How to Set Up Continue: A Practical Guide

TL;DR

• Install Continue in VS Code or JetBrains in 2 minutes via marketplace
• Connect to OpenAI, Anthropic, or Ollama (local) in config.json
• Use tab autocomplete and chat for code generation/editing
• Customize with slash commands (e.g., /explain, /test) and context providers (@file, @docs)
• Enterprise? Enable SSO, audit logs, and self-hosting for compliance

1. Installation

VS Code

1. Open VS Code.
2. Press Ctrl+Shift+X (Windows/Linux) or Cmd+Shift+X (Mac) to open Extensions.
3. Search for Continue and click Install.
   • Expected: Sidebar icon appears (rocket emoji).
4. Verify:

   code --list-extensions | grep continue
   # Output: continue.continue
   

   Copy

JetBrains (IntelliJ, PyCharm, etc.)

1. Open Settings (Ctrl+Alt+S / Cmd+,).
2. Navigate to Plugins > Marketplace.
3. Search for Continue and click Install.
4. Restart the IDE.
   • Expected: Continue panel appears in the right sidebar.

Gotcha: On Windows, ensure WSL2 is installed if using local models:

wsl --install

2. Configure Models

Edit ~/.continue/config.json (auto-created on first launch). Here’s a production-ready template:

{
  "models": [
    {
      "title": "OpenAI GPT-4 Turbo",
      "provider": "openai",
      "model": "gpt-4-turbo",
      "apiKey": "${OPENAI_API_KEY}"
    },
    {
      "title": "Anthropic Claude 3 Opus",
      "provider": "anthropic",
      "model": "claude-3-opus-20240229",
      "apiKey": "${ANTHROPIC_API_KEY}"
    },
    {
      "title": "Local Ollama (Llama3)",
      "provider": "ollama",
      "model": "llama3"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Local Ollama (CodeLlama)",
    "provider": "ollama",
    "model": "codellama"
  }
}

Key Fields:

• apiKey: Use ${ENV_VAR} for security (set in .bashrc/~/.zshrc).
• tabAutocompleteModel: Separate model for faster completions.

Verify Models:

1. Open Continue sidebar.
2. Click the model dropdown (top-right).
   • Expected: All configured models appear.

Common Errors:

• 401 Unauthorized: Check API keys (echo $OPENAI_API_KEY).
• Ollama not found: Install Ollama first:

  curl -fsSL https://ollama.com/install.sh | sh
  ollama pull llama3
  

  Copy

3. Tab Autocomplete Setup

Enable in config.json:

{
  "tabAutocompleteEnabled": true,
  "tabAutocompleteModel": {
    "title": "Ollama CodeLlama",
    "provider": "ollama",
    "model": "codellama"
  }
}

Test It:

1. Open a Python file.
2. Type def hello_world( and press Tab.
   • Expected: Continue suggests a function body.

Pro Tip: For faster completions, use smaller models (e.g., starcoder:1b).

4. Chat and Inline Editing

Chat Interface

1. Open Continue sidebar (Ctrl+Shift+L / Cmd+Shift+L).
2. Type a prompt (e.g., "Explain this React component").
   • Expected: Streaming response with code snippets.

Inline Editing:

1. Highlight code.
2. Press Cmd+I (Mac) / Ctrl+I (Windows/Linux).
3. Type an instruction (e.g., "Add error handling").
   • Expected: Code updates in-place.

[AI Security](https://hyperion-<a href="/services/coaching-vs-consulting">consulting</a>.io/services/cybersecurity-for-ai) Posture Framework™ (DETECT Phase):

• Enable audit logs in config.json to track LLM interactions:

  {
    "auditLogEnabled": true,
    "auditLogPath": "/var/log/continue/audit.log"
  }
  

  Copy
  • Logs include: timestamps, user IDs, and model inputs/outputs.

5. Context Providers

Use @ symbols to reference context in chat:

Example:

@file src/api.ts
Refactor this to use async/await instead of callbacks.

Gotcha: @codebase requires Pro/Enterprise for large projects.

6. Custom Slash Commands

Define reusable commands in config.json:

{
  "customCommands": [
    {
      "name": "explain",
      "prompt": "Explain the following code in simple terms:\n{{code}}",
      "description": "Explain selected code"
    },
    {
      "name": "test",
      "prompt": "Write unit tests for:\n{{code}}\n\nUse Jest for JavaScript or pytest for Python.",
      "description": "Generate tests"
    }
  ]
}

Usage:

1. Highlight code.
2. Type /explain in the chat.
   • Expected: Plain-English explanation.

AI Security Posture Framework™ (PROTECT Phase):

• Restrict commands via role-based access control (Enterprise):

  {
    "enterprise": {
      "allowedCommands": ["explain", "test"],
      "blockedCommands": ["delete", "exec"]
    }
  }
  

  Copy

7. config.json Deep Dive

Full Enterprise Example

{
  "models": [
    {
      "title": "Azure OpenAI GPT-4",
      "provider": "openai",
      "model": "gpt-4",
      "apiBase": "https://your-azure-endpoint.openai.azure.com",
      "apiKey": "${AZURE_OPENAI_KEY}",
      "apiVersion": "2024-02-15-preview"
    }
  ],
  "tabAutocompleteEnabled": true,
  "tabAutocompleteModel": {
    "title": "Ollama DeepSeek Coder",
    "provider": "ollama",
    "model": "deepseek-coder:6.7b"
  },
  "customCommands": [
    {
      "name": "audit",
      "prompt": "Audit this code for security vulnerabilities:\n{{code}}\n\nFollow OWASP Top 10 guidelines.",
      "description": "Security audit"
    }
  ],
  "enterprise": {
    "ssoProvider": "okta",
    "auditLogEnabled": true,
    "allowedModels": ["gpt-4", "claude-3-opus"]
  },
  "contextProviders": [
    {
      "name": "docs",
      "params": {
        "maxTokens": 4096
      }
    }
  ]
}

Key Enterprise Features:

• SSO Integration: Okta/SAML for authentication.
• Model Whitelisting: Restrict to approved models.
• Audit Logs: Compliance-ready logging (AI Security Posture Framework™ COMPLY Phase).

Alternatives Comparison

What's Next?

1. Benchmark Models: Compare gpt-4 vs. claude-3-opus for your use case.

   time continue --model gpt-4 --prompt "Write a Python Flask API"
   time continue --model claude-3-opus --prompt "Write a Python Flask API"
   

   Copy
2. Set Up CI/CD: Add config.json to your repo and validate it in CI:

   npm install -g @continue/cli
   continue validate-config
   

   Copy
3. Explore Self-Hosting: Deploy Continue on-prem for <a href="/services/on-premise-ai">air-gapped</a> environments:

   docker run -p 3000:3000 -v ~/.continue:/root/.continue continuedev/continue
   

   Copy

Need help deploying Continue at scale? Hyperion Consulting offers AI tools consulting to optimize your setup for security, cost, and performance. Visit hyperion-consulting.io to learn more.