Next Edit Suggestion is a context-aware autocomplete extension that predicts the next edit you are about to make — not the characters after your cursor — from your last edit, and offers it as inline ghost text.
Quick Start
To use NES, you must configure the following settings:
nextEditSuggestion.endpointUrl: The API endpoint for your model server.
nextEditSuggestion.model: The specific model ID you wish to use.
Note: If no model is configured, no suggestions will be provided.
Configuration Paths:
Settings Panel: File > Preferences > Settings → Search Next Edit Suggestion → Configure endpointUrl and model.
Command Palette: Ctrl+Shift+P → Next Edit Suggestion: Settings → Configure endpoint, API key, model, and format.
Key Commands
Accept: Apply the suggested edit.
Accept Inline: Insert the suggestion at the current cursor position.
Dismiss: Hide the current suggestion.
Open Menu: Access settings or trigger a suggestion via the command palette (Next Edit Suggestion: Settings).
Status Bar Indicators
The status bar provides real-time feedback on the extension's state:
(sparkle) Next Edit: Active.
(circle-outline) NES: Waiting for input.
(clock) NES Waiting: Analyzing the current context.
(sync~spin) NES Requesting: Fetching a suggestion from the server.
(warning) NES Error: A connection or request error occurred.
Configuration Options
debounceDelayMs: Delay after an edit before a suggestion is generated (Default: 250ms).
maxContextLines: Maximum context lines sent to the model (Default: 200).
maxTokens: Maximum generated tokens (Default: 256).
requestTimeoutMs: Time to wait for a server response (Default: 30,000ms).
promptTemplate: The prompt template used (Default: auto).
showStatusBar: Toggle the status bar icons.
Additional Notes
Whitespace suppression: Suggestions are not triggered by whitespace-only input (e.g., repeated space or tab presses).
