adding-a-command

分类: 数据与AI | 上传者: rely-ai-org | 下载: 0 | 版本: v1.0（最新）

How to add a new CLI command to @rely-ai/caliber. Use when creating new Commander.js subcommands, wiring up LLM calls, using spinners/chalk, or writing command tests.

更新日志: Source: GitHub https://github.com/rely-ai-org/caliber

目录结构

当前层级: .claude/skills/adding-a-command/

📄 SKILL.md 2.6 KB

SKILL.md

---
name: adding-a-command
description: How to add a new CLI command to @rely-ai/caliber. Use when creating new Commander.js subcommands, wiring up LLM calls, using spinners/chalk, or writing command tests.
---
# Adding a New CLI Command

Follow this checklist when adding a new command to `@rely-ai/caliber`.

## 1. Create the command file

Create `src/commands/<name>.ts`. Export a single async function:

```typescript
import chalk from 'chalk';
import ora from 'ora';
import { loadConfig } from '../llm/config.js';

export async function myNewCommand(options: { dryRun?: boolean }) {
  const config = loadConfig();
  if (!config) {
    console.log(chalk.red('No LLM provider configured. Run `caliber config` or set ANTHROPIC_API_KEY.'));
    throw new Error('__exit__');
  }

  const spinner = ora('Working...').start();
  try {
    // ... implementation
    spinner.succeed(chalk.green('Done!'));
  } catch (err) {
    spinner.fail(chalk.red('Something went wrong'));
    throw err;
  }
}
```

**Key conventions:**
- Import paths must use `.js` extension: `'../llm/config.js'`
- Check `loadConfig()` for commands that require an LLM provider
- Use `ora` for spinners on async operations
- Use `chalk` for coloured output (red = error, green = success, yellow = warning)
- Throw `new Error('__exit__')` for clean exits (no stack trace)

## 2. Register in cli.ts

```typescript
import { myNewCommand } from './commands/my-new-command.js';

program
  .command('my-new-command')
  .description('What this command does')
  .option('--dry-run', 'Preview changes without writing')
  .action(myNewCommand);
```

## 3. Use the LLM layer (if needed)

```typescript
import { llmCall, llmJsonCall } from '../llm/index.js';

// Plain text response
const text = await llmCall({ system: 'You are...', prompt: 'Analyze...' });

// Parsed JSON response
const result = await llmJsonCall<MyType>({ system: '...', prompt: '...' });
```

See the `llm-provider` skill for streaming patterns and adding new providers.

## 4. Write a test

Create `src/commands/__tests__/my-new-command.test.ts`. The LLM provider is globally mocked in `src/test/setup.ts`.

```typescript
import { describe, it, expect, vi, beforeEach } from 'vitest';
import { myNewCommand } from '../my-new-command.js';

describe('myNewCommand', () => {
  beforeEach(() => vi.clearAllMocks());

  it('throws __exit__ when no config', async () => {
    await expect(myNewCommand({})).rejects.toThrow('__exit__');
  });
});
```

## 5. Update README.md

Add a row to the Commands table in `README.md`.

## 6. Commit

```bash
git commit -m "feat: add my-new-command command"
```

登录后下载/点赞/收藏 ❤ 4 | ★ 0

adding-a-command

目录结构

SKILL.md

举报内容

提示