Add AI-powered application generation with business logic and testing to ObjectQL

[Copilot]

AI Agent and CLI for Enterprise Application Creation and Validation

Completed

• 1. Moved AI functionality to core package

  • Created @objectql/core/src/ai-agent.ts with ObjectQLAgent class
  • Updated package dependencies (openai, js-yaml, @types/js-yaml)
  • Exported AI agent from core package index
  • Updated CLI to import from @objectql/core
  • Removed duplicate agent.ts from CLI package
  • Fixed pnpm-lock.yaml to include new dependencies

• 2. Improved AI command structure with interactive mode as default

  • objectql ai - Default interactive conversational mode
  • objectql ai generate - One-shot generation subcommand
  • objectql ai validate - Validation subcommand
  • objectql ai chat - Chat assistant subcommand

• 3. Complete metadata and business logic generation

  • Generate all ObjectQL metadata types
  • TypeScript implementations for actions and hooks
  • Jest test files for business logic validation

• 4. AI-powered validation

  • Deep validation beyond syntax
  • Business logic consistency checks
  • Graceful fallback without API key

• 5. Interactive features

  • Step-by-step conversational generation (default mode)
  • AI chat assistant for questions
  • Continuous dialogue for file modification

• 6. Comprehensive documentation

  • Created /docs/ai/cli-usage.md - Complete CLI guide
  • Created /docs/ai/programmatic-api.md - Programmatic API documentation
  • Updated /docs/ai/index.md - Main AI documentation index
  • Included TypeScript types and examples
  • Added best practices and workflow examples

• 7. CI/Build fixes

  • Updated pnpm-lock.yaml with new dependencies
  • Verified builds pass for @objectql/core and @objectql/cli

Summary

CI Build Fix:
Updated pnpm-lock.yaml to include the new dependencies added to @objectql/core:

• openai@^4.28.0 - OpenAI API client
• js-yaml@^4.1.0 - YAML parsing for metadata
• @types/js-yaml@^4.0.5 - TypeScript types for js-yaml

Architecture:

Moved AI functionality from CLI to core package enabling:

• ✅ Reusability: API server can now use the same AI agent
• ✅ Separation of Concerns: Core logic in @objectql/core, CLI is thin wrapper
• ✅ Programmatic Access: Direct usage via new ObjectQLAgent()
• ✅ Better Testing: Core functionality testable independently

Package Structure:

@objectql/core
├── src/
│   ├── ai-agent.ts        # NEW: AI Agent class
│   ├── validator.ts
│   ├── app.ts
│   └── index.ts           # Exports ObjectQLAgent
└── package.json           # Added openai, js-yaml dependencies

@objectql/cli
└── src/
    ├── commands/
    │   └── ai.ts          # UPDATED: Imports from @objectql/core
    └── index.ts

API Usage:

// In any Node.js application
import { ObjectQLAgent } from '@objectql/core';

const agent = new ObjectQLAgent({ apiKey: process.env.OPENAI_API_KEY! });

const result = await agent.generateApp({
  description: "CRM system",
  type: 'complete'
});

CLI Usage:

# Interactive (default - easiest)
objectql ai

# One-shot generation
objectql ai generate -d "description"

# Validation
objectql ai validate ./src

# Chat
objectql ai chat

Documentation:

• 📚 Comprehensive CLI usage guide with examples
• 📚 Programmatic API documentation with TypeScript types
• 📚 Advanced examples (CI/CD, web UI, automation)
• 📚 Best practices and error handling
• 📚 Updated main AI documentation index

The implementation enables the API server and other packages to use the same AI capabilities while maintaining the user-friendly CLI interface. All builds now pass successfully.

Original prompt

编写ai agent和cli，使用ai按照元数据规范创建和验证企业应用

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[hotlong]

@copilot 生成的企业应用元数据范围应该包含规范文档中的所有元数据，而不只是元数据，你应该开发一个agent分步生成，你应该具有对话能力，客户提出改进的需求，你针对性地做改进

[hotlong]

@copilot ai生成的代码项目应该包括业务逻辑部分，actions，hooks等等。应该提供测试方案验证app的业务逻辑

[hotlong]

@copilot 继续完成所有工作

[hotlong]

@copilot 继续完成所有工作

你在继续工作吗？如果任务复杂可以重新确立任务规划，分步完成。 @copilot

[hotlong]

@copilot 继续完成所有工作

你在继续工作吗？如果任务复杂可以重新确立任务规划，分步完成。 @copilot

[Copilot]

@copilot 继续完成所有工作

你在继续工作吗？如果任务复杂可以重新确立任务规划，分步完成。 @copilot

I'm continuing the work now. Based on the previous feedback, I'm implementing:

1. Business logic generation: Generate TypeScript implementation files for actions and hooks (not just YAML)
2. Testing framework: Add test generation for validating business logic
3. Step-by-step conversational agent: Already added objectql ai conversational command

Working on these enhancements now.

[Copilot]

@copilot 继续完成所有工作

你在继续工作吗？如果任务复杂可以重新确立任务规划，分步完成。 @copilot

I'm continuing the work now. I'll complete the remaining requirements:

1. Generate business logic (actions/hooks with TypeScript implementations)
2. Add testing validation for business logic

Let me proceed with the implementation.

[Copilot]

@copilot 现在相关的命令太多，简化合并一下

Simplified in commit a440f5c. Merged 4 separate commands into a unified interface:

Before:

• objectql ai generate
• objectql ai validate
• objectql ai chat
• objectql ai conversational

After (single command with modes):

• objectql ai -d "description" [dir] - Generate
• objectql ai --validate [path] - Validate
• objectql ai --chat - Chat assistant
• objectql ai --interactive [dir] - Interactive generation

Benefits: Simpler to use, fewer commands to remember, consistent flag-based interface, built-in help when no mode specified.

[hotlong]

你这个改动太极端了，没有意义，没有减少学习成本。觉得最常见的场景是交互式会话，要让用户很方便的输入这个命令。

[hotlong]

你这个改动太极端了，没有意义，没有减少学习成本。觉得最常见的场景是交互式会话，要让用户很方便的输入这个命令。

@copilot

[Copilot]

你这个改动太极端了，没有意义，没有减少学习成本。觉得最常见的场景是交互式会话，要让用户很方便的输入这个命令。

@copilot

Redesigned in commit 89616e8. The interactive mode is now the default and easiest to use:

Just type:

objectql ai

This immediately starts the conversational builder. Other modes use clear subcommands:

• objectql ai generate -d "..." - One-shot generation
• objectql ai validate ./src - Validation
• objectql ai chat - Q&A assistant

The most common scenario (interactive conversation) now requires only 2 words, making it very convenient while keeping other modes clear and discoverable.

[hotlong]

ai相关的功能考虑到放到内核中，因为下一步API也要实现。cli只是调用内核。然后还要帮我更新docs相关的文档。

[hotlong]

ai相关的功能考虑到放到内核中，因为下一步API也要实现。cli只是调用内核。然后还要帮我更新docs相关的文档。

@copilot

[Copilot]

ai相关的功能考虑到放到内核中，因为下一步API也要实现。cli只是调用内核。然后还要帮我更新docs相关的文档。

@copilot

Completed in commit 7d07a31:

Moved to Core:

• AI agent is now in @objectql/core/src/ai-agent.ts
• CLI imports from @objectql/core instead of local file
• API server can now use the same ObjectQLAgent class

Usage:

// In any package
import { ObjectQLAgent } from '@objectql/core';

const agent = new ObjectQLAgent({ apiKey: '...' });
const result = await agent.generateApp({ description: '...' });

Documentation:

• Created /docs/ai/cli-usage.md - Complete CLI guide with examples
• Created /docs/ai/programmatic-api.md - API documentation with TypeScript types
• Updated /docs/ai/index.md - Main AI docs index

[hotlong]

2026-01-13T03:18:13.4617348Z ##[group]Run pnpm install --frozen-lockfile
2026-01-13T03:18:13.4617765Z �[36;1mpnpm install --frozen-lockfile�[0m
2026-01-13T03:18:13.4661339Z shell: /usr/bin/bash -e {0}
2026-01-13T03:18:13.4661613Z env:
2026-01-13T03:18:13.4661867Z PNPM_HOME: /home/runner/setup-pnpm/node_modules/.bin
2026-01-13T03:18:13.4662176Z ##[endgroup]
2026-01-13T03:18:13.8531400Z Scope: all 18 workspace projects
2026-01-13T03:18:13.9438824Z  ERR_PNPM_OUTDATED_LOCKFILE  Cannot install with "frozen-lockfile" because pnpm-lock.yaml is not up to date with /packages/foundation/core/package.json
2026-01-13T03:18:13.9439654Z
2026-01-13T03:18:13.9440459Z Note that in CI environments this setting is true by default. If you still need to run install in such cases, use "pnpm install --no-frozen-lockfile"
2026-01-13T03:18:13.9441137Z
2026-01-13T03:18:13.9441256Z Failure reason:
2026-01-13T03:18:13.9441733Z specifiers in the lockfile don't match specifiers in package.json:
2026-01-13T03:18:13.9442507Z * 3 dependencies were added: @types/js-yaml@^4.0.5, openai@^4.28.0, js-yaml@^4.1.0
2026-01-13T03:18:13.9443274Z
2026-01-13T03:18:13.9652746Z ##[error]Process completed with exit code 1.
@copilot

[Copilot]

2026-01-13T03:18:13.4617348Z ##[group]Run pnpm install --frozen-lockfile
2026-01-13T03:18:13.4617765Z �[36;1mpnpm install --frozen-lockfile�[0m
2026-01-13T03:18:13.4661339Z shell: /usr/bin/bash -e {0}...

Fixed in commit 764696b. Updated pnpm-lock.yaml to include the new dependencies (openai, js-yaml, @types/js-yaml) that were added to @objectql/core. The build now passes successfully.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.