$devvkit learn --librarie continue.dev-guide

Continue.dev Guide

[ai][open-source][privacy][vibe-coding]

AI / LLM Tools

Install

# VS Code:
# Install "Continue" from the marketplace

# JetBrains:
# Install from JetBrains Marketplace

# Or:
npx create-continue
# Open-source: github.com/continuedev/continue

Continue is the open-source alternative to Cursor and Copilot. It gives you the same AI coding experience (tab completion, inline chat, multi-file editing) but with full control over your data and models. Use GPT-4o, Claude, Ollama local models, or a custom endpoint.

The key feature is model flexibility. Run Tab autocomplete with a small local model (StarCoder2-3B or CodeGemma-2B) for fast, private completions, and use Claude/GPT-4o for complex chat and edits. All config is in `~/.continue/config.json`.

Continue supports context providers: @file (reference a file), @folder (reference a folder), @codebase (search the entire codebase), @terminal (include terminal output), @diff (show git diff), and @problems (include VS Code errors). Great for privacy-sensitive environments where code must not leave the network.

Setup

Install and configure— Set up Continue.

# Install VS Code extension from marketplace
# Or JetBrains plugin from JetBrains Marketplace

# Open Continue sidebar:
# Click the Continue icon in the activity bar
# Or press Cmd+Shift+R (Ctrl+Shift+R)

Tab Autocomplete

Tab completion— Inline AI suggestions.

# Configure autocomplete model in ~/.continue/config.json:
{
  "tabAutocompleteModel": {
    "title": "Tab Autocomplete",
    "provider": "ollama",
    "model": "starcoder2:3b"
  }
}

# Or use free API:
"tabAutocompleteModel": {
  "provider": "free-trial",
  "model": "codestral-latest"
}

Local-only mode— Air-gapped development.

# ~/.continue/config.json
{
  "models": [{
    "title": "Local LLM",
    "provider": "ollama",
    "model": "codellama:7b"
  }],
  "tabAutocompleteModel": {
    "title": "Local Tab",
    "provider": "ollama",
    "model": "starcoder2:3b"
  },
  "disableIndexing": true,  # No codebase search (avoids network)
  "allowAnonymousTelemetry": false
}

# Only local Ollama requests, no API keys needed

Chat & Edit

Chat — ask and edit— Select code, ask AI.

# Select code, press Cmd+I (Ctrl+I)
# Ask in the input box:
"What does this function do?"
"Add TypeScript types to this"
"Refactor to use async/await"
"Find potential bugs"

# Use the edit mode to apply changes:
# Highlight code → Ctrl+Shift+E
# Type: "Rewrite this with better error handling"

Multi-file edits (Apply)— Edit multiple files at once.

# In chat, describe the change:
"Add pagination to all list endpoints"

# Continue shows diffs for each file
# Review and accept with Ctrl+Enter
# Or reject specific files

# For complex changes, use the Apply feature:
# Chat asks: "Which files do you want me to edit?"
# List the files or use @file references

Context Providers

Context providers— Reference files and project.

# Type @ in the chat input:
@file src/app.ts         # Reference a specific file
@folder src/components   # Reference an entire folder
@codebase                # Search entire codebase
@terminal                # Include terminal output
@diff                    # Show git diff
@problems                # Include VS Code errors

# Example:
"@codebase Find all places where we fetch users"
"@file src/app.ts explain the routing setup"

Model Config

Custom model config— Use local or cloud models.

# ~/.continue/config.json
{
  "models": [
    {
      "title": "Claude 3.5 Sonnet",
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "apiKey": "sk-ant-..."
    },
    {
      "title": "Local Llama",
      "provider": "ollama",
      "model": "llama3.2"
    },
    {
      "title": "OpenAI Compatible",
      "provider": "openai",
      "apiBase": "http://localhost:8080/v1",
      "model": "my-model"
    }
  ]
}