python-manager-discovery

作者: microsoft

環境管理器專用的發現模式與已知問題。在處理或審查 conda、poetry、pipenv、pyenv 等環境發現程式碼時使用。

npx skills add https://github.com/microsoft/vscode-python-environments --skill python-manager-discovery

Environment Manager Discovery Patterns

This skill documents manager-specific discovery patterns, environment variable precedence, and known issues.

Manager Quick Reference

ManagerConfig FilesCache LocationKey Env Vars
Poetrypoetry.toml, pyproject.toml, config.tomlPlatform-specificPOETRY_VIRTUALENVS_IN_PROJECT, POETRY_CACHE_DIR
PipenvPipfile, Pipfile.lockXDG or WORKON_HOMEWORKON_HOME, XDG_DATA_HOME
Pyenv.python-version, versions/~/.pyenv/ or pyenv-winPYENV_ROOT, PYENV_VERSION
Condaenvironment.yml, conda-meta/Registries + pathsCONDA_PREFIX, CONDA_DEFAULT_ENV
venvpyvenv.cfgIn-projectNone

Poetry

Discovery Locations

Virtualenvs cache (default):

  • Windows: %LOCALAPPDATA%\pypoetry\Cache\virtualenvs
  • macOS: ~/Library/Caches/pypoetry/virtualenvs
  • Linux: ~/.cache/pypoetry/virtualenvs

In-project (when enabled):

  • .venv/ in project root

Config Precedence (highest to lowest)

  1. Local config: poetry.toml in project root
  2. Environment variables: POETRY_VIRTUALENVS_*
  3. Global config: ~/.config/pypoetry/config.toml

Known Issues

IssueDescriptionFix
{cache-dir} placeholderNot resolved in paths from configResolve placeholder before use
Wrong default pathWindows/macOS differ from LinuxUse platform-specific defaults
In-project detectionPOETRY_VIRTUALENVS_IN_PROJECT must be checkedCheck env var first, then config

Code Pattern

async function getPoetryVirtualenvsPath(): Promise<string> {
    // 1. Check environment variable first
    const envVar = process.env.POETRY_VIRTUALENVS_PATH;
    if (envVar) return envVar;

    // 2. Check local poetry.toml
    const localConfig = await readPoetryToml(projectRoot);
    if (localConfig?.virtualenvs?.path) {
        return resolvePoetryPath(localConfig.virtualenvs.path);
    }

    // 3. Use platform-specific default
    return getDefaultPoetryCache();
}

function resolvePoetryPath(configPath: string): string {
    // Handle {cache-dir} placeholder
    if (configPath.includes('{cache-dir}')) {
        const cacheDir = getDefaultPoetryCache();
        return configPath.replace('{cache-dir}', cacheDir);
    }
    return configPath;
}

Pipenv

Discovery Locations

Default:

  • Linux: ~/.local/share/virtualenvs/ (XDG_DATA_HOME)
  • macOS: ~/.local/share/virtualenvs/
  • Windows: ~\.virtualenvs\

When WORKON_HOME is set:

  • Use $WORKON_HOME/ directly

Environment Variables

VarPurpose
WORKON_HOMEOverride virtualenv location
XDG_DATA_HOMEBase for Linux default
PIPENV_VENV_IN_PROJECTCreate .venv/ in project

Known Issues

IssueDescriptionFix
Missing WORKON_HOME supportEnv var not checkedRead env var before defaults
Missing XDG_DATA_HOME supportNot used on LinuxCheck XDG spec

Code Pattern

function getPipenvVirtualenvsPath(): string {
    // Check WORKON_HOME first
    if (process.env.WORKON_HOME) {
        return process.env.WORKON_HOME;
    }

    // Check XDG_DATA_HOME on Linux
    if (process.platform === 'linux') {
        const xdgData = process.env.XDG_DATA_HOME || path.join(os.homedir(), '.local', 'share');
        return path.join(xdgData, 'virtualenvs');
    }

    // Windows/macOS defaults
    return path.join(os.homedir(), '.virtualenvs');
}

PyEnv

Discovery Locations

Unix:

  • ~/.pyenv/versions/ (default)
  • $PYENV_ROOT/versions/ (if PYENV_ROOT set)

Windows (pyenv-win):

  • %USERPROFILE%\.pyenv\pyenv-win\versions\
  • Different directory structure than Unix!

Key Differences: Unix vs Windows

AspectUnixWindows (pyenv-win)
Commandpyenvpyenv.bat
Root~/.pyenv/%USERPROFILE%\.pyenv\pyenv-win\
Shims~/.pyenv/shims/%USERPROFILE%\.pyenv\pyenv-win\shims\

Known Issues

IssueDescriptionFix
path.normalize() vs path.resolve()Windows drive letter missingUse path.resolve() on both sides
Wrong command on WindowsLooking for pyenv instead of pyenv.batCheck for .bat extension

Code Pattern

function getPyenvRoot(): string {
    if (process.env.PYENV_ROOT) {
        return process.env.PYENV_ROOT;
    }

    if (process.platform === 'win32') {
        // pyenv-win uses different structure
        return path.join(os.homedir(), '.pyenv', 'pyenv-win');
    }

    return path.join(os.homedir(), '.pyenv');
}

function getPyenvVersionsPath(): string {
    const root = getPyenvRoot();
    return path.join(root, 'versions');
}

// Use path.resolve() for comparisons!
function comparePyenvPaths(pathA: string, pathB: string): boolean {
    return path.resolve(pathA) === path.resolve(pathB);
}

Conda

Discovery Locations

Environment locations:

  • Base install envs/ directory
  • ~/.conda/envs/
  • Paths in ~/.condarc envs_dirs

Windows Registry:

  • HKCU\Software\Python\ContinuumAnalytics\
  • HKLM\SOFTWARE\Python\ContinuumAnalytics\

Shell Activation

ShellActivation Command
bash, zshsource activate envname
fishconda activate envname (NOT source!)
PowerShellconda activate envname
cmdactivate.bat envname

Known Issues

IssueDescriptionFix
Fish shell activationUses bash-style commandUse fish-compatible syntax
Registry pathsMay be stale/invalidVerify paths exist
Base vs named envsDifferent activationCheck if activating base

Code Pattern

function getCondaActivationCommand(shell: ShellType, envName: string): string {
    switch (shell) {
        case 'fish':
            // Fish uses different syntax!
            return `conda activate ${envName}`;
        case 'cmd':
            return `activate.bat ${envName}`;
        case 'powershell':
            return `conda activate ${envName}`;
        default:
            // bash, zsh
            return `source activate ${envName}`;
    }
}

venv

Discovery

Identification:

  • Look for pyvenv.cfg file in directory
  • Contains home and optionally version keys

Version Extraction Priority

  1. version field in pyvenv.cfg
  2. Parse from home path (e.g., Python311)
  3. Spawn Python executable (last resort)

Code Pattern

async function getVenvVersion(venvPath: string): Promise<string | undefined> {
    const cfgPath = path.join(venvPath, 'pyvenv.cfg');

    try {
        const content = await fs.readFile(cfgPath, 'utf-8');
        const lines = content.split('\n');

        for (const line of lines) {
            const [key, value] = line.split('=').map((s) => s.trim());
            if (key === 'version') {
                return value;
            }
        }

        // Fall back to parsing home path
        const homeLine = lines.find((l) => l.startsWith('home'));
        if (homeLine) {
            const home = homeLine.split('=')[1].trim();
            const match = home.match(/(\d+)\.(\d+)/);
            if (match) {
                return `${match[1]}.${match[2]}`;
            }
        }
    } catch {
        // Config file not found or unreadable
    }

    return undefined;
}

PET Server (Native Finder)

JSON-RPC Communication

The PET server is a Rust-based locator that communicates via JSON-RPC over stdio.

Known Issues

IssueDescriptionFix
No timeoutJSON-RPC can hang foreverAdd Promise.race with timeout
Silent spawn errorsExtension continues without envsSurface spawn errors to user
Resource leaksWorker pool not cleaned upDispose on deactivation
Type guard missingResponse types not validatedAdd runtime type checks
Cache key collisionPaths normalize to same keyUse consistent normalization

Code Pattern

async function fetchFromPET<T>(method: string, params: unknown): Promise<T> {
    const timeout = 30000; // 30 seconds

    const result = await Promise.race([
        this.client.request(method, params),
        new Promise<never>((_, reject) => setTimeout(() => reject(new Error('PET server timeout')), timeout)),
    ]);

    // Validate response type
    if (!isValidResponse<T>(result)) {
        throw new Error(`Invalid response from PET: ${JSON.stringify(result)}`);
    }

    return result;
}

來自 microsoft 的更多技能

oss-growth
microsoft
開源增長駭客角色
agent-framework-azure-ai-py
microsoft
使用Microsoft Agent Framework Python SDK(agent-framework-azure-ai)构建Azure AI Foundry代理。适用于使用AzureAIAgentsProvider创建持久化代理、使用托管工具(代码解释器、文件搜索、网络搜索)、集成MCP服务器、管理对话线程或实现流式响应。涵盖函数工具、结构化输出和多工具代理。
development
airunway-aks-setup
microsoft
在AKS上設定AI Runway——從裸叢集到執行模型。涵蓋叢集驗證、控制器安裝、GPU評估、供應商設定及首次部署。時機:「設定AI Runway」、「上線AKS叢集」、「安裝AI Runway」、「airunway設定」、「部署模型至AKS」、「在AKS上進行GPU推論」、「在AKS上設定KAITO」、「在AKS上執行LLM」、「在AKS上使用vLLM」、「在AKS上設定模型服務」、「AI Runway控制器」。
devops
appinsights-instrumentation
microsoft
使用Azure Application Insights檢測Web應用程式的指南。提供遙測模式、SDK設定與組態參考。適用時機:如何檢測應用程式、App Insights SDK、遙測模式、什麼是App Insights、Application Insights指南、檢測範例、APM最佳實踐。
devops
applicationinsights-web-ts
microsoft
使用Application Insights JavaScript SDK(@microsoft/applicationinsights-web)為瀏覽器/Web應用程式進行檢測。適用於真實使用者監控(RUM)——頁面檢視、點擊、AJAX/fetch依賴、例外、自訂事件,以及與後端OpenTelemetry追蹤關聯的瀏覽器端GenAI代理追蹤。涵蓋SDK載入器指令碼與npm設定、框架擴充(React、React Native、Angular)、點擊分析、遙測初始化器,以及從瀏覽器發出的代理/工具/模型span的OTel GenAI語意慣例。
devops
azure-ai-anomalydetector-java
microsoft
使用適用於 Java 的 Azure AI 異常偵測器 SDK 建置異常偵測應用程式。在實作單變量/多變量異常偵測、時間序列分析或 AI 驅動監控時使用。
development
azure-ai-language-conversations-py
microsoft
使用 azure-ai-language-conversations Python SDK 實作對話語言理解(CLU)。當使用 ConversationAnalysisClient 分析對話意圖與實體、建置 NLP 功能,或將語言理解整合至應用程式時使用。
development
azure-ai-ml-py
microsoft
Azure Machine Learning SDK v2 for Python。用於機器學習工作區、作業、模型、資料集、計算資源與管線。 觸發詞:「azure-ai-ml」、「MLClient」、「workspace」、「model registry」、「training jobs」、「datasets」。
development