AIRGAP StudioAIRGAP Studio

AIRGAP Lite Assistant

The default AI coding assistant — inline autocomplete and RAG-based chat.

Overview

AIRGAP Lite Assistant is the default AI coding assistant that activates automatically the first time you launch AIRGAP Studio. It is based on a Continue 1.3.39 fork and delivers lightweight inline autocomplete and RAG-based chat.

  • The first assistant new users encounter
  • Automatically active on first launch (default)
  • Fully offline — every feature runs against the local llama-server
  • Recommended when you do not need heavy autonomous agent behavior

If you need autonomous multi-file refactoring or automatic terminal command execution, switch to AIRGAP Assistant (Cline fork).

Key Features

Inline Autocomplete

Suggests the next line or block in real time while you type. Responds immediately after a short debounce — press Tab to accept, Esc to dismiss.

Quick Edit

Modify a selected block with a natural-language instruction. Changes are previewed as a diff and applied on approval.

Inline Edit

Open an inline editing prompt inside the editor and apply natural-language instructions to the current cursor position or selection.

RAG-based Chat

Workspace-aware chat with context attached automatically from the indexed codebase. Use @ mentions to add files, symbols, folders, or documents to the context.

Inject context explicitly using mentions such as @codebase, @file, @folder, @docs, and @terminal.

Keybindings

ActionShortcut
Quick Edit (edit selection)Ctrl+I
Open chat panel / send messageCtrl+L
Trigger inline autocomplete manuallyCtrl+Alt+Space
Accept autocomplete suggestionTab
Dismiss autocomplete suggestionEsc
New chat sessionCtrl+Shift+L

Shortcuts may differ depending on your OS and keyboard layout. Open the Keyboard Shortcuts panel in VSCodium and filter by Continue: to view and rebind every command.

AI Model Selection

You cannot change the model from inside AIRGAP Lite Assistant. Model selection is the sole responsibility of AIRGAP Monitor, and is performed via:

  • Command Palette (Ctrl+Shift+P) → AIRGAP: Select LLM Model

llama-server restarts automatically once a new model is selected, and the current-model indicator in Lite Assistant refreshes on its own. See the AIRGAP Monitor document for the full flow.

External Providers

The default provider is the local llama-server. External foundation providers (OpenAI / Anthropic / Gemini, etc.) can be opt-in exposed through the foundationProvidersEnabled flag in airgap-env.json.

{
  "foundationProvidersEnabled": true
}

Foundation providers are not recommended in air-gapped environments. Keep the default (false) so that all inference stays inside the local llama-server.

Comparison with AIRGAP Assistant (Cline fork)

ItemAIRGAP Lite AssistantAIRGAP Assistant
BaseContinue 1.3.39 forkCline 3.56.2 fork
BehaviorInline autocomplete + Quick Edit + RAG chatAutonomous agent (multi-file edits, terminal execution, Checkpoint)
Default active✓ (default)—
Provider policyllama-server default + foundation opt-inllama-server + 4 OpenAI-compatible providers (openai / openai-native / openai-codex / lmstudio)
Recommended forEveryday autocomplete, short edits, RAG queriesMulti-file refactoring, autonomous tasks

The two assistants are mutually exclusive — only one is active at any time. To switch to the other, follow the assistant-switching procedure.

Switching Assistants

To switch to AIRGAP Assistant, run AIRGAP: Switch Assistant from the Command Palette. The switch requires a VSCodium restart (the --disable-extension CLI flag is only applied at spawn time).

See Switching AI Assistants for the full procedure.