项目需要大量建模工程，为节约成本，结合网上教程与 Codex 方案整理整套部署流程，步骤较多特此归档。

基础软件清单

• 建模软件：Blender 4.2
• Python 环境：Python 3.12（Conda 环境管理）
• Cli 工具：CLI-Anything

环境变量配置

Python 环境构建

conda create -n cli-anything python=3.12 pip -y
conda run -n cli-anything python --version
conda run -n cli-anything python -m pip --version

Blender Cli 依赖安装

conda run -n cli-anything python -m pip install "D:\CLI-Anything\blender\agent-harness"

Blender 与 Cli 环境变量配置

PowerShell 中文乱码临时处理

chcp 65001
[Console]::InputEncoding = [System.Text.UTF8Encoding]::new()
[Console]::OutputEncoding = [System.Text.UTF8Encoding]::new()
$env:PYTHONUTF8 = "1"
$env:PYTHONIOENCODING = "utf-8"

持久写入 Conda 环境变量

conda env config vars set -n cli-anything `
  PYTHONUTF8=1 `
  PYTHONIOENCODING=utf-8

自动配置用户 Path 环境变量

$condaEnvironmentData = conda env list --json | ConvertFrom-Json

$cliEnvPrefix = $condaEnvironmentData.envs |
    Where-Object { $_ -match "[\\/]cli-anything$" } |
    Select-Object -First 1

$cliEnvPrefix

# ------------------------------

$cliScriptsDir = Join-Path $cliEnvPrefix "Scripts"

$cliScriptsDir
Test-Path $cliScriptsDir
Get-ChildItem $cliScriptsDir -Filter "cli-anything-blender*"

# ------------------------------

$blenderDir = "D:\Blender Foundation\Blender 4.2"

if (-not (Test-Path $cliScriptsDir)) {
    throw "CLI Scripts 目录不存在：$cliScriptsDir"
}

if (-not (Test-Path (Join-Path $blenderDir "blender.exe"))) {
    throw "找不到 blender.exe：$blenderDir"
}

$currentUserPath = [Environment]::GetEnvironmentVariable("Path", "User")
$currentUserParts = @(
    $currentUserPath -split ";" |
    Where-Object { -not [string]::IsNullOrWhiteSpace($_) }
)

foreach ($requiredPath in @($blenderDir, $cliScriptsDir)) {
    if ($currentUserParts -notcontains $requiredPath) {
        $currentUserParts += $requiredPath
    }
}

[Environment]::SetEnvironmentVariable(
    "Path",
    ($currentUserParts -join ";"),
    "User"
)

$env:Path = "$blenderDir;$cliScriptsDir;$env:Path"

环境验证

where.exe blender
where.exe cli-anything-blender

blender --version
cli-anything-blender --help

验证通过后，即可结合 Codex 批量处理模型。