Copilot-style tab autocomplete for VS Code, powered by a local Ollama server. Your code never leaves your machine. Inline (ghost-text) completions appear as you type; press Tab to accept.
A fill-in-the-middle (FIM) capable code model. tab.zone doesn't download models for you — pull one first. The default model is qwen2.5-coder:3b:
ollama pull qwen2.5-coder:3b
3b is a good speed/quality balance for line completion. For faster, lighter suggestions use qwen2.5-coder:1.5b; for higher quality use qwen2.5-coder:7b. Whichever you pull, set tabZone.model to match (and tabZone.fimTemplate to the model family).
Ollama FIM-capable code model to use. Smaller (1.5b/3b) = faster.
tabZone.fimTemplate
qwen
FIM prompt template matching the model family (qwen, codellama, starcoder, deepseek).
tabZone.debounceMs
250
Delay after the last keystroke before requesting a completion.
tabZone.maxPrefixChars
3000
Max characters of context taken from before the cursor.
tabZone.maxSuffixChars
1000
Max characters of context taken from after the cursor.
tabZone.maxTokens
128
Max tokens generated per completion. Lower is faster.
tabZone.maxLines
4
Max lines kept from a completion. 1 = strict single-line, 0 = no limit.
tabZone.temperature
0.1
Sampling temperature. Lower is more deterministic.
Usage
As you type, suggestions appear as grey ghost text — press Tab to accept.
Toggle completions on/off by clicking the tab.zone item in the status bar, or by running tab.zone: Toggle inline completions from the Command Palette. The status-bar icon shows the current state:
Development
npm install
npm run watch # rebuild on change
Press F5 in VS Code to launch the Extension Development Host.
npm run typecheck # tsc --noEmit
npm run package # minified production build
