]; then
                echo "为了防止显存溢出,tokens上限默认设置为1400"
                OFFCUT_TOKEN=2700
            else
                echo "为了防止显存溢出,tokens上限默认设置为2300"
                OFFCUT_TOKEN=1800
            fi
        else
            echo "tokens上限默认设置为4096"
            OFFCUT_TOKEN=0
        fi
    elif [ "$runtime_backend" = "vllm" ]; then  # 使用VLLM后端
        if [ "$model_size_num" -gt 3 ]; then  # 模型大小大于3B
            echo "您的显存不足以使用vllm后端部署 $model_size 模型"
            exit 1
        else
            echo "tokens上限默认设置为4096"
            OFFCUT_TOKEN=0
        fi
    fi
elif [ "$GPU1_MEMORY_SIZE" -ge 22000 ] && [ "$GPU1_MEMORY_SIZE" -le 25000 ]; then  # [22GB, 24GB]
    echo "您当前的显存为 $GPU1_MEMORY_SIZE MiB 推荐部署7B模型"
    if [ "$model_size_num" -gt 7 ]; then  # 模型大小大于7B
        echo "您的显存不足以部署 $model_size 模型,请重新选择模型大小"
        exit 1
    fi
    OFFCUT_TOKEN=0
elif [ "$GPU1_MEMORY_SIZE" -gt 25000 ]; then  # 显存大于24GB
    OFFCUT_TOKEN=0
fi

update_or_append_to_env "OFFCUT_TOKEN" "$OFFCUT_TOKEN"

if [ $llm_api = 'cloud' ]; then
  need_input_openai_info=1
  OPENAI_API_KEY=$(grep OPENAI_API_KEY .env | cut -d '=' -f2)
  # 如果.env中已存在OPENAI_API_KEY的值(不为空),则询问用户是否使用上次默认值:$OPENAI_API_KEY,$OPENAI_API_BASE, $OPENAI_API_MODEL_NAME, $OPENAI_API_CONTEXT_LENGTH
  if [ -n "$OPENAI_API_KEY" ]; then
    read -p "Do you want to use the previous OPENAI_API_KEY: $OPENAI_API_KEY? (yes/no) 是否使用上次的OPENAI_API_KEY: $OPENAI_API_KEY?(yes/no) 回车默认选yes,请输入:" use_previous
    use_previous=${use_previous:-yes}
    if [ "$use_previous" = "yes" ]; then
      need_input_openai_info=0
    fi
  fi
  if [ $need_input_openai_info -eq 1 ]; then
    read -p "Please enter OPENAI_API_KEY: " OPENAI_API_KEY
    read -p "Please enter OPENAI_API_BASE (default: https://api.openai.com/v1):" OPENAI_API_BASE
    read -p "Please enter OPENAI_API_MODEL_NAME (default: gpt-3.5-turbo):" OPENAI_API_MODEL_NAME
    read -p "Please enter OPENAI_API_CONTEXT_LENGTH (default: 4096):" OPENAI_API_CONTEXT_LENGTH

    if [ -z "$OPENAI_API_KEY" ]; then  # 如果OPENAI_API_KEY为空,则退出
    echo "OPENAI_API_KEY is empty, please enter OPENAI_API_KEY."
    exit 1
    fi
    if [ -z "$OPENAI_API_BASE" ]; then  # 如果OPENAI_API_BASE为空,则设置默认值
      OPENAI_API_BASE="https://api.openai.com/v1"
    fi
    if [ -z "$OPENAI_API_MODEL_NAME" ]; then  # 如果OPENAI_API_MODEL_NAME为空,则设置默认值
      OPENAI_API_MODEL_NAME="gpt-3.5-turbo"
    fi
    if [ -z "$OPENAI_API_CONTEXT_LENGTH" ]; then  # 如果OPENAI_API_CONTEXT_LENGTH为空,则设置默认值
      OPENAI_API_CONTEXT_LENGTH=4096
    fi

    update_or_append_to_env "OPENAI_API_KEY" "$OPENAI_API_KEY"
    update_or_append_to_env "OPENAI_API_BASE" "$OPENAI_API_BASE"
    update_or_append_to_env "OPENAI_API_MODEL_NAME" "$OPENAI_API_MODEL_NAME"
    update_or_append_to_env "OPENAI_API_CONTEXT_LENGTH" "$OPENAI_API_CONTEXT_LENGTH"
  else
    OPENAI_API_BASE=$(grep OPENAI_API_BASE .env | cut -d '=' -f2)
    OPENAI_API_MODEL_NAME=$(grep OPENAI_API_MODEL_NAME .env | cut -d '=' -f2)
    OPENAI_API_CONTEXT_LENGTH=$(grep OPENAI_API_CONTEXT_LENGTH .env | cut -d '=' -f2)
    echo "使用上次的配置:"
    echo "OPENAI_API_KEY: $OPENAI_API_KEY"
    echo "OPENAI_API_BASE: $OPENAI_API_BASE"
    echo "OPENAI_API_MODEL_NAME: $OPENAI_API_MODEL_NAME"
    echo "OPENAI_API_CONTEXT_LENGTH: $OPENAI_API_CONTEXT_LENGTH"
  fi
fi

echo "llm_api is set to [$llm_api]"
echo "device_id is set to [$device_id]"
echo "runtime_backend is set to [$runtime_backend]"
echo "model_name is set to [$model_name]"
echo "conv_template is set to [$conv_template]"
echo "tensor_parallel is set to [$tensor_parallel]"
echo "gpu_memory_utilization is set to [$gpu_memory_utilization]"

update_or_append_to_env "LLM_API" "$llm_api"
update_or_append_to_env "DEVICE_ID" "$device_id"
update_or_append_to_env "RUNTIME_BACKEND" "$runtime_backend"
update_or_append_to_env "MODEL_NAME" "$model_name"
update_or_append_to_env "CONV_TEMPLATE" "$conv_template"
update_or_append_to_env "TP" "$tensor_parallel"
update_or_append_to_env "GPU_MEM_UTILI" "$gpu_memory_utilization"

# 检查是否存在 models 文件夹,且models下是否存在embed,rerank,base三个文件夹
if [ ! -d "models" ] || [ ! -d "models/embed" ] || [ ! -d "models/rerank" ] || [ ! -d "models/base" ]; then
  echo "models文件夹不完整 开始克隆和解压模型..."
  echo "===================================================="
  echo "******************** 重要提示 ********************"
  echo "===================================================="
  echo ""
  echo "模型大小为8G左右,下载+解压时间可能较长,请耐心等待10分钟,"
  echo "仅首次启动需下载模型。"
  echo "The model size is about 8GB, the download and decompression time may be long, "
  echo "please wait patiently for 10 minutes."
  echo "Only the model needs to be downloaded for the first time."
  echo ""
  echo "===================================================="
  echo "如果你在下载过程中遇到任何问题,请及时联系技术支持。"
  echo "===================================================="
  # 记录下载和解压的时间
万物皆可问 — 私有部署网易有道QAnything_第8张图片 检测环境

第二步:下载模型

模型下载的主要脚本代码如下。如果已经下载过网络模型,则不会再次下载。

  # 如果存在QAanything/models.zip,不用下载
  if [ ! -f "QAnything/models.zip" ]; then
    echo "Downloading models.zip..."
    echo "开始下载模型文件..."
    git lfs install
    git clone https://www.modelscope.cn/netease-youdao/QAnything.git
    d_end_time=$(date +%s)
    elapsed=$((d_end_time - d_start_time))  # 计算经过的时间(秒)
    echo "Download Time elapsed: ${elapsed} seconds."
    echo "下载耗时: ${elapsed} 秒."
  else
    echo "models.zip already exists, no need to download."
    echo "models.zip已存在,无需下载。"
  fi

  # 解压模型文件
  # 判断是否存在unzip,不存在建议使用sudo apt-get install unzip安装
  if ! command -v unzip &> /dev/null; then
    echo "Error: unzip 命令不存在,请使用 sudo apt update && sudo apt-get install unzip 安装,再重新启动"
    exit 1
  fi

  unzip_start_time=$(date +%s)
  unzip QAnything/models.zip

  unzip_end_time=$(date +%s)
  elapsed=$((unzip_end_time - unzip_start_time))  # 计算经过的时间(秒)
  echo "unzip Time elapsed: ${elapsed} seconds."
  echo "解压耗时: ${elapsed} 秒."

  # 删除克隆的仓库
  # rm -rf QAnything
else
  echo "models 文件夹已存在,无需下载。"
fi

check_version_file() {
  local version_file="models/version.txt"
  local expected_version="$1"

  # 检查 version.txt 文件是否存在
  if [ ! -f "$version_file" ]; then
    echo "QAnything/models/ 不存在version.txt 请检查您的模型文件是否完整。"
    exit 1
  fi

  # 读取 version.txt 文件中的版本号
  local version_in_file=$(cat "$version_file")

  # 检查版本号是否为 v2.1.0
  if [ "$version_in_file" != "$expected_version" ]; then
    echo "当前版本为 $version_in_file ,不是期望的 $expected_version 版本。请更新您的模型文件。"
    exit 1
  fi

  echo "检查模型版本成功,当前版本为 $expected_version。"
}

check_version_file "v2.1.0"
echo "Model directories check passed. (0/8)"
echo "模型路径和模型版本检查通过. (0/8)"

万物皆可问 — 私有部署网易有道QAnything_第9张图片 下载模型

下载和部署QAnything docker

# 检查是否存在用户文件
if [[ -f "$user_file" ]]; then
    # 读取上次的配置
    host=$(cat "$user_file")
    read -p "Do you want to use the previous host: $host? (yes/no) 是否使用上次的host: $host?(yes/no) 回车默认选yes,请输入:" use_previous
    use_previous=${use_previous:-yes}
    if [[ $use_previous != "yes" && $use_previous != "是" ]]; then
        read -p "Are you running the code on a remote server or on your local machine? (remote/local) 您是在远程服务器上还是本地机器上启动代码?(remote/local) " answer
        if [[ $answer == "local" || $answer == "本地" ]]; then
            host="localhost"
        else
            read -p "Please enter the server IP address 请输入服务器公网IP地址(示例:10.234.10.144): " host
            echo "当前设置的远程服务器IP地址为 $host, QAnything启动后,本地前端服务(浏览器打开[http://$host:5052/qanything/])将远程访问[http://$host:8777]上的后端服务,请知悉!"
            sleep 5
        fi
        # 保存新的配置到用户文件
        echo "$host" > "$user_file"
    fi
else
    # 如果用户文件不存在,询问用户并保存配置
    read -p "Are you running the code on a remote server or on your local machine? (remotelocal) 您是在云服务器上还是本地机器上启动代码?(remote/local) " answer
    if [[ $answer == "local" || $answer == "本地" ]]; then
        host="localhost"
    else
        read -p "Please enter the server IP address 请输入服务器公网IP地址(示例:10.234.10.144): " host
        echo "当前设置的远程服务器IP地址为 $host, QAnything启动后,本地前端服务(浏览器打开[http://$host:5052/qanything/])将远程访问[http://$host:8777]上的后端服务,请知悉!"
        sleep 5
    fi
    # 保存配置到用户文件
    echo "$host" > "$user_file"
fi

if [ -e /proc/version ]; then
  if grep -qi microsoft /proc/version || grep -qi MINGW /proc/version; then
    if grep -qi microsoft /proc/version; then
        echo "Running under WSL"
    else
        echo "Running under git bash"
    fi
    
    if docker-compose -p user -f docker-compose-windows.yaml down |& tee /dev/tty | grep -q "services.qanything_local.deploy.resources.reservations value 'devices' does not match any of the regexes"; then
        echo "检测到 Docker Compose 版本过低,请升级到v2.23.3或更高版本。执行docker-compose -v查看版本。"
    fi
    docker-compose -p user -f docker-compose-windows.yaml up -d
    docker-compose -p user -f docker-compose-windows.yaml logs -f qanything_local
  else
    echo "Running under native Linux"
    if docker-compose -p user -f docker-compose-linux.yaml down |& tee /dev/tty | grep -q "services.qanything_local.deploy.resources.reservations value 'devices' does not match any of the regexes"; then
        echo "检测到 Docker Compose 版本过低,请升级到v2.23.3或更高版本。执行docker-compose -v查看版本。"
    fi
    docker-compose -p user -f docker-compose-linux.yaml up -d
    docker-compose -p user -f docker-compose-linux.yaml logs -f qanything_local
    # 检查日志输出
  fi
else
  echo "/proc/version 文件不存在。请确认自己位于Linux或Windows的WSL环境下"
fi

万物皆可问 — 私有部署网易有道QAnything_第10张图片

下载并运行docker

当看到最后这条打印消息的时候,整个的部署工作就已经成功了。这时候通过Portainer可以看到总共部署了5个docker container:milvus-etcd-local,milvus-minio-local,milvus-standalone-local,mysql-container-local和qanything-container-local。

万物皆可问 — 私有部署网易有道QAnything_第11张图片

这时候访问上述红框里的URL就可以使用QAnything了。下面的视频是我部署的QAnything的使用demo。我给它输入了一个Tesla Model Y的用户手册,然后就可以向它问各种Model Y的使用问题啦。

如果想体检QAnything的私有化部署和它的各种功能,可以访问我私有化部署的服务器。服务器的地址如下,可以访问我的博文原文:万物皆可问 — 私有部署网易有道QAnything - HY's Blog

作者个人Blog(HY's Blog):https://blog.yanghong.dev

你可能感兴趣的:(技术分享,人工智能,llama)