Skip to content

editor.md

Latest commit

 

History

History
69 lines (52 loc) · 2.77 KB

editor.md

File metadata and controls

69 lines (52 loc) · 2.77 KB
Raw

Editor Extension (VS Code & forks)

The LLMStack extension brings your local gateway into the editor: a chat sidebar, AI edits you review as a diff before they touch your code, inline completions, and one-click explanations — all talking to your gateway. Code never leaves your machine.

Install

=== "VS Code"

Install **LLMStack — Local AI** from the Visual Studio Marketplace, or:

```
code --install-extension llmstack.llmstack-vscode
```

=== "Cursor / Windsurf / VSCodium"

These VS Code forks cannot use Microsoft's marketplace; they use the
[Open VSX Registry](https://open-vsx.org). Search **LLMStack** in the
Extensions view, or:

```
# Cursor / Windsurf / VSCodium all accept VSIX installs too:
codium --install-extension llmstack.llmstack-vscode
```

=== "From source"

```
cd editors/vscode
npm install
npm run package          # produces llmstack-vscode-*.vsix
code --install-extension llmstack-vscode-*.vsix
```

!!! tip "First, start the gateway" The extension talks to a local gateway. Run llmstack up (or llmstack serve) first. The status-bar item shows whether the gateway is reachable.

Features

Surface What it does
Chat sidebar Streaming chat in the activity bar, with markdown + code blocks. Pick the model, optionally include the active selection/file as context.
Apply / Insert / Copy Every code block in a reply can be applied over your selection, inserted at the cursor, or copied.
Edit with AI (Cmd/Ctrl+Alt+E) Select code, describe a change, and review the result as a native diff — nothing is written until you approve.
Revert last AI edit One command restores the file to its pre-edit state.
Inline completion Opt-in ghost-text completions from your local model.
Ask / Explain (Cmd/Ctrl+Alt+A) Quick questions and selection explanations.
👍 / 👎 feedback Rate replies; feedback flows into the gateway's adaptive-learning pipeline.

Settings

Setting Default Description
llmstack.gatewayUrl http://localhost:8000 Base URL of your local gateway.
llmstack.apiKey "" API key, only if gateway.auth=api_key.
llmstack.model llama3.2 Default model (override per-chat in the panel).
llmstack.inlineCompletion.enabled false Enable ghost-text completions.
llmstack.inlineCompletion.contextLines 50 Lines of surrounding context to send.
llmstack.inlineCompletion.maxTokens 200 Max tokens per completion.

Privacy

Every request goes to the gateway URL you configure — by default localhost. Run llmstack verify-private to audit that nothing in your configuration sends data off the machine.