Continue - VS Code AI Programming Assistant Complete Configuration

Open-source AI code assistant, supporting code completion, chat, editing, compatible with all mainstream LLMs, completely free

Smart Completion

Tab key triggers code completion

Multi-Model Support

GPT, Claude, Local Models

Context Awareness

Understands project code structure

Highly Customizable

Custom commands and providers

1. Installation and Initial Configuration

1. Install Plugin

  1. Open VS Code
  2. Go to Extensions Store (Ctrl/Cmd + Shift + X)
  3. Search for "Continue"
  4. Click Install on the plugin published by Continue
  5. Restart VS Code

2. Shortcuts

  • Ctrl/Cmd + L - Open chat panel
  • Ctrl/Cmd + I - Edit selected code
  • Tab - Accept code completion
  • Ctrl/Cmd + K - GenerateCode

2. Configuration File Details

config.json Complete Configuration

{
  "models": [
    {
      "title": "GPT-4o",
      "provider": "openai",
      "model": "gpt-4o",
      "apiKey": "your-api-key",
      "apiBase": "https://api.n1n.ai/v1"
    },
    {
      "title": "Claude 3.5 Sonnet",
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "apiKey": "your-api-key",
      "apiBase": "https://api.n1n.ai/v1"
    },
    {
      "title": "DeepSeek Coder",
      "provider": "openai",
      "model": "deepseek-coder",
      "apiKey": "your-api-key",
      "apiBase": "https://api.deepseek.com/v1"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Codestral",
    "provider": "mistral",
    "model": "codestral-latest",
    "apiKey": "your-api-key"
  },
  "embeddingsProvider": {
    "provider": "openai",
    "model": "text-embedding-3-small",
    "apiKey": "your-api-key",
    "apiBase": "https://api.n1n.ai/v1"
  },
  "contextProviders": [
    {
      "name": "code",
      "params": {}
    },
    {
      "name": "docs",
      "params": {}
    },
    {
      "name": "terminal",
      "params": {}
    },
    {
      "name": "problems",
      "params": {}
    }
  ],
  "slashCommands": [
    {
      "name": "edit",
      "description": "Edit selected code"
    },
    {
      "name": "comment",
      "description": "Add comments to code"
    },
    {
      "name": "test",
      "description": "Generate unit tests"
    },
    {
      "name": "fix",
      "description": "Fix problems in code"
    }
  ]
}

Chat Models

Configure multiple models for selection, supports quick switching

Code Completion

Use dedicated code models like Codestral

Embedding Models

Used for code search and context understanding

3. Code Auto-completion Optimization

VS Code Settings

// VS Code settings.json Configure
{
  // Continue code completion settings
  "continue.enableTabAutocomplete": true,
  "continue.tabAutocompleteOptions": {
    "multilineCompletions": "always",
    "maxPromptTokens": 1500,
    "debounceDelay": 350,
    "maxSuffixPercentage": 0.4,
    "prefixPercentage": 0.85,
    "template": "Please complete the following code:\n{{{prefix}}}[BLANK]{{{suffix}}}\n\nFill in the [BLANK]"
  },
  
  // Trigger conditions
  "continue.tabAutocompleteOptions.triggers": [
    {
      "language": "python",
      "triggerWords": ["def", "class", "if", "for", "while"]
    },
    {
      "language": "javascript",
      "triggerWords": ["function", "const", "let", "if", "for"]
    }
  ],
  
  // Disable auto-completion for certain files
  "continue.tabAutocompleteOptions.disableInFiles": [
    "*.md",
    "*.txt",
    "package-lock.json"
  ]
}

💡 Performance Optimization Tips

  • • Use faster models like Codestral or DeepSeek Coder
  • • Adjust debounceDelay to reduce latency
  • • Limit maxPromptTokens to improve response speed
  • • Disable auto-completion for large files

4. Custom Context Providers

Create Custom Providers

// Custom context provider example
// ~/.continue/contextProviders/myCustomProvider.js

module.exports = {
  title: "Project Documentation",
  description: "Include project-related documentation",
  
  async getContext(query, extras) {
    const docs = await loadProjectDocs();
    
    return docs.map(doc => ({
      name: doc.title,
      description: doc.summary,
      content: doc.content
    }));
  }
};

// Register in config.json
{
  "contextProviders": [
    {
      "name": "myCustomProvider",
      "params": {
        "docsPath": "./docs"
      }
    }
  ]
}

Built-in Providers

  • • code: Current code file
  • • docs: Project documentation
  • • terminal: Terminal output
  • • problems: Errors and warnings
  • • git: Git history and diffs

Use Cases

  • • Include project-specific documentation
  • • Integrate external API documentation
  • • Add team coding standards
  • • including TestData
  • • Reference database schemas

5. Custom Slash Commands

Create Refactor Command

// Custom slash command
// ~/.continue/slashCommands/refactor.js

module.exports = {
  name: "refactor",
  description: "Refactor selected code",
  
  async run(sdk) {
    const selection = await sdk.getSelectedCode();
    
    const prompt = `Please refactor the following code to make it clearer and more efficient: 
    
${selection}

Requirements: 
1. Extract duplicate code into functions
2. Improve variable naming
3. Add necessary type annotations
4. Optimize performance`;
    
    const result = await sdk.llm.complete(prompt);
    await sdk.applyEdit(result);
  }
};

Common Command Ideas

/optimize - Optimize performance
/security - Security check
/docs - GenerateDocumentation
/convert - Convert code
/review - Code review
/clean - Clean code

6. Best Practices and Tips

🎯 Improve Efficiency

  • ✅ Use @ symbol to quickly reference files
  • ✅ Configure multiple models, switch based on task
  • ✅ Create project-specific slash commands
  • ✅ Use context providers to include relevant documentation
  • ✅ Set appropriate token limits

âš¡ Performance Optimization

  • ✅ Run Ollama locally to reduce latency
  • ✅ Use smaller models for completion
  • ✅ Limit context window size
  • ✅ Disable unnecessary context providers
  • ✅ Adjust debounce delay parameter

7. FAQ Solutions

Code completion not working?

1. Check if tabAutocompleteModel configuration is correct

2. Confirm API Key is valid

3. Check error messages in output panel

4. Try restarting VS Code

How to use local models?
// Configure Ollama local models
{
  "models": [{
    "title": "Local Llama",
    "provider": "ollama",
    "model": "llama2:13b"
  }]
}
How to reduce API costs?

• Use cheaper models like GPT-3.5-turbo

• Reduce maxTokens limit

• Enable caching feature

• Use local models for simple tasks

Cursor Configuration Tutorial

Another powerful AI programming assistant configuration guide

Aider Configuration Tutorial

Command-line AI pair programming tool