按 U-Claw-海外化改造方案.md 与 范围决策记录.md 实施。这是 fork，不回上游： 海外版删掉的正是上游的中国市场默认值。 阶段 0 地基 - 下载源全部改国际：脚本/CI 61 处 + lockfile 880 条 npmmirror URL 归零 （lockfile 那 880 条是 npm 的 resolved 字段，脚本层参数化根本绕不过它） - 移除 install.ps1 里三个第三方 GitHub 加速代理，bundle 改直连 + SHA256 校验 （原来只检查"文件大于 1MB"就解压运行） - 技能内容与分发分离：skills/manifest.json 单一来源，install.sh 1170→658 行、 install.ps1 721→546 行，两者技能内容归零 实测原来是三份不一致：skills-cn 完整、install.sh 约 40%、install.ps1 约 17%， 且 7 个通用技能只有 U 盘版有 —— 一键安装的用户一个能用的技能都没有 - Node 版本三种（v22.14/16/22.1）统一，新建 NODE_VERSION 单一来源 - Config 页三份合一。portable/Config.html 用根相对路径调 API 却只从 file:// 打开， 保存功能已静默失效两个月；现缩为 120 行重定向壳 - 测试接入 CI（此前 node --test 无人运行，所有断言形同虚设） 阶段 1 双语可用 - 浏览器侧 i18n：JSON 为源、生成经典 script（file:// 下 fetch 本地 JSON 被拦） 语言跟盘走不跟机器走：启动器写 data/.openclaw/locale.js - 8 处硬编码 lang="zh-CN" 归零，data-i18n 覆盖 213 处，词条 en/zh 各 279 条 - B3 单框 Key：12 张模型卡 → 一个输入框，前缀识别 provider， 服务端 /api/test-key 发 1-token 请求实测，错误映射成人话 Key 填错到得知：从"直到对话失败"降到 ≤1 秒 - 区域格式 SG：DD/MM/YYYY、12 小时、S$、Asia/Singapore （ICU 在 en-SG 下把 SGD 渲染成裸 $，与美元无法区分，故自行拼 S$） - README 内容分叉而非翻译，§1.3 证据清单逐条清零 阶段 2 降门槛 - 启动逻辑上移 lib/start.mjs：Windows-Start.bat 220→28 行、 Mac-Start.command 235→33 行 修掉 Mac 侧两个 bug：控制台端口硬编码 18788（回落时打开死页）、 微信插件从未在 Mac 上安装 - U 盘根目录 23 → 3 个可点文件，其余进 advanced/ - 首启向导：语言 → 用途（7 角色，manifest 驱动）→ 密钥，答过不再问 - 三档界面，Simple 档隐藏一切技术名词 - 自动自愈：启动失败先自查自修，修不好导出脱敏诊断包 （Doctor 从"用户要知道去点的工具"变成后台机制） 阶段 3 技能库 - 19 个英文技能，planned 归零。sg-weather / sg-transport 的端点均实测过 - SkillHub 从 56 张手写第三方卡片改为 manifest 生成：703→125 行，中文归零 其他 - origin.json 收拢所有运行时地址，tests/origin.test.mjs 保证迁移不会漏 - portable/ 下用户可见中文归零（由断言保证） - 82 项测试 未验证（本机无 Windows / 无 pwsh）： - install.ps1、setup.ps1 约 210 行改动从未经 PowerShell 解析器 - 完整启动路径仅在假 node + 假 openclaw 上冒烟 - 8 个 .bat 的盘根推导仅静态断言 详见 U盘实测清单.md 受阻： - 隐藏黑窗口 —— 需代码签名证书（.vbs 已被 Windows 弃用，替代方案都要签名） - 场景卡 —— OpenClaw 上游 Dashboard 无预填 prompt 接口 - 官网 36 条 —— 上游 2026-04-14 拆到私有仓库，无权限
This commit is contained in:
@@ -1,14 +1,21 @@
|
||||
// setup-local-model.mjs — 内网/本地模型一键配置（不碰 Control UI）
|
||||
// setup-local-model.mjs — set up a local or self-hosted model from the command line.
|
||||
//
|
||||
// 给"内网/离线"场景:用纯命令行问几个问题，直接写好 openclaw.json，
|
||||
// 再当场实测能不能连上、能不能回话。全程不依赖会挂的 dashboard 网页。
|
||||
// 支持两类本地/内网模型：
|
||||
// 1) Ollama（本机，http://127.0.0.1:11434）
|
||||
// 2) newapi / 任意 OpenAI 兼容中转（内网 IP + token）
|
||||
// For offline and corporate-network situations: ask a few questions, write
|
||||
// openclaw.json directly, then actually test that the model answers. Deliberately
|
||||
// does not depend on the dashboard, which is the thing that tends to be broken
|
||||
// when someone reaches for this tool.
|
||||
//
|
||||
// 写入只 merge 模型相关字段，保留 gateway 等原有配置；写前自动备份。
|
||||
// Two kinds of endpoint:
|
||||
// 1) Ollama on this machine (http://127.0.0.1:11434)
|
||||
// 2) Any OpenAI-compatible endpoint — a company server, a relay (URL + token)
|
||||
//
|
||||
// 用法：node setup-local-model.mjs <CONFIG_PATH>
|
||||
// Only the model-related fields are merged in; gateway and other settings are
|
||||
// left alone, and the previous file is backed up first.
|
||||
//
|
||||
// Usage: node setup-local-model.mjs <CONFIG_PATH>
|
||||
//
|
||||
// Input: a real terminal gets interactive readline; piped input (tests, scripts)
|
||||
// is read once and dequeued line by line.
|
||||
|
||||
import { readFileSync, writeFileSync, existsSync, copyFileSync, mkdirSync } from 'node:fs';
|
||||
import { dirname } from 'node:path';
|
||||
@@ -17,7 +24,6 @@ import { stdin as input, stdout as output } from 'node:process';
|
||||
import http from 'node:http';
|
||||
import https from 'node:https';
|
||||
|
||||
// 输入抽象：真控制台(TTY)走交互式 readline；被管道喂入(测试/脚本)则一次读完按行出队。
|
||||
function makePrompter() {
|
||||
if (input.isTTY) {
|
||||
const rl = createInterface({ input, output });
|
||||
@@ -69,7 +75,7 @@ async function chatTest(baseUrl, apiKey, modelId) {
|
||||
const res = await requestText(url, {
|
||||
method: 'POST',
|
||||
headers: { 'Content-Type': 'application/json', ...(apiKey ? { Authorization: `Bearer ${apiKey}` } : {}) },
|
||||
body: JSON.stringify({ model: modelId, messages: [{ role: 'user', content: '请回复四个字：连接成功' }], max_tokens: 64, stream: false }),
|
||||
body: JSON.stringify({ model: modelId, messages: [{ role: 'user', content: 'Reply with exactly: connection ok' }], max_tokens: 64, stream: false }),
|
||||
timeoutMs: CHAT_TIMEOUT_MS,
|
||||
});
|
||||
const ms = Date.now() - started;
|
||||
@@ -111,10 +117,10 @@ function requestText(rawUrl, { method = 'GET', headers = {}, body, timeoutMs })
|
||||
|
||||
async function main() {
|
||||
const configPath = process.argv[2] || process.env.OPENCLAW_CONFIG_PATH;
|
||||
if (!configPath) { line('用法: node setup-local-model.mjs <CONFIG_PATH>'); process.exitCode = 2; return; }
|
||||
if (!configPath) { line('Usage: node setup-local-model.mjs <CONFIG_PATH>'); process.exitCode = 2; return; }
|
||||
|
||||
line('========================================');
|
||||
line(' U-Claw 内网/本地模型 一键配置');
|
||||
line(' U-Claw — set up a local or self-hosted model');
|
||||
line('========================================');
|
||||
line('');
|
||||
|
||||
@@ -122,32 +128,32 @@ async function main() {
|
||||
const ask = rl.ask;
|
||||
|
||||
try {
|
||||
line('选择模型类型：');
|
||||
line(' 1) Ollama（本机部署，http://127.0.0.1:11434）');
|
||||
line(' 2) newapi / 其它 OpenAI 兼容中转（内网 IP + token）');
|
||||
const kind = await ask('输入 1 或 2', '1');
|
||||
line('Which kind of model?');
|
||||
line(' 1) Ollama running on this machine (http://127.0.0.1:11434)');
|
||||
line(' 2) A self-hosted or company endpoint (any OpenAI-compatible URL + token)');
|
||||
const kind = await ask('Enter 1 or 2', '1');
|
||||
|
||||
let providerKey, baseUrl, apiKey, modelId;
|
||||
if (kind === '2') {
|
||||
providerKey = 'newapi';
|
||||
line('');
|
||||
line('提示：baseUrl 通常形如 http://192.168.1.50:3000/v1（注意大多要带 /v1）');
|
||||
baseUrl = await ask('newapi 地址 baseUrl', 'http://192.168.1.50:3000/v1');
|
||||
line('The endpoint usually looks like http://192.168.1.50:3000/v1 — most need the /v1 on the end.');
|
||||
baseUrl = await ask('Endpoint URL', 'http://192.168.1.50:3000/v1');
|
||||
apiKey = await ask('token / API Key', '');
|
||||
modelId = await ask('模型 ID（管理员给的，如 deepseek-v3）', '');
|
||||
modelId = await ask('Model ID — your admin will have given you this, e.g. deepseek-v3', '');
|
||||
} else {
|
||||
providerKey = 'ollama';
|
||||
line('');
|
||||
baseUrl = await ask('Ollama 地址（一般本机默认即可）', 'http://127.0.0.1:11434/v1');
|
||||
// Ollama 的 OpenAI 兼容端点在 /v1；自动补上
|
||||
baseUrl = await ask('Ollama URL — the default is right unless you changed it', 'http://127.0.0.1:11434/v1');
|
||||
// Ollama's OpenAI-compatible endpoint lives at /v1; add it if missing
|
||||
if (!/\/v1\/?$/.test(baseUrl)) baseUrl = baseUrl.replace(/\/+$/, '') + '/v1';
|
||||
apiKey = 'ollama'; // 本地占位 key，任意值即可
|
||||
modelId = await ask('模型名（先用 ollama list 查，如 qwen2.5 / llama3.1）', 'qwen2.5');
|
||||
apiKey = 'ollama'; // Ollama ignores auth; any placeholder works
|
||||
modelId = await ask('Model name — run `ollama list` to see yours, e.g. qwen2.5', 'qwen2.5');
|
||||
}
|
||||
|
||||
if (!baseUrl || !modelId) { line(''); line('地址或模型 ID 为空，已取消。'); process.exitCode = 2; return; }
|
||||
if (!baseUrl || !modelId) { line(''); line('No endpoint or model ID given — nothing was changed.'); process.exitCode = 2; return; }
|
||||
|
||||
// 读取并合并现有配置（保留 gateway 等），写前备份
|
||||
// Merge into the existing config so gateway and other settings survive; back up first
|
||||
let config = {};
|
||||
if (existsSync(configPath)) {
|
||||
try { config = JSON.parse(readFileSync(configPath, 'utf8')); }
|
||||
@@ -174,39 +180,39 @@ async function main() {
|
||||
|
||||
writeFileSync(configPath, JSON.stringify(config, null, 2), 'utf8');
|
||||
line('');
|
||||
line(`✓ 已写入配置：${configPath}`);
|
||||
line(`✓ Saved to ${configPath}`);
|
||||
line(` provider=${providerKey} baseUrl=${baseUrl} model=${modelId}`);
|
||||
if (existsSync(configPath + '.bak')) line(` （原配置已备份为 openclaw.json.bak）`);
|
||||
if (existsSync(configPath + '.bak')) line(' Your previous settings were kept as openclaw.json.bak');
|
||||
line('');
|
||||
|
||||
// 当场实测
|
||||
line('正在实测：发一条对话给模型...');
|
||||
// Prove it actually works rather than just claiming it was saved
|
||||
line('Testing it — sending one message to the model…');
|
||||
const r = await chatTest(baseUrl, apiKey, modelId);
|
||||
line('');
|
||||
if (r.ok) {
|
||||
line(`✓✓ 跑通了！模型回复 (${r.ms}ms)：${r.reply.slice(0, 120) || '(空回复但请求成功)'}`);
|
||||
line(`✓ It works. The model replied in ${r.ms}ms: ${r.reply.slice(0, 120) || '(empty reply, but the request succeeded)'}`);
|
||||
line('');
|
||||
line('配置完成。现在双击 Windows-Start.bat 即可正常使用（对话可走 CLI 或 Dashboard）。');
|
||||
line('You are set. Start U-Claw normally and it will use this model.');
|
||||
} else if (r.status) {
|
||||
line(`✗ 服务端 HTTP ${r.status} (${r.ms}ms)：${r.body}`);
|
||||
line(`✗ The server answered with HTTP ${r.status} in ${r.ms}ms: ${r.body}`);
|
||||
line('');
|
||||
if (r.status === 401 || r.status === 403) line('→ 网络通，但 token / key 不对（Ollama 可忽略鉴权，newapi 请核对 token）。');
|
||||
else if (r.status === 404) line('→ 网络通，但路径或模型 ID 不对（检查 baseUrl 是否要带 /v1、模型名是否正确）。');
|
||||
else line('→ 网络通，服务端报错，把上面内容发管理员。');
|
||||
if (r.status === 401 || r.status === 403) line(' The network is fine but the token was rejected. Ollama ignores auth entirely; for a company endpoint, check the token with whoever issued it.');
|
||||
else if (r.status === 404) line(' The network is fine but the path or model ID is wrong. Check whether the endpoint needs /v1 on the end, and that the model name matches exactly.');
|
||||
else line(' The network is fine; the server itself errored. Send the text above to whoever runs it.');
|
||||
} else {
|
||||
const host = (() => { try { return new URL(withScheme(baseUrl)).hostname; } catch { return baseUrl; } })();
|
||||
line(`✗ 连不上：${r.error} (${r.ms}ms)`);
|
||||
line(`✗ Could not reach it: ${r.error} (${r.ms}ms)`);
|
||||
line('');
|
||||
if (providerKey === 'ollama') {
|
||||
line('→ 本机 Ollama 没连上，多半是 Ollama 没启动或模型没拉。请在本机执行：');
|
||||
line(' ollama serve （启动服务，若已是后台服务可跳过）');
|
||||
line(` ollama pull ${modelId} （把模型拉到本地，离线需提前准备好）`);
|
||||
line(' 然后重新运行本工具。');
|
||||
line(' Ollama is not answering. It is usually not running, or the model was never pulled. On this machine, run:');
|
||||
line(' ollama serve starts it, skip if it already runs in the background');
|
||||
line(` ollama pull ${modelId} downloads the model — do this while you still have internet`);
|
||||
line(' Then run this tool again.');
|
||||
} else {
|
||||
line('→ 是地址错 / 内网不通 / 防火墙 / 模型服务没起。');
|
||||
line(' 在这台机器上自测：');
|
||||
line(' Either the address is wrong, the network cannot reach it, a firewall is blocking it, or the service is down.');
|
||||
line(' Test it from this machine with:');
|
||||
line(` ping ${host}`);
|
||||
line(' 让机房管理员确认 IP、端口、防火墙放行。');
|
||||
line(' If that also fails, ask whoever runs the server to confirm the address, port and firewall rules.');
|
||||
}
|
||||
}
|
||||
} finally {
|
||||
|
||||
Reference in New Issue
Block a user