awel 0.2.0 → 0.2.1

package/README.md

@@ -8,16 +8,24 @@ AI-powered development overlay for Next.js. Awel runs a proxy in front of your d
 
 ## Quick Start
 
-```bash
-# Skip if you're already in a Next.js app
-npx create-next-app@latest my-app && cd my-app
+### Create a new project
 
+```bash
 # Set up at least one AI provider (pick one):
 export ANTHROPIC_API_KEY="sk-ant-..."   # Anthropic API
 export OPENAI_API_KEY="sk-..."          # OpenAI
 export GOOGLE_GENERATIVE_AI_API_KEY="..." # Google AI
 # Or install the Claude CLI: https://docs.anthropic.com/en/docs/claude-code
 
+npx awel create
+```
+
+This scaffolds a new Next.js project and marks it for creation mode. Follow the instructions to `cd` into the project and run `npx awel dev`. You'll see a full-page creation UI where you describe what you want to build — the AI agent generates the entire app for you, then transitions to the normal Awel overlay.
+
+### Use with an existing project
+
+```bash
+cd my-existing-next-app
 npx awel dev
 ```
 
@@ -25,10 +33,11 @@ Awel needs at least one configured provider to function. See [Supported Models](
 
 This starts Awel on port 3001 and proxies your Next.js dev server on port 3000. Open `http://localhost:3001` to see your app with the Awel overlay.
 
-### Options
+### Commands
 
 ```
-awel dev [options]
+awel create              Create a new Next.js project with Awel
+awel dev [options]       Start the development server with Awel overlay
 
   -p, --port <port>    Target app port (default: 3000)
   -v, --verbose        Print LLM stream events to stderr
@@ -65,6 +74,13 @@ Awel uses the [Vercel AI SDK](https://sdk.vercel.ai) and supports multiple provi
 
 Switch models at any time from the dropdown in the dashboard header.
 
+### Additional Environment Variables
+
+| Variable | Description |
+|----------|-------------|
+| `OPENAI_BASE_URL` | Custom base URL for the OpenAI provider (e.g. a proxy or compatible API). Defaults to `https://api.openai.com/v1`. |
+| `AWEL_MAX_OUTPUT_TOKENS` | Maximum number of tokens the model can generate per response. Applies to all providers. |
+
 ## Agent Tools
 
 The AI agent has access to:
@@ -94,6 +110,7 @@ The AI agent has access to:
 - **Diff review** — review a summary of all file changes before accepting
 - **Dark mode** — follows your system preference
 - **i18n** — English and Chinese
+- **Creation mode** — `awel create` scaffolds a new project and launches a full-page AI chat where you describe your app and the agent builds it from scratch
 
 ## Development
 

package/README.zh-CN.md

@@ -8,16 +8,24 @@
 
 ## 快速开始
 
-```bash
-# 如果你已经在一个 Next.js 项目中，可以跳过这一步
-npx create-next-app@latest my-app && cd my-app
+### 创建新项目
 
+```bash
 # 至少配置一个 AI 服务商（任选其一）：
 export ANTHROPIC_API_KEY="sk-ant-..."   # Anthropic API
 export OPENAI_API_KEY="sk-..."          # OpenAI
 export GOOGLE_GENERATIVE_AI_API_KEY="..." # Google AI
 # 或安装 Claude CLI：https://docs.anthropic.com/en/docs/claude-code
 
+npx awel create
+```
+
+此命令会创建一个新的 Next.js 项目并标记为创建模式。按照提示 `cd` 进入项目目录并运行 `npx awel dev`。你会看到一个全屏创建界面，描述你想构建的应用——AI 智能体会为你生成整个应用，完成后自动切换到正常的 Awel 浮层模式。
+
+### 在已有项目中使用
+
+```bash
+cd my-existing-next-app
 npx awel dev
 ```
 
@@ -25,10 +33,11 @@ Awel 需要至少一个已配置的服务商才能运行。完整列表见[支
 
 Awel 会在端口 3001 启动，并代理运行在端口 3000 的 Next.js 开发服务器。打开 `http://localhost:3001` 即可看到带有 Awel 浮层的应用。
 
-### 选项
+### 命令
 
 ```
-awel dev [options]
+awel create              创建一个带有 Awel 的新 Next.js 项目
+awel dev [options]       启动带有 Awel 浮层的开发服务器
 
   -p, --port <port>    目标应用端口（默认：3000）
   -v, --verbose        将 LLM 流式事件输出到 stderr
@@ -65,6 +74,13 @@ Awel 使用 [Vercel AI SDK](https://sdk.vercel.ai)，支持多个服务商。设
 
 可随时在面板顶部的下拉菜单中切换模型。
 
+### 额外环境变量
+
+| 变量 | 说明 |
+|------|------|
+| `OPENAI_BASE_URL` | OpenAI 服务商的自定义 Base URL（例如代理或兼容 API）。默认为 `https://api.openai.com/v1`。 |
+| `AWEL_MAX_OUTPUT_TOKENS` | 模型单次响应的最大生成 token 数。对所有服务商生效。 |
+
 ## 智能体工具
 
 AI 智能体可使用以下工具：
@@ -94,6 +110,7 @@ AI 智能体可使用以下工具：
 - **Diff 审查** — 在接受变更前查看所有文件修改的摘要
 - **深色模式** — 跟随系统偏好
 - **国际化** — 支持英文和中文
+- **创建模式** — `awel create` 创建新项目并启动全屏 AI 对话界面，描述你的应用，智能体从零开始为你构建
 
 ## 开发
 

package/dist/cli/onboarding.js

@@ -1,25 +1,50 @@
 import { readAwelConfig, writeAwelConfig } from './awel-config.js';
 import { getAvailableProviders, PROVIDER_ENV_KEYS, PROVIDER_LABELS } from './providers/registry.js';
 import { awel } from './logger.js';
-function printSetupInstructions() {
-    awel.log('No LLM providers are configured.');
-    awel.log('Awel needs at least one AI provider to function.');
-    awel.log('');
-    awel.log('Set up a provider by exporting its API key:');
-    awel.log('');
-    // List all providers with their setup commands
-    for (const [provider, envKey] of Object.entries(PROVIDER_ENV_KEYS)) {
+async function promptProviderSetup() {
+    const p = await import('@clack/prompts');
+    p.log.warn('No LLM providers are configured.');
+    p.log.message('Awel needs at least one AI provider to function.\n');
+    const provider = await p.select({
+        message: 'Which provider would you like to set up?',
+        options: [
+            { value: 'claude-code', label: 'Claude Code', hint: 'Uses Claude CLI binary' },
+            { value: 'anthropic', label: 'Anthropic API', hint: 'ANTHROPIC_API_KEY' },
+            { value: 'openai', label: 'OpenAI', hint: 'OPENAI_API_KEY' },
+            { value: 'google-ai', label: 'Google AI', hint: 'GOOGLE_GENERATIVE_AI_API_KEY' },
+            { value: 'vercel-gateway', label: 'Vercel AI Gateway', hint: 'AI_GATEWAY_API_KEY' },
+            { value: 'qwen', label: 'Qwen', hint: 'DASHSCOPE_API_KEY' },
+            { value: 'minimax', label: 'MiniMax', hint: 'MINIMAX_API_KEY' },
+        ],
+    });
+    if (p.isCancel(provider)) {
+        p.cancel('Cancelled');
+        process.exit(0);
+    }
+    if (provider === 'claude-code') {
+        p.note('Install the Claude CLI:\n\n' +
+            '  npm install -g @anthropic-ai/claude-code\n\n' +
+            'Then authenticate:\n\n' +
+            '  claude login', 'Claude Code Setup');
+    }
+    else if (provider === 'openai') {
+        const envKey = PROVIDER_ENV_KEYS[provider];
+        p.note(`Export your API key:\n\n` +
+            `  export ${envKey}="your-api-key"\n\n` +
+            `Or add it to your .env file:\n\n` +
+            `  ${envKey}=your-api-key\n\n` +
+            `To use a custom base URL (e.g. a proxy or compatible API):\n\n` +
+            `  export OPENAI_BASE_URL="https://your-proxy.com/v1"`, 'OpenAI Setup');
+    }
+    else {
+        const envKey = PROVIDER_ENV_KEYS[provider];
         const label = PROVIDER_LABELS[provider] ?? provider;
-        awel.log(`  ${label}`);
-        awel.log(`  export ${envKey}="..."`);
-        awel.log('');
+        p.note(`Export your API key:\n\n` +
+            `  export ${envKey}="your-api-key"\n\n` +
+            `Or add it to your .env file:\n\n` +
+            `  ${envKey}=your-api-key`, `${label} Setup`);
     }
-    // Claude Code is special — no env var, needs CLI install
-    const label = PROVIDER_LABELS['claude-code'] ?? 'Claude Code';
-    awel.log(`  ${label}`);
-    awel.log('  Install the Claude CLI: https://docs.anthropic.com/en/docs/claude-code');
-    awel.log('');
-    awel.log('Then run `awel dev` again.');
+    p.log.message('Then run `awel dev` again.');
 }
 export async function ensureProvider(projectCwd) {
     const config = readAwelConfig(projectCwd);
@@ -41,17 +66,17 @@ export async function ensureProvider(projectCwd) {
         return;
     }
     if (isFirstRun && available.length === 0) {
-        // First run with NO providers — show welcome + instructions, exit
+        // First run with NO providers — show welcome + interactive setup, exit
         awel.log('');
         awel.log('Welcome to Awel!');
         awel.log('AI-powered development overlay for Next.js');
         awel.log('');
-        printSetupInstructions();
+        await promptProviderSetup();
         process.exit(1);
     }
     if (!isFirstRun && available.length === 0) {
-        // Subsequent run with NO providers — instructions only, exit
-        printSetupInstructions();
+        // Subsequent run with NO providers — interactive setup, exit
+        await promptProviderSetup();
         process.exit(1);
     }
     // Subsequent run with providers available — silent pass-through

package/dist/cli/providers/vercel.js

@@ -149,7 +149,9 @@ function createModel(modelId, providerType, cwd) {
         return anthropic(modelId);
     }
     else if (providerType === 'openai') {
-        const openai = createOpenAI({});
+        const openai = createOpenAI({
+            baseURL: process.env.OPENAI_BASE_URL,
+        });
         return openai(modelId);
     }
     else if (providerType === 'google-ai') {
@@ -230,12 +232,16 @@ export function createVercelProvider(modelId, providerType) {
                 const systemPrompt = isSelfContained
                     ? undefined
                     : `${basePrompt}\n\nThe user's project directory is: ${config.projectCwd}`;
+                const maxOutputTokens = process.env.AWEL_MAX_OUTPUT_TOKENS
+                    ? parseInt(process.env.AWEL_MAX_OUTPUT_TOKENS, 10)
+                    : undefined;
                 const streamTextArgs = {
                     model,
                     ...(systemPrompt && { system: systemPrompt }),
                     messages,
                     tools,
                     ...(!isSelfContained && { stopWhen: stepCountIs(25) }),
+                    ...(maxOutputTokens && { maxOutputTokens }),
                     abortSignal: abortController.signal,
                 };
                 logEvent('stream:start', `model=${modelId} provider=${providerType} messages=${messages.length}`);

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "awel",
-    "version": "0.2.0",
+    "version": "0.2.1",
     "private": false,
     "type": "module",
     "bin": {