Skip to content

ai.md

Latest commit

 

History

History
152 lines (109 loc) · 3.4 KB

ai.md

File metadata and controls

152 lines (109 loc) · 3.4 KB

fnm-ai

fnm-ai is a Windows-first npm wrapper around fnm for managing Node.js versions with natural-language requests.

It installs two executable names:

fnm-api
fnm-ai

Use fnm-api for normal npm-based usage. The lower-level Rust binary also exposes the same feature as fnm ai and fnm api.

Install

npm install -g fnm-ai

The npm package currently targets Windows x64:

{
  "os": ["win32"],
  "cpu": ["x64"]
}

Configure the AI provider

Common requests are parsed locally first. If the local parser cannot understand the request, fnm calls an OpenAI-compatible chat/completions provider.

The default model is Kimi kimi-k2.6. Save the provider config once:

fnm-api config set --base-url https://api.moonshot.ai/v1 --api-key <your-kimi-api-key>

You can also set the model explicitly:

fnm-api config set --base-url https://api.moonshot.ai/v1 --api-key <your-kimi-api-key> --model kimi-k2.6

Inspect the saved config without printing the full API key:

fnm-api config get

Print the config file path:

fnm-api config path

On Windows this is usually:

%APPDATA%\fnm-ai\config.json

Environment variables override the saved config:

$env:FNM_AI_BASE_URL = "https://api.moonshot.ai/v1"
$env:FNM_AI_API_KEY = "<your-kimi-api-key>"
$env:FNM_AI_MODEL = "kimi-k2.6"

Use

Run a single request:

fnm-api "帮我切换到 Node 20"
fnm-api "安装最新 LTS，并切换过去"
fnm-api "use node 22, install it if missing"
fnm-api "检查当前环境配置"
fnm-api "列出已安装版本"

Start the interactive prompt:

fnm-api

The prompt has two built-in presets:

1. switch environment
2. check current environment config

Preset 1 applies the current project or default Node version using .node-version, .nvmrc, package.json#engines.node, or the default alias.

Preset 2 prints the fnm directory, Node mirror, architecture, version-file strategy, corepack setting, resolve-engines setting, multishell path, PATH status, and current Node version.

Environment bootstrap

When launched through npm, fnm-api reads %APPDATA%\fnm-ai\config.json and maps it to FNM_AI_BASE_URL, FNM_AI_API_KEY, and FNM_AI_MODEL.

If FNM_MULTISHELL_PATH is missing, fnm-api also runs:

fnm env --json

It merges the returned FNM_* variables into the child process and prepends the multishell path to PATH. This lets fnm-api "检查当前环境配置" work even when the current PowerShell session has not loaded fnm env.

One limitation remains: a child process cannot permanently mutate the parent terminal environment. If you want node, npm, and pnpm in the current PowerShell window to keep using the selected version after fnm-api exits, add fnm to your PowerShell profile:

fnm env --use-on-cd --shell powershell | Out-String | Invoke-Expression

Without that profile setup, fnm-api can initialize and use the environment for its own process and update fnm's multishell link, but the already-running parent shell may not refresh its own PATH.

Safety

AI output is translated into a fixed set of actions:

help
exit
switch_environment
check_environment
install
use
list_local
list_remote
current
default
uninstall

The model cannot return arbitrary shell commands for execution.