Add test coverage for AI provider implementations

[Copilot]

The issue reported incomplete AI provider implementations. Investigation shows all 5 providers (Groq, OpenRouter, HuggingFace, Google Gemini, Ollama) and interactive mode functions are fully implemented with proper error handling.

Changes

• Test suite (test/smoke-test.js): 21 tests validating provider implementations, error handling, and model routing
• Test script (package.json): Added npm test
• Dependencies (package-lock.json): Locked 90 packages

Test Coverage

// Validates all providers exist and handle missing API keys
await asyncTest('sendToGroq throws error without API key', async () => {
  const provider = new AIProvider('groq-llama-70b');
  try {
    await provider.sendMessage([{ role: 'user', content: 'test' }]);
    throw new Error('Should have thrown error');
  } catch (error) {
    if (!error.message.includes('API-Key')) {
      throw new Error('Expected API key error');
    }
  }
});

Tests use named constants and thresholds rather than exact values for resilience to model additions/removals.

Security

CodeQL scan: 0 vulnerabilities detected.

Original prompt

🐛 Problems Found in copilot-cli

The copilot-cli project has several critical incomplete implementations:

1. Incomplete AI Provider Implementation

File: src/ai-provider.js

• Line 50: Function sendMessage() is incomplete - cuts off mid-implementation
• Missing implementations for all AI providers:
  • Groq (sendToGroq method)
  • OpenRouter
  • Hugging Face
  • Google Gemini
  • Ollama

2. Incomplete Interactive Mode

File: src/interactive.js

• Line 120: Function showConfig() is incomplete
• Missing function: handleAIRequest() - called on line 49 but never defined

3. Missing Error Handling

Throughout the codebase:

• No try-catch blocks for API calls
• No validation for API keys
• No graceful error messages for users

4. Missing Functions

Several functions are referenced but not implemented:

• handleAIRequest() in interactive.js
• sendToGroq() in ai-provider.js
• sendToOpenRouter() in ai-provider.js
• sendToHuggingFace() in ai-provider.js
• sendToGoogle() in ai-provider.js
• sendToOllama() in ai-provider.js

5. Subprojects Not Integrated

Three separate subprojects exist but aren't integrated:

• desktop/ - Electron desktop app
• omni-control/ - Control suite
• universal/ - Universal components

✅ Required Fixes

Priority 1: Complete Core Functionality

1. Complete src/ai-provider.js:

   • Implement sendMessage() fully
   • Implement all provider methods:
     • sendToGroq() - Use Groq API
     • sendToOpenRouter() - Use OpenRouter API
     • sendToHuggingFace() - Use HuggingFace Inference API
     • sendToGoogle() - Use Google Gemini API
     • sendToOllama() - Use Ollama local API
   • Add proper error handling for each provider
   • Add API key validation

2. Complete src/interactive.js:

   • Implement handleAIRequest() function
   • Complete showConfig() function
   • Add spinner/loading indicators for API calls
   • Add proper error messages for users

3. Add Error Handling:

   • Try-catch blocks around all API calls
   • User-friendly error messages
   • Fallback behavior when API fails
   • API key validation on startup

Priority 2: Code Quality

4. Add Input Validation:

   • Validate API keys before making calls
   • Check for empty responses
   • Handle rate limits gracefully

5. Add Missing Documentation:

   • JSDoc comments for all functions
   • Update README with complete setup instructions
   • Add examples for each AI provider

Priority 3: Integration (Optional)

6. Integrate Subprojects (if time permits):
   • Link desktop app to CLI
   • Document omni-control and universal components

📋 Implementation Details

For sendToGroq():

async sendToGroq(conversationHistory) {
  if (!this.config.groqApiKey) {
    throw new Error('Groq API key not configured');
  }
  
  const response = await fetch('https://api.groq.com/openai/v1/chat/completions', {
    method: 'POST',
    headers: {
      'Authorization': `Bearer ${this.config.groqApiKey}`,
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      model: this.modelName.replace('groq-', ''),
      messages: conversationHistory,
      temperature: 0.7
    })
  });
  
  // Handle response and errors
}

Similar implementation needed for other providers.

For handleAIRequest():

async function handleAIRequest(input) {
  conversationHistory.push({ role: 'user', content: input });
  
  const spinner = ora('Thinking...').start();
  
  try {
    const response = await aiProvider.sendMessage(conversationHistory);
    spinner.stop();
    
    conversationHistory.push({ role: 'assistant', content: response });
    console.log(chalk.green('\nAssistant: ') + response + '\n');
  } catch (error) {
    spinner.stop();
    console.log(chalk.red(`\n❌ Error: ${error.message}\n`));
  }
}

🎯 Expected Outcome

A fully functional CLI tool that:

• ✅ Works with all advertised AI providers
• ✅ Has proper error handling
• ✅ Validates API keys on startup
• ✅ Provides clear user feedback
• ✅ Can be installed globally with npm install -g
• ✅ Has complete implementation (no hanging functions)

🧪 Testing Requirements

After fixes, the tool should:

1. Start without errors: npm start
2. Accept and respond to user input
3. Handle missing API keys gracefully
4. Support model switching with /model command
5. Work with at least Groq provider (most accessible free option)

This pull request was created from Copilot chat.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.