@model-go/cli 0.1.2-rc.1 → 0.1.3

package/README.md

@@ -2,7 +2,7 @@
 
 The official CLI for modelgo. Pairs with AI agent skills (Claude Code, Codex, Gemini CLI, etc.) so AI agents can operate modelgo on your behalf.
 
-> **v0 framework stage.** Business APIs are not wired up yet; the only command is `hello` (a demo to verify the install pipeline).
+> **v0 framework stage.** Device-flow `auth` commands are available. `modelgo pay request` supports anonymous x402 pay-per-call calls to model gateway APIs; API key management is not wired up yet.
 
 ## Install
 
@@ -12,23 +12,92 @@ npx @model-go/cli@latest install
 
 This runs an interactive wizard that:
 
-1. Installs `@model-go/cli` globally via npm (which downloads the Go binary from GitHub Releases).
+1. Installs `@model-go/cli` globally via npm (which downloads the Go binary from GitHub Releases as `modelgo`).
 2. Distributes `modelgo-*` skills to every AI agent installed on your machine (Claude Code, Trae, Trae CN, and other agents in the [skills](https://www.npmjs.com/package/skills) ecosystem).
 
 After install, restart your AI agent (open a new chat / session) and try:
 
-> "Have modelgo-cli say hello to me."
+> "Check my modelgo login status."
 
-Your AI should find the `modelgo-hello` skill and run `modelgo-cli hello`.
+Your AI should find the `modelgo-shared` skill and run `modelgo auth status`.
 
 ## Direct commands
 
 ```bash
-modelgo-cli --version
-modelgo-cli hello [--name NAME]
-modelgo-cli --help
+modelgo --version
+modelgo auth login
+modelgo auth status
+modelgo auth logout
+modelgo pay methods
+modelgo pay request --env cn --path /v1/chat/completions --method POST --data '{"model":"gpt-4o","messages":[]}' --json
+modelgo --help
 ```
 
+`modelgo auth login` without `--no-wait` is a blocking command: it prints the
+authorization URL, then keeps polling until the browser approval completes or
+the device code expires.
+
+### Environments
+
+The CLI talks to a named environment. Two are built in:
+
+| Env    | Base URL                     |
+| ------ | ---------------------------- |
+| `cn`   | `https://api.modelgo.com`    |
+| `intl` | `https://api.modelgo.global` |
+
+The active environment defaults to `cn`. Switch it, or register your own:
+
+```bash
+modelgo env list                 # show built-in + custom envs (active marked with *)
+modelgo env current              # print the active env
+modelgo env use intl             # switch the active env
+modelgo env add <name> --base-url https://...   # register or override an env URL
+modelgo env remove <name>        # remove a custom env or override
+```
+
+Environment definitions and the active selection live in `~/.modelgo/config.json`.
+Credentials are stored per-env in `~/.modelgo/auth.json`, so switching environments
+preserves each login. There are no environment variables to set.
+
+Auth commands operate on the active env; pass `--env <name>` to target a different one:
+
+```bash
+modelgo auth login --env intl
+modelgo auth status --env intl
+modelgo auth logout --all          # clear every env
+```
+
+### Non-streaming agent flows
+
+For AI-agent harnesses that cannot stream intermediate output to the user, use:
+
+```bash
+modelgo auth login --no-wait --json
+modelgo auth login --device-code <DEVICE_CODE>
+```
+
+Do not show the URL and then immediately block on `--device-code` in the same
+turn. In final-message-only harnesses, that pattern prevents the user from ever
+seeing the URL before the agent starts waiting.
+
+> **`--json` emits newline-delimited JSON (NDJSON).** In the default waiting
+> flow (`modelgo auth login --json` without `--no-wait`), stdout receives two
+> JSON objects: first the device-code object (with `verification_url`), then
+> the authenticated object once approved. Parse line by line — don't
+> `JSON.parse` the whole stream. With `--no-wait`, only the device-code object
+> is printed.
+
+### x402 pay-per-call model APIs
+
+Use `modelgo pay request` when the user wants to call a model API without logging in to ModelGo first. The CLI opts into the gateway x402 path with `X-Payment-Protocol: x402`.
+
+```bash
+modelgo pay request --env cn --path /v1/chat/completions --method POST --data '{"model":"gpt-4o","messages":[{"role":"user","content":"hello"}]}' --intent "调用 gpt-4o 完成聊天补全" --json
+```
+
+In the `cn` environment, if the gateway returns a 402 that advertises `alipay:*`, the CLI writes the `PAYMENT-REQUIRED` payload to a local file and returns a structured handoff for `alipay-payment-skill`. In `intl`, the CLI does not route to Alipay; it returns the advertised x402 options for a future non-Alipay rail.
+
 ## Upgrade
 
 Re-run the installer; it detects an out-of-date install and upgrades in place:

package/checksums.txt

@@ -1,6 +1,6 @@
-915ed381929af9346da75a8cd05cc2e07e2dffad87a2611273e62950888d3a13  modelgo-cli-0.1.2-rc.1-darwin-amd64.tar.gz
-db43edac818eba5d1275c47e7a8cf6bce87d46565a54fc9db276bce92b861051  modelgo-cli-0.1.2-rc.1-darwin-arm64.tar.gz
-fbc7e92c959f5d62ced3940481afbd444655f91129acbb7bc6ff5f8d9a200729  modelgo-cli-0.1.2-rc.1-linux-amd64.tar.gz
-d9fb3ccf896bb4b493b3af2a095d9406c54809d01b97734929135ae6347c07ec  modelgo-cli-0.1.2-rc.1-linux-arm64.tar.gz
-0cc9d8681d4be1718641ae1dfb609c4e967aa741150cff9a5b598c7a962931e2  modelgo-cli-0.1.2-rc.1-windows-amd64.zip
-608930e7904701a794f3350ca594fbc992b35eaee619c25e6c082172c05c69ac  modelgo-cli-0.1.2-rc.1-windows-arm64.zip
+d2301f86bdccb4225e959062382d20108eaac2ce023456f306485b8addcacacc  modelgo-0.1.2-rc.1-SNAPSHOT-7701312-darwin-amd64.tar.gz
+88480f50f971139c72866d73a69a67368a08a260c96b71e04add4eb7d6676ecf  modelgo-0.1.2-rc.1-SNAPSHOT-7701312-darwin-arm64.tar.gz
+9bc1719a182e3b996a793491775fd501da419a6c526d2b8a7cb2e72479cd9e1c  modelgo-0.1.2-rc.1-SNAPSHOT-7701312-linux-amd64.tar.gz
+534554cb7b7fff278260b4e030fcbe30a0029c146d46980493a1d902e65c7669  modelgo-0.1.2-rc.1-SNAPSHOT-7701312-linux-arm64.tar.gz
+13cfa7761480b595fab0e1c3400c299013d29eaca3643282f2817b872b7d89d1  modelgo-0.1.2-rc.1-SNAPSHOT-7701312-windows-amd64.zip
+3d57522bf6cfaeaded3ea3b6db4cf4afd8ddd4f0fbf0bcea71241531898a939c  modelgo-0.1.2-rc.1-SNAPSHOT-7701312-windows-arm64.zip

package/package.json

@@ -1,14 +1,17 @@
 {
   "name": "@model-go/cli",
-  "version": "0.1.2-rc.1",
+  "version": "0.1.3",
   "description": "The official CLI for modelgo. Pairs with AI agent skills so AI agents can operate modelgo on your behalf.",
   "bin": {
-    "modelgo-cli": "scripts/run.js"
+    "modelgo": "scripts/run.js"
   },
   "scripts": {
     "postinstall": "node scripts/install.js",
     "test": "node --test scripts/__tests__/*.test.mjs",
-    "lint:skills": "node scripts/lint-skills.mjs"
+    "lint:skills": "node scripts/lint-skills.mjs",
+    "sync:skills": "node scripts/sync-skill-version.mjs",
+    "generate:reference": "node scripts/generate-reference.mjs",
+    "sync:skill-assets": "node scripts/generate-reference.mjs && node scripts/sync-skill-version.mjs"
   },
   "files": [
     "scripts/install.js",

package/scripts/install-wizard.js

@@ -31,7 +31,7 @@ const messages = {
     step2Spinner:  "正在安装 Skills...",
     step2Done:     "Skills 已安装",
     step2Fail:     "Skills 安装失败。运行以下命令重试: npx skills add %s -y -g",
-    done:          "安装完成！\n试试跟你的 AI 工具（Claude Code、Codex 等）说：\"用 modelgo-cli 跟我打个招呼\"",
+    done:          "安装完成！\n试试跟你的 AI 工具（Claude Code、Codex 等）说：\"用 modelgo 跟我打个招呼\"",
     cancelled:     "安装已取消",
     nonTtyHint:    "非交互模式，已完成 npm 全局安装 + skills 安装。",
   },
@@ -46,7 +46,7 @@ const messages = {
     step2Spinner:  "Installing skills...",
     step2Done:     "Skills installed",
     step2Fail:     "Failed to install skills. Run manually: npx skills add %s -y -g",
-    done:          "You are all set!\nTry asking your AI tool (Claude Code, Codex, etc.): \"Have modelgo-cli say hello to me\"",
+    done:          "You are all set!\nTry asking your AI tool (Claude Code, Codex, etc.): \"Have modelgo say hello to me\"",
     cancelled:     "Installation cancelled",
     nonTtyHint:    "Non-interactive mode. Completed global install + skills install.",
   },

package/scripts/install.js

@@ -9,7 +9,7 @@ const crypto = require("crypto");
 
 const VERSION = require("../package.json").version;
 const REPO = "modelgo/modelgo-cli";
-const NAME = "modelgo-cli";
+const NAME = "modelgo";
 const DEFAULT_MIRROR_HOST = "https://registry.npmmirror.com";
 
 // Allowlist gates the initial request host. curl --location follows redirects
@@ -224,7 +224,7 @@ function install() {
   const downloadUrls = [GITHUB_URL, ...mirrorUrls];
 
   fs.mkdirSync(binDir, { recursive: true });
-  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), "modelgo-cli-"));
+  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), "modelgo-"));
   const archivePath = path.join(tmpDir, archiveName);
 
   try {

package/scripts/run.js

@@ -3,8 +3,8 @@
 // SPDX-License-Identifier: MIT
 //
 // npm bin entrypoint. Routes:
-//   - `modelgo-cli install`  → install-wizard.js (interactive setup)
-//   - everything else        → exec bin/modelgo-cli (the downloaded Go binary)
+//   - `modelgo install`  → install-wizard.js (interactive setup)
+//   - everything else    → exec bin/modelgo (the downloaded Go binary)
 
 const { execFileSync } = require("child_process");
 const path = require("path");
@@ -16,12 +16,12 @@ if (args[0] === "install") {
   require("./install-wizard.js");
 } else {
   const isWindows = process.platform === "win32";
-  const binName = "modelgo-cli" + (isWindows ? ".exe" : "");
+  const binName = "modelgo" + (isWindows ? ".exe" : "");
   const binPath = path.join(__dirname, "..", "bin", binName);
 
   if (!fs.existsSync(binPath)) {
     console.error(
-      `modelgo-cli binary not found at ${binPath}\n` +
+      `modelgo binary not found at ${binPath}\n` +
       `Please reinstall: npm install -g @model-go/cli`
     );
     process.exit(1);