Continue

Continue - Open Source AI Code Assistant

Czym jest Continue?

Continue to open-source'owy asystent AI do programowania dla VS Code i JetBrains IDE (IntelliJ, PyCharm, WebStorm, etc.). W przeciwieństwie do GitHub Copilot, Continue pozwala na używanie dowolnego modelu językowego - od komercyjnych API (GPT-4, Claude, Gemini) po lokalne modele uruchamiane przez Ollama, LM Studio czy vLLM.

Continue został stworzony z myślą o deweloperach, którzy chcą pełnej kontroli nad swoimi danymi i preferują elastyczność w wyborze modeli AI. Projekt jest w pełni open-source (Apache 2.0), co oznacza, że możesz go modyfikować, hostować samodzielnie i używać bez żadnych opłat - płacisz tylko za API modeli, które wybierzesz.

Dlaczego Continue?

Kluczowe zalety Continue:

100% Open Source - Apache 2.0 license, pełna transparentność kodu
Dowolny model LLM - GPT-4, Claude, Gemini, Llama, Mistral, CodeLlama, i więcej
Lokalne modele - Pełna prywatność z Ollama, LM Studio, vLLM
Bring Your Own Key - Używaj swoich API keys, płać tylko za użycie
Context providers - Dodaj dokumentację, terminal, search do kontekstu
Slash commands - Szybkie akcje przez /edit, /comment, /share
Tab autocomplete - Inteligentne uzupełnianie kodu jak Copilot

Continue vs GitHub Copilot

Cecha	Continue	GitHub Copilot
Open Source	✅ Tak (Apache 2.0)	❌ Nie
Modele	Dowolne (GPT-4, Claude, lokalne)	Tylko modele GitHub/OpenAI
Lokalne modele	✅ Ollama, LM Studio, vLLM	❌ Brak
Prywatność danych	✅ Pełna kontrola	⚠️ Dane na serwerach GitHub
Cena	Darmowy (BYOK)	$10-19/mo
Self-hosting	✅ Tak	❌ Nie
Chat	✅ Wbudowany	✅ Copilot Chat
Autocomplete	✅ Tab completion	✅ Tab completion
IDE	VS Code, JetBrains	VS Code, JetBrains, Vim

Instalacja

VS Code

Code

Bash

# Z marketplace
code --install-extension Continue.continue

# Lub przez VS Code:
# 1. Otwórz Extensions (Ctrl+Shift+X)
# 2. Szukaj "Continue"
# 3. Kliknij Install

JetBrains (IntelliJ, PyCharm, WebStorm)

Code

Bash

# Z JetBrains Marketplace:
# 1. Settings → Plugins → Marketplace
# 2. Szukaj "Continue"
# 3. Install
# 4. Restart IDE

Weryfikacja instalacji

Po instalacji:

Otwórz panel Continue (Ctrl+L lub Cmd+L na Mac)
Powinieneś zobaczyć chat interface
Skonfiguruj model w ~/.continue/config.json

Konfiguracja modeli

Struktura config.json

Główny plik konfiguracyjny znajduje się w ~/.continue/config.json:

Code

JSON

{
  "models": [
    {
      "title": "Claude 3.5 Sonnet",
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "apiKey": "sk-ant-xxx"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Starcoder",
    "provider": "ollama",
    "model": "starcoder2:3b"
  },
  "embeddingsProvider": {
    "provider": "ollama",
    "model": "nomic-embed-text"
  },
  "contextProviders": [],
  "slashCommands": []
}

Konfiguracja Claude (Anthropic)

Code

JSON

{
  "models": [
    {
      "title": "Claude 3.5 Sonnet",
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "apiKey": "sk-ant-api03-xxx",
      "contextLength": 200000,
      "completionOptions": {
        "temperature": 0.7,
        "maxTokens": 4096
      }
    },
    {
      "title": "Claude 3 Opus",
      "provider": "anthropic",
      "model": "claude-3-opus-20240229",
      "apiKey": "sk-ant-api03-xxx"
    },
    {
      "title": "Claude 3 Haiku (szybki)",
      "provider": "anthropic",
      "model": "claude-3-haiku-20240307",
      "apiKey": "sk-ant-api03-xxx"
    }
  ]
}

Konfiguracja OpenAI (GPT-4)

Code

JSON

{
  "models": [
    {
      "title": "GPT-4 Turbo",
      "provider": "openai",
      "model": "gpt-4-turbo",
      "apiKey": "sk-xxx"
    },
    {
      "title": "GPT-4o",
      "provider": "openai",
      "model": "gpt-4o",
      "apiKey": "sk-xxx"
    },
    {
      "title": "GPT-4o Mini",
      "provider": "openai",
      "model": "gpt-4o-mini",
      "apiKey": "sk-xxx"
    }
  ]
}

Konfiguracja lokalnych modeli (Ollama)

Code

JSON

{
  "models": [
    {
      "title": "Llama 3.1 70B",
      "provider": "ollama",
      "model": "llama3.1:70b"
    },
    {
      "title": "CodeLlama 34B",
      "provider": "ollama",
      "model": "codellama:34b"
    },
    {
      "title": "Mistral 7B",
      "provider": "ollama",
      "model": "mistral:7b"
    },
    {
      "title": "DeepSeek Coder",
      "provider": "ollama",
      "model": "deepseek-coder:33b"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Starcoder 3B",
    "provider": "ollama",
    "model": "starcoder2:3b"
  }
}

Konfiguracja wielu providerów

Code

JSON

{
  "models": [
    {
      "title": "Claude Sonnet (główny)",
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "apiKey": "sk-ant-xxx"
    },
    {
      "title": "GPT-4 (backup)",
      "provider": "openai",
      "model": "gpt-4-turbo",
      "apiKey": "sk-xxx"
    },
    {
      "title": "Llama Local (offline)",
      "provider": "ollama",
      "model": "llama3:8b"
    }
  ]
}

Ollama - Lokalne modele

Instalacja Ollama

Code

Bash

# macOS / Linux
curl -fsSL https://ollama.ai/install.sh | sh

# macOS (Homebrew)
brew install ollama

# Windows
# Pobierz installer z https://ollama.com/download

Uruchomienie Ollama

Code

Bash

# Uruchom serwer Ollama
ollama serve

# W nowym terminalu - pobierz modele
ollama pull llama3:8b
ollama pull codellama:7b
ollama pull starcoder2:3b
ollama pull deepseek-coder:6.7b

Popularne modele do kodowania

Code

Bash

# Modele ogólne
ollama pull llama3:8b          # 4.7GB - dobry balans
ollama pull llama3:70b         # 40GB - najlepsza jakość
ollama pull mistral:7b         # 4.1GB - szybki i dobry

# Modele do kodu
ollama pull codellama:7b       # 3.8GB - Meta code model
ollama pull codellama:34b      # 19GB - lepszy dla kodu
ollama pull deepseek-coder:6.7b # 3.8GB - świetny do kodu
ollama pull starcoder2:3b      # 1.7GB - autocomplete

# Modele do chat
ollama pull llama3.1:8b        # 4.7GB - najnowsza wersja
ollama pull phi3:14b           # 7.9GB - Microsoft model

Konfiguracja Ollama w Continue

Code

JSON

{
  "models": [
    {
      "title": "Llama 3 Local",
      "provider": "ollama",
      "model": "llama3:8b",
      "apiBase": "http://localhost:11434"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Starcoder Autocomplete",
    "provider": "ollama",
    "model": "starcoder2:3b",
    "apiBase": "http://localhost:11434"
  },
  "embeddingsProvider": {
    "provider": "ollama",
    "model": "nomic-embed-text",
    "apiBase": "http://localhost:11434"
  }
}

Ollama z GPU (NVIDIA)

Code

Bash

# Ollama automatycznie wykrywa GPU
# Sprawdź status GPU
ollama ps

# Uruchom z konkretnym GPU
CUDA_VISIBLE_DEVICES=0 ollama serve

Główne funkcje

Chat (Cmd+L / Ctrl+L)

Chat to główna funkcja Continue pozwalająca na rozmowę z AI o kodzie:

Code

TEXT

// Zaznacz kod i naciśnij Cmd+L

User: Explain this function
[zaznaczony kod]

Continue: This function does XYZ...
It takes parameters A and B...
The return value is...

Przydatne użycia chatu:

Wyjaśnienie złożonego kodu
Pytania o najlepsze praktyki
Debugging - "dlaczego ten kod nie działa?"
Propozycje refaktoringu

Edit (Cmd+I / Ctrl+I)

Edit pozwala na bezpośrednią modyfikację kodu przez AI:

Code

TypeScript

// Zaznacz kod, naciśnij Cmd+I
// Wpisz instrukcję:
"Refactor to use async/await instead of callbacks"

// Continue wygeneruje diff, który możesz zaakceptować lub odrzucić

Przykładowe instrukcje edit:

"Add error handling"
"Convert to TypeScript"
"Add input validation"
"Optimize this loop"
"Make this function pure"

Tab Autocomplete

Autocomplete działa automatycznie podczas pisania:

Code

TypeScript

// Zacznij pisać...
function calculateTax(

// Continue zasugeruje:
function calculateTax(amount: number, rate: number): number {
  return amount * rate
}

// Naciśnij Tab aby zaakceptować

Konfiguracja autocomplete:

Code

JSON

{
  "tabAutocompleteModel": {
    "title": "Fast Autocomplete",
    "provider": "ollama",
    "model": "starcoder2:3b"
  },
  "tabAutocompleteOptions": {
    "useCopyBuffer": true,
    "useFileSuffix": true,
    "maxPromptTokens": 1024,
    "debounceDelay": 300
  }
}

Slash Commands

Slash commands to szybkie akcje dostępne przez /nazwa:

Code

TEXT

/edit refactor this to be more readable
/comment add JSDoc comments to all functions
/share create a shareable link for this code
/cmd generate a terminal command to do X
/test write unit tests for this function
/fix fix the errors in this code

Konfiguracja custom slash commands:

Code

JSON

{
  "slashCommands": [
    {
      "name": "test",
      "description": "Write unit tests",
      "prompt": "Write comprehensive unit tests for the selected code using Jest. Include edge cases and error scenarios."
    },
    {
      "name": "doc",
      "description": "Add documentation",
      "prompt": "Add detailed JSDoc/TSDoc documentation to this code. Include @param, @returns, and @example where appropriate."
    },
    {
      "name": "review",
      "description": "Code review",
      "prompt": "Review this code for potential issues, bugs, and improvements. Consider performance, security, and best practices."
    },
    {
      "name": "explain",
      "description": "Explain code",
      "prompt": "Explain this code step by step. What does each part do? What's the overall purpose?"
    }
  ]
}

Context Providers

Context providers dodają dodatkowy kontekst do promptów:

Wbudowane context providers

Code

JSON

{
  "contextProviders": [
    {
      "name": "code",
      "params": {}
    },
    {
      "name": "docs",
      "params": {}
    },
    {
      "name": "terminal",
      "params": {}
    },
    {
      "name": "search",
      "params": {}
    },
    {
      "name": "open",
      "params": {}
    },
    {
      "name": "problems",
      "params": {}
    }
  ]
}

URL Context Provider

Code

JSON

{
  "contextProviders": [
    {
      "name": "url",
      "params": {
        "preset": "docs"
      }
    }
  ]
}

Użycie w chacie:

Code

TEXT

@url https://docs.react.dev/learn/state-a-components-memory
How do I use useState in React?

Codebase Context Provider

Code

JSON

{
  "contextProviders": [
    {
      "name": "codebase",
      "params": {
        "nRetrieve": 25,
        "nFinal": 10,
        "useReranking": true
      }
    }
  ]
}

Użycie:

Code

TEXT

@codebase How is authentication implemented in this project?

File Context Provider

Code

JSON

{
  "contextProviders": [
    {
      "name": "file",
      "params": {}
    }
  ]
}

Użycie:

Code

TEXT

@file src/utils/auth.ts explain this file

Folder Context Provider

Code

JSON

{
  "contextProviders": [
    {
      "name": "folder",
      "params": {}
    }
  ]
}

Użycie:

Code

TEXT

@folder src/components what components do we have?

Custom Documentation

Code

JSON

{
  "contextProviders": [
    {
      "name": "docs",
      "params": {
        "sites": [
          {
            "title": "Next.js Docs",
            "startUrl": "https://nextjs.org/docs",
            "rootUrl": "https://nextjs.org/docs"
          },
          {
            "title": "Prisma Docs",
            "startUrl": "https://www.prisma.io/docs",
            "rootUrl": "https://www.prisma.io/docs"
          }
        ]
      }
    }
  ]
}

Embeddings i RAG

Konfiguracja embeddings

Code

JSON

{
  "embeddingsProvider": {
    "provider": "ollama",
    "model": "nomic-embed-text",
    "apiBase": "http://localhost:11434"
  }
}

Alternatywne providers

Code

JSON

{
  "embeddingsProvider": {
    "provider": "openai",
    "model": "text-embedding-3-small",
    "apiKey": "sk-xxx"
  }
}

Code

JSON

{
  "embeddingsProvider": {
    "provider": "transformers.js",
    "model": "Xenova/all-MiniLM-L6-v2"
  }
}

Indexing codebase

Code

Bash

# Continue automatycznie indeksuje codebase
# Możesz też wymusić reindex:
# Cmd+Shift+P → "Continue: Rebuild codebase index"

.continueignore

Wyklucz pliki z indeksowania:

Code

GITIGNORE

# .continueignore
node_modules/
dist/
build/
*.min.js
*.lock
coverage/
.next/

Zaawansowana konfiguracja

System prompt customization

Code

JSON

{
  "systemMessage": "You are an expert TypeScript developer working on a Next.js project. Follow these guidelines:\n- Use TypeScript strict mode\n- Prefer functional components with hooks\n- Use Tailwind CSS for styling\n- Follow the project's existing patterns"
}

Model completion options

Code

JSON

{
  "models": [
    {
      "title": "Claude Custom",
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "completionOptions": {
        "temperature": 0.3,
        "maxTokens": 4096,
        "topP": 0.95,
        "topK": 40,
        "presencePenalty": 0,
        "frequencyPenalty": 0
      }
    }
  ]
}

Prompt templates

Code

JSON

{
  "promptTemplates": {
    "edit": "Please edit the following code according to these instructions:\n\n{{{userInput}}}\n\nCode:\n```\n{{{selectedText}}}\n```",
    "chat": "Context:\n{{{contextItems}}}\n\nQuestion: {{{userInput}}}",
    "autocomplete": "{{{prefix}}}"
  }
}

Request options

Code

JSON

{
  "requestOptions": {
    "timeout": 60000,
    "verifySsl": true,
    "headers": {
      "X-Custom-Header": "value"
    }
  }
}

Integracja z narzędziami

Git integration

Continue automatycznie wykrywa kontekst git:

Code

TEXT

@git show me recent changes
@git what changed in the last commit?
@diff explain these changes

Terminal integration

Code

TEXT

@terminal show recent commands
@terminal what was the last error?

Problem integration

Code

TEXT

@problems fix all TypeScript errors
@problems what errors do we have?

Keyboard shortcuts

VS Code shortcuts

Skrót	Akcja
`Cmd+L`	Otwórz chat
`Cmd+I`	Edit inline
`Cmd+Shift+L`	Dodaj do chatu
`Cmd+Shift+R`	Regeneruj odpowiedź
`Tab`	Akceptuj autocomplete
`Esc`	Zamknij autocomplete
`Cmd+Shift+Enter`	Submit
`Cmd+Delete`	Wyczyść chat

JetBrains shortcuts

Skrót	Akcja
`Ctrl+L`	Otwórz chat
`Ctrl+I`	Edit inline
`Ctrl+Shift+L`	Dodaj do chatu
`Tab`	Akceptuj autocomplete

Customization shortcuts

Code

JSON

// VS Code keybindings.json
[
  {
    "key": "cmd+shift+c",
    "command": "continue.chat"
  },
  {
    "key": "cmd+shift+e",
    "command": "continue.edit"
  }
]

Provider-specific setup

Azure OpenAI

Code

JSON

{
  "models": [
    {
      "title": "Azure GPT-4",
      "provider": "azure",
      "model": "gpt-4",
      "apiKey": "your-azure-key",
      "apiBase": "https://your-resource.openai.azure.com",
      "apiVersion": "2024-02-15-preview",
      "deployment": "your-deployment-name"
    }
  ]
}

AWS Bedrock

Code

JSON

{
  "models": [
    {
      "title": "Claude via Bedrock",
      "provider": "bedrock",
      "model": "anthropic.claude-3-sonnet-20240229-v1:0",
      "region": "us-east-1"
    }
  ]
}

Google AI (Gemini)

Code

JSON

{
  "models": [
    {
      "title": "Gemini Pro",
      "provider": "google",
      "model": "gemini-1.5-pro",
      "apiKey": "your-google-api-key"
    }
  ]
}

Groq (ultra-fast inference)

Code

JSON

{
  "models": [
    {
      "title": "Llama via Groq",
      "provider": "groq",
      "model": "llama-3.1-70b-versatile",
      "apiKey": "gsk_xxx"
    }
  ]
}

Together AI

Code

JSON

{
  "models": [
    {
      "title": "Mistral via Together",
      "provider": "together",
      "model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
      "apiKey": "your-together-key"
    }
  ]
}

LM Studio

Code

JSON

{
  "models": [
    {
      "title": "LM Studio Model",
      "provider": "lmstudio",
      "model": "local-model",
      "apiBase": "http://localhost:1234/v1"
    }
  ]
}

OpenRouter

Code

JSON

{
  "models": [
    {
      "title": "Claude via OpenRouter",
      "provider": "openrouter",
      "model": "anthropic/claude-3.5-sonnet",
      "apiKey": "sk-or-xxx"
    }
  ]
}

Przypadki użycia

Code review

Code

TEXT

/review

# Lub bardziej szczegółowo:
Review this code for:
- Security vulnerabilities
- Performance issues
- Code smells
- TypeScript best practices

Refactoring

Code

TEXT

/edit Extract this logic into a custom hook

# Lub przez chat:
@file src/components/UserForm.tsx
How can I refactor this to be more maintainable?

Documentation

Code

TEXT

/doc

# Lub:
Add comprehensive documentation including:
- Function purpose
- Parameters with types
- Return value
- Usage examples

Testing

Code

TEXT

/test

# Lub:
Write unit tests using Vitest for this function.
Include edge cases and error scenarios.

Debugging

Code

TEXT

I'm getting this error:
[paste error]

@file src/utils/api.ts
What might be causing this?

Learning

Code

TEXT

@docs https://react.dev
Explain how useMemo works and when I should use it

Bezpieczeństwo i prywatność

Lokalne modele (maksymalna prywatność)

Code

JSON

{
  "models": [
    {
      "title": "Local Llama",
      "provider": "ollama",
      "model": "llama3:8b"
    }
  ],
  "tabAutocompleteModel": {
    "provider": "ollama",
    "model": "starcoder2:3b"
  },
  "embeddingsProvider": {
    "provider": "ollama",
    "model": "nomic-embed-text"
  }
}

Telemetry

Code

JSON

{
  "allowAnonymousTelemetry": false
}

Data handling

Continue:

Nie przechowuje twojego kodu na serwerach
Nie trenuje modeli na twoim kodzie
Wysyła kod tylko do wybranego providera API
Z lokalnymi modelami - żadne dane nie opuszczają twojego komputera

Porównanie z innymi narzędziami

Continue vs Cursor

Cecha	Continue	Cursor
Open Source	✅ Tak	❌ Nie
IDE	VS Code, JetBrains	Custom (fork VS Code)
Lokalne modele	✅ Tak	⚠️ Ograniczone
Cena	Free (BYOK)	$20/mo
Customization	Pełna	Ograniczona

Continue vs Cody

Cecha	Continue	Cody
Open Source	✅ Tak	✅ Tak
Codebase context	✅ Embeddings	✅ Sourcegraph
Lokalne modele	✅ Tak	⚠️ Ograniczone
Enterprise	Self-host	Sourcegraph
Cena	Free	Free + Pro $9/mo

Continue vs Tabnine

Cecha	Continue	Tabnine
Open Source	✅ Tak	❌ Nie
Chat	✅ Tak	✅ Tak
Lokalne modele	✅ Dowolne	⚠️ Tylko Tabnine
Cena	Free (BYOK)	Free + Pro $12/mo

Rozwiązywanie problemów

Continue nie widzi Ollama

Code

Bash

# Sprawdź czy Ollama działa
curl http://localhost:11434/api/tags

# Jeśli nie odpowiada:
ollama serve

# Sprawdź dostępne modele
ollama list

Autocomplete nie działa

Sprawdź czy tabAutocompleteModel jest skonfigurowany
Sprawdź czy model jest pobrany (dla Ollama)
Sprawdź logi: Continue: Show Logs

Wolne odpowiedzi

Dla lokalnych modeli - użyj mniejszego modelu (7B zamiast 70B)
Sprawdź czy GPU jest wykorzystywane
Zwiększ debounceDelay dla autocomplete

API errors

Code

JSON

{
  "requestOptions": {
    "timeout": 120000,
    "verifySsl": false
  }
}

Reset konfiguracji

Code

Bash

# Backup
cp ~/.continue/config.json ~/.continue/config.json.bak

# Reset
rm ~/.continue/config.json

# Restart VS Code/IDE

Migracja z GitHub Copilot

Krok 1: Instalacja Continue

Code

Bash

code --install-extension Continue.continue

Krok 2: Konfiguracja podobna do Copilot

Code

JSON

{
  "models": [
    {
      "title": "GPT-4 Turbo",
      "provider": "openai",
      "model": "gpt-4-turbo",
      "apiKey": "sk-xxx"
    }
  ],
  "tabAutocompleteModel": {
    "title": "GPT-3.5 Turbo",
    "provider": "openai",
    "model": "gpt-3.5-turbo",
    "apiKey": "sk-xxx"
  }
}

Krok 3: Wyłączenie Copilot

Code

Bash

code --uninstall-extension GitHub.copilot

Cennik

100% darmowy - Apache 2.0 license
BYOK (Bring Your Own Key)
Płacisz tylko za API modeli

Koszty API (przykładowe)

Model	Provider	Cena (input/output)
Claude 3.5 Sonnet	Anthropic	$3/$15 per 1M tokens
GPT-4 Turbo	OpenAI	$10/$30 per 1M tokens
Llama 3 70B	Groq	$0.59/$0.79 per 1M tokens
Lokalne modele	Ollama	$0 (tylko hardware)

Oszczędności vs Copilot

Copilot: $10-19/mo = $120-228/rok
Continue + Lokalne modele: $0
Continue + API: ~$5-20/mo (zależnie od użycia)

FAQ - Najczęściej zadawane pytania

Czy Continue jest tak dobry jak Copilot?

Z odpowiednim modelem (Claude 3.5 Sonnet, GPT-4) - tak, często lepszy. Dodatkowo masz pełną kontrolę nad modelami i prywatnością.

Jakie modele polecacie do kodowania?

Najlepsza jakość: Claude 3.5 Sonnet lub GPT-4 Turbo
Balans jakość/cena: Llama 3.1 70B (Groq) lub DeepSeek Coder
Lokalne: Llama 3 8B lub CodeLlama 34B
Autocomplete: Starcoder2 3B lub DeepSeek Coder 6.7B

Czy mogę używać Continue offline?

Tak, z lokalnymi modelami przez Ollama. Potrzebujesz wystarczającej mocy obliczeniowej (min. 8GB RAM dla 7B modeli).

Czy Continue działa z moim językiem programowania?

Tak, Continue działa z każdym językiem. Jakość zależy od wybranego modelu - większość modeli dobrze radzi sobie z popularnymi językami.

Jak przejść z Cursor na Continue?

Continue jest rozszerzeniem do VS Code - nie musisz zmieniać IDE. Po prostu zainstaluj Continue w VS Code i skonfiguruj preferowane modele.

Czy mogę używać Continue w zespole?

Tak, możesz:

Współdzielić config.json w repo
Użyć environment variables dla API keys
Self-hostować modele dla całego zespołu

Continue - open source AI code assistant

What is Continue?

Continue is an open-source AI coding assistant for VS Code and JetBrains IDEs (IntelliJ, PyCharm, WebStorm, etc.). Unlike GitHub Copilot, Continue lets you use any language model - from commercial APIs (GPT-4, Claude, Gemini) to local models running through Ollama, LM Studio, or vLLM.

Continue was built with developers in mind who want full control over their data and prefer flexibility in choosing AI models. The project is fully open-source (Apache 2.0), which means you can modify it, self-host it, and use it without any fees - you only pay for the API of the models you choose.

Why Continue?

Key advantages of Continue:

100% Open Source - Apache 2.0 license, full code transparency
Any LLM model - GPT-4, Claude, Gemini, Llama, Mistral, CodeLlama, and more
Local models - Full privacy with Ollama, LM Studio, vLLM
Bring Your Own Key - Use your own API keys, pay only for usage
Context providers - Add documentation, terminal, search to context
Slash commands - Quick actions via /edit, /comment, /share
Tab autocomplete - Intelligent code completion like Copilot

Continue vs GitHub Copilot

Feature	Continue	GitHub Copilot
Open Source	Yes (Apache 2.0)	No
Models	Any (GPT-4, Claude, local)	GitHub/OpenAI models only
Local models	Ollama, LM Studio, vLLM	None
Data privacy	Full control	Data on GitHub servers
Price	Free (BYOK)	$10-19/mo
Self-hosting	Yes	No
Chat	Built-in	Copilot Chat
Autocomplete	Tab completion	Tab completion
IDE	VS Code, JetBrains	VS Code, JetBrains, Vim

Installation

VS Code

Code

Bash

# From the marketplace
code --install-extension Continue.continue

# Or through VS Code:
# 1. Open Extensions (Ctrl+Shift+X)
# 2. Search for "Continue"
# 3. Click Install

JetBrains (IntelliJ, PyCharm, WebStorm)

Code

Bash

# From JetBrains Marketplace:
# 1. Settings → Plugins → Marketplace
# 2. Search for "Continue"
# 3. Install
# 4. Restart IDE

Verifying the installation

After installation:

Open the Continue panel (Ctrl+L or Cmd+L on Mac)
You should see the chat interface
Configure your model in ~/.continue/config.json

Model configuration

config.json structure

The main configuration file is located at ~/.continue/config.json:

Code

JSON

{
  "models": [
    {
      "title": "Claude 3.5 Sonnet",
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "apiKey": "sk-ant-xxx"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Starcoder",
    "provider": "ollama",
    "model": "starcoder2:3b"
  },
  "embeddingsProvider": {
    "provider": "ollama",
    "model": "nomic-embed-text"
  },
  "contextProviders": [],
  "slashCommands": []
}

Claude (Anthropic) configuration

Code

JSON

{
  "models": [
    {
      "title": "Claude 3.5 Sonnet",
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "apiKey": "sk-ant-api03-xxx",
      "contextLength": 200000,
      "completionOptions": {
        "temperature": 0.7,
        "maxTokens": 4096
      }
    },
    {
      "title": "Claude 3 Opus",
      "provider": "anthropic",
      "model": "claude-3-opus-20240229",
      "apiKey": "sk-ant-api03-xxx"
    },
    {
      "title": "Claude 3 Haiku (fast)",
      "provider": "anthropic",
      "model": "claude-3-haiku-20240307",
      "apiKey": "sk-ant-api03-xxx"
    }
  ]
}

OpenAI (GPT-4) configuration

Code

JSON

{
  "models": [
    {
      "title": "GPT-4 Turbo",
      "provider": "openai",
      "model": "gpt-4-turbo",
      "apiKey": "sk-xxx"
    },
    {
      "title": "GPT-4o",
      "provider": "openai",
      "model": "gpt-4o",
      "apiKey": "sk-xxx"
    },
    {
      "title": "GPT-4o Mini",
      "provider": "openai",
      "model": "gpt-4o-mini",
      "apiKey": "sk-xxx"
    }
  ]
}

Local models (Ollama) configuration

Code

JSON

{
  "models": [
    {
      "title": "Llama 3.1 70B",
      "provider": "ollama",
      "model": "llama3.1:70b"
    },
    {
      "title": "CodeLlama 34B",
      "provider": "ollama",
      "model": "codellama:34b"
    },
    {
      "title": "Mistral 7B",
      "provider": "ollama",
      "model": "mistral:7b"
    },
    {
      "title": "DeepSeek Coder",
      "provider": "ollama",
      "model": "deepseek-coder:33b"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Starcoder 3B",
    "provider": "ollama",
    "model": "starcoder2:3b"
  }
}

Multiple providers configuration

Code

JSON

{
  "models": [
    {
      "title": "Claude Sonnet (primary)",
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "apiKey": "sk-ant-xxx"
    },
    {
      "title": "GPT-4 (backup)",
      "provider": "openai",
      "model": "gpt-4-turbo",
      "apiKey": "sk-xxx"
    },
    {
      "title": "Llama Local (offline)",
      "provider": "ollama",
      "model": "llama3:8b"
    }
  ]
}

Ollama - local models

Installing Ollama

Code

Bash

# macOS / Linux
curl -fsSL https://ollama.ai/install.sh | sh

# macOS (Homebrew)
brew install ollama

# Windows
# Download the installer from https://ollama.com/download

Running Ollama

Code

Bash

# Start the Ollama server
ollama serve

# In a new terminal - pull models
ollama pull llama3:8b
ollama pull codellama:7b
ollama pull starcoder2:3b
ollama pull deepseek-coder:6.7b

Popular models for coding

Code

Bash

# General-purpose models
ollama pull llama3:8b          # 4.7GB - good balance
ollama pull llama3:70b         # 40GB - best quality
ollama pull mistral:7b         # 4.1GB - fast and capable

# Code-specific models
ollama pull codellama:7b       # 3.8GB - Meta code model
ollama pull codellama:34b      # 19GB - better for code
ollama pull deepseek-coder:6.7b # 3.8GB - excellent for code
ollama pull starcoder2:3b      # 1.7GB - autocomplete

# Chat models
ollama pull llama3.1:8b        # 4.7GB - latest version
ollama pull phi3:14b           # 7.9GB - Microsoft model

Ollama configuration in Continue

Code

JSON

{
  "models": [
    {
      "title": "Llama 3 Local",
      "provider": "ollama",
      "model": "llama3:8b",
      "apiBase": "http://localhost:11434"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Starcoder Autocomplete",
    "provider": "ollama",
    "model": "starcoder2:3b",
    "apiBase": "http://localhost:11434"
  },
  "embeddingsProvider": {
    "provider": "ollama",
    "model": "nomic-embed-text",
    "apiBase": "http://localhost:11434"
  }
}

Ollama with GPU (NVIDIA)

Code

Bash

# Ollama automatically detects the GPU
# Check GPU status
ollama ps

# Run with a specific GPU
CUDA_VISIBLE_DEVICES=0 ollama serve

Core features

Chat (Cmd+L / Ctrl+L)

Chat is the main Continue feature that lets you have a conversation with AI about your code:

Code

TEXT

// Select code and press Cmd+L

User: Explain this function
[selected code]

Continue: This function does XYZ...
It takes parameters A and B...
The return value is...

Useful chat applications:

Explaining complex code
Asking about best practices
Debugging - "why isn't this code working?"
Refactoring suggestions

Edit (Cmd+I / Ctrl+I)

Edit allows direct code modification through AI:

Code

TypeScript

// Select code, press Cmd+I
// Type an instruction:
"Refactor to use async/await instead of callbacks"

// Continue will generate a diff that you can accept or reject

Example edit instructions:

"Add error handling"
"Convert to TypeScript"
"Add input validation"
"Optimize this loop"
"Make this function pure"

Tab Autocomplete

Autocomplete works automatically as you type:

Code

TypeScript

// Start typing...
function calculateTax(

// Continue will suggest:
function calculateTax(amount: number, rate: number): number {
  return amount * rate
}

// Press Tab to accept

Autocomplete configuration:

Code

JSON

{
  "tabAutocompleteModel": {
    "title": "Fast Autocomplete",
    "provider": "ollama",
    "model": "starcoder2:3b"
  },
  "tabAutocompleteOptions": {
    "useCopyBuffer": true,
    "useFileSuffix": true,
    "maxPromptTokens": 1024,
    "debounceDelay": 300
  }
}

Slash commands

Slash commands are quick actions accessible via /name:

Code

TEXT

/edit refactor this to be more readable
/comment add JSDoc comments to all functions
/share create a shareable link for this code
/cmd generate a terminal command to do X
/test write unit tests for this function
/fix fix the errors in this code

Custom slash commands configuration:

Code

JSON

{
  "slashCommands": [
    {
      "name": "test",
      "description": "Write unit tests",
      "prompt": "Write comprehensive unit tests for the selected code using Jest. Include edge cases and error scenarios."
    },
    {
      "name": "doc",
      "description": "Add documentation",
      "prompt": "Add detailed JSDoc/TSDoc documentation to this code. Include @param, @returns, and @example where appropriate."
    },
    {
      "name": "review",
      "description": "Code review",
      "prompt": "Review this code for potential issues, bugs, and improvements. Consider performance, security, and best practices."
    },
    {
      "name": "explain",
      "description": "Explain code",
      "prompt": "Explain this code step by step. What does each part do? What's the overall purpose?"
    }
  ]
}

Context providers

Context providers add additional context to your prompts:

Built-in context providers

Code

JSON

{
  "contextProviders": [
    {
      "name": "code",
      "params": {}
    },
    {
      "name": "docs",
      "params": {}
    },
    {
      "name": "terminal",
      "params": {}
    },
    {
      "name": "search",
      "params": {}
    },
    {
      "name": "open",
      "params": {}
    },
    {
      "name": "problems",
      "params": {}
    }
  ]
}

URL context provider

Code

JSON

{
  "contextProviders": [
    {
      "name": "url",
      "params": {
        "preset": "docs"
      }
    }
  ]
}

Usage in chat:

Code

TEXT

@url https://docs.react.dev/learn/state-a-components-memory
How do I use useState in React?

Codebase context provider

Code

JSON

{
  "contextProviders": [
    {
      "name": "codebase",
      "params": {
        "nRetrieve": 25,
        "nFinal": 10,
        "useReranking": true
      }
    }
  ]
}

Usage:

Code

TEXT

@codebase How is authentication implemented in this project?

File context provider

Code

JSON

{
  "contextProviders": [
    {
      "name": "file",
      "params": {}
    }
  ]
}

Usage:

Code

TEXT

@file src/utils/auth.ts explain this file

Folder context provider

Code

JSON

{
  "contextProviders": [
    {
      "name": "folder",
      "params": {}
    }
  ]
}

Usage:

Code

TEXT

@folder src/components what components do we have?

Custom documentation

Code

JSON

{
  "contextProviders": [
    {
      "name": "docs",
      "params": {
        "sites": [
          {
            "title": "Next.js Docs",
            "startUrl": "https://nextjs.org/docs",
            "rootUrl": "https://nextjs.org/docs"
          },
          {
            "title": "Prisma Docs",
            "startUrl": "https://www.prisma.io/docs",
            "rootUrl": "https://www.prisma.io/docs"
          }
        ]
      }
    }
  ]
}

Embeddings and RAG

Embeddings configuration

Code

JSON

{
  "embeddingsProvider": {
    "provider": "ollama",
    "model": "nomic-embed-text",
    "apiBase": "http://localhost:11434"
  }
}

Alternative providers

Code

JSON

{
  "embeddingsProvider": {
    "provider": "openai",
    "model": "text-embedding-3-small",
    "apiKey": "sk-xxx"
  }
}

Code

JSON

{
  "embeddingsProvider": {
    "provider": "transformers.js",
    "model": "Xenova/all-MiniLM-L6-v2"
  }
}

Indexing the codebase

Code

Bash

# Continue automatically indexes the codebase
# You can also force a reindex:
# Cmd+Shift+P → "Continue: Rebuild codebase index"

.continueignore

Exclude files from indexing:

Code

GITIGNORE

# .continueignore
node_modules/
dist/
build/
*.min.js
*.lock
coverage/
.next/

Advanced configuration

System prompt customization

Code

JSON

{
  "systemMessage": "You are an expert TypeScript developer working on a Next.js project. Follow these guidelines:\n- Use TypeScript strict mode\n- Prefer functional components with hooks\n- Use Tailwind CSS for styling\n- Follow the project's existing patterns"
}

Model completion options

Code

JSON

{
  "models": [
    {
      "title": "Claude Custom",
      "provider": "anthropic",
      "model": "claude-3-5-sonnet-20241022",
      "completionOptions": {
        "temperature": 0.3,
        "maxTokens": 4096,
        "topP": 0.95,
        "topK": 40,
        "presencePenalty": 0,
        "frequencyPenalty": 0
      }
    }
  ]
}

Prompt templates

Code

JSON

{
  "promptTemplates": {
    "edit": "Please edit the following code according to these instructions:\n\n{{{userInput}}}\n\nCode:\n```\n{{{selectedText}}}\n```",
    "chat": "Context:\n{{{contextItems}}}\n\nQuestion: {{{userInput}}}",
    "autocomplete": "{{{prefix}}}"
  }
}

Request options

Code

JSON

{
  "requestOptions": {
    "timeout": 60000,
    "verifySsl": true,
    "headers": {
      "X-Custom-Header": "value"
    }
  }
}

Tool integrations

Git integration

Continue automatically detects git context:

Code

TEXT

@git show me recent changes
@git what changed in the last commit?
@diff explain these changes

Terminal integration

Code

TEXT

@terminal show recent commands
@terminal what was the last error?

Problem integration

Code

TEXT

@problems fix all TypeScript errors
@problems what errors do we have?

Keyboard shortcuts

VS Code shortcuts

Shortcut	Action
`Cmd+L`	Open chat
`Cmd+I`	Inline edit
`Cmd+Shift+L`	Add to chat
`Cmd+Shift+R`	Regenerate response
`Tab`	Accept autocomplete
`Esc`	Dismiss autocomplete
`Cmd+Shift+Enter`	Submit
`Cmd+Delete`	Clear chat

JetBrains shortcuts

Shortcut	Action
`Ctrl+L`	Open chat
`Ctrl+I`	Inline edit
`Ctrl+Shift+L`	Add to chat
`Tab`	Accept autocomplete

Customizing shortcuts

Code

JSON

// VS Code keybindings.json
[
  {
    "key": "cmd+shift+c",
    "command": "continue.chat"
  },
  {
    "key": "cmd+shift+e",
    "command": "continue.edit"
  }
]

Provider-specific setup

Azure OpenAI

Code

JSON

{
  "models": [
    {
      "title": "Azure GPT-4",
      "provider": "azure",
      "model": "gpt-4",
      "apiKey": "your-azure-key",
      "apiBase": "https://your-resource.openai.azure.com",
      "apiVersion": "2024-02-15-preview",
      "deployment": "your-deployment-name"
    }
  ]
}

AWS Bedrock

Code

JSON

{
  "models": [
    {
      "title": "Claude via Bedrock",
      "provider": "bedrock",
      "model": "anthropic.claude-3-sonnet-20240229-v1:0",
      "region": "us-east-1"
    }
  ]
}

Google AI (Gemini)

Code

JSON

{
  "models": [
    {
      "title": "Gemini Pro",
      "provider": "google",
      "model": "gemini-1.5-pro",
      "apiKey": "your-google-api-key"
    }
  ]
}

Groq (ultra-fast inference)

Code

JSON

{
  "models": [
    {
      "title": "Llama via Groq",
      "provider": "groq",
      "model": "llama-3.1-70b-versatile",
      "apiKey": "gsk_xxx"
    }
  ]
}

Together AI

Code

JSON

{
  "models": [
    {
      "title": "Mistral via Together",
      "provider": "together",
      "model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
      "apiKey": "your-together-key"
    }
  ]
}

LM Studio

Code

JSON

{
  "models": [
    {
      "title": "LM Studio Model",
      "provider": "lmstudio",
      "model": "local-model",
      "apiBase": "http://localhost:1234/v1"
    }
  ]
}

OpenRouter

Code

JSON

{
  "models": [
    {
      "title": "Claude via OpenRouter",
      "provider": "openrouter",
      "model": "anthropic/claude-3.5-sonnet",
      "apiKey": "sk-or-xxx"
    }
  ]
}

Use cases

Code review

Code

TEXT

/review

# Or more specifically:
Review this code for:
- Security vulnerabilities
- Performance issues
- Code smells
- TypeScript best practices

Refactoring

Code

TEXT

/edit Extract this logic into a custom hook

# Or through chat:
@file src/components/UserForm.tsx
How can I refactor this to be more maintainable?

Documentation

Code

TEXT

/doc

# Or:
Add comprehensive documentation including:
- Function purpose
- Parameters with types
- Return value
- Usage examples

Testing

Code

TEXT

/test

# Or:
Write unit tests using Vitest for this function.
Include edge cases and error scenarios.

Debugging

Code

TEXT

I'm getting this error:
[paste error]

@file src/utils/api.ts
What might be causing this?

Learning

Code

TEXT

@docs https://react.dev
Explain how useMemo works and when I should use it

Security and privacy

Local models (maximum privacy)

Code

JSON

{
  "models": [
    {
      "title": "Local Llama",
      "provider": "ollama",
      "model": "llama3:8b"
    }
  ],
  "tabAutocompleteModel": {
    "provider": "ollama",
    "model": "starcoder2:3b"
  },
  "embeddingsProvider": {
    "provider": "ollama",
    "model": "nomic-embed-text"
  }
}

Telemetry

Code

JSON

{
  "allowAnonymousTelemetry": false
}

Data handling

Continue:

Does not store your code on any servers
Does not train models on your code
Sends code only to the API provider you choose
With local models - no data leaves your computer

Comparison with other tools

Continue vs Cursor

Feature	Continue	Cursor
Open Source	Yes	No
IDE	VS Code, JetBrains	Custom (VS Code fork)
Local models	Yes	Limited
Price	Free (BYOK)	$20/mo
Customization	Full	Limited

Continue vs Cody

Feature	Continue	Cody
Open Source	Yes	Yes
Codebase context	Embeddings	Sourcegraph
Local models	Yes	Limited
Enterprise	Self-host	Sourcegraph
Price	Free	Free + Pro $9/mo

Continue vs Tabnine

Feature	Continue	Tabnine
Open Source	Yes	No
Chat	Yes	Yes
Local models	Any	Tabnine only
Price	Free (BYOK)	Free + Pro $12/mo

Troubleshooting

Continue cannot see Ollama

Code

Bash

# Check if Ollama is running
curl http://localhost:11434/api/tags

# If it doesn't respond:
ollama serve

# Check available models
ollama list

Autocomplete is not working

Check that tabAutocompleteModel is configured
Check that the model has been pulled (for Ollama)
Check the logs: Continue: Show Logs

Slow responses

For local models - use a smaller model (7B instead of 70B)
Check if the GPU is being utilized
Increase the debounceDelay for autocomplete

API errors

Code

JSON

{
  "requestOptions": {
    "timeout": 120000,
    "verifySsl": false
  }
}

Resetting the configuration

Code

Bash

# Backup
cp ~/.continue/config.json ~/.continue/config.json.bak

# Reset
rm ~/.continue/config.json

# Restart VS Code/IDE

Migrating from GitHub Copilot

Step 1: Install Continue

Code

Bash

code --install-extension Continue.continue

Step 2: Copilot-like configuration

Code

JSON

{
  "models": [
    {
      "title": "GPT-4 Turbo",
      "provider": "openai",
      "model": "gpt-4-turbo",
      "apiKey": "sk-xxx"
    }
  ],
  "tabAutocompleteModel": {
    "title": "GPT-3.5 Turbo",
    "provider": "openai",
    "model": "gpt-3.5-turbo",
    "apiKey": "sk-xxx"
  }
}

Step 3: Disable Copilot

Code

Bash

code --uninstall-extension GitHub.copilot

Pricing

Continue

100% free - Apache 2.0 license
BYOK (Bring Your Own Key)
You only pay for model API usage

API costs (examples)

Model	Provider	Price (input/output)
Claude 3.5 Sonnet	Anthropic	$3/$15 per 1M tokens
GPT-4 Turbo	OpenAI	$10/$30 per 1M tokens
Llama 3 70B	Groq	$0.59/$0.79 per 1M tokens
Local models	Ollama	$0 (hardware only)

Savings vs Copilot

Copilot: $10-19/mo = $120-228/year
Continue + Local models: $0
Continue + API: ~$5-20/mo (depending on usage)

FAQ - frequently asked questions

Is Continue as good as Copilot?

With the right model (Claude 3.5 Sonnet, GPT-4) - yes, often better. On top of that, you get full control over models and privacy.

What models do you recommend for coding?

Best quality: Claude 3.5 Sonnet or GPT-4 Turbo
Quality/price balance: Llama 3.1 70B (Groq) or DeepSeek Coder
Local: Llama 3 8B or CodeLlama 34B
Autocomplete: Starcoder2 3B or DeepSeek Coder 6.7B

Can I use Continue offline?

Yes, with local models through Ollama. You need sufficient computing power (at least 8GB RAM for 7B models).

Does Continue work with my programming language?

Yes, Continue works with any language. Quality depends on the chosen model - most models handle popular languages well.

How do I switch from Cursor to Continue?

Continue is a VS Code extension - you don't need to change your IDE. Simply install Continue in VS Code and configure your preferred models.

Can I use Continue in a team?

Yes, you can:

Share config.json in your repo
Use environment variables for API keys
Self-host models for the entire team