EditVolt Complete
Inline completions run on a bundled local engine — no API key, no cloud round-trip, and they keep working with no internet connection at all.
How it works
As you type, suggestions stream in as ghost text:
- Tab accepts the whole suggestion
- EditVolt: Accept Next Word / Next Line of Suggestion accept part of one
- Suggestions are multi-line and consider nearby code and your recent edits
- Tab-to-jump predicts your next edit elsewhere in the file and takes you there
- Change propagation offers follow-up edits related to the change you just made
Completion context is latency-tiered — proximity and recency of your own code. It deliberately does not query the semantic index; that is reserved for chat, @Workspace, and search.
Settings
| Setting | Default | Description |
|---|---|---|
product.completions.enabled | true | Master switch |
product.completions.source | local | local, cloud, or auto. local never contacts the cloud |
product.completions.streaming.enabled | true | Stream the suggestion as it is generated |
product.completions.debounceMs | 120 | Delay before requesting a suggestion (minimum 80) |
product.completions.nextEdit.enabled | true | Tab-to-jump next-edit prediction |
product.completions.tabPredictions.enabled | true | Tab prediction surface |
product.completions.tabPredictions.cloud | auto | auto, local, or cloud for tab predictions |
product.completions.propagation.enabled | false | Offer follow-up edits related to your last change |
product.models.localTier | auto | Completion model size: auto, 7b, 14b, off |
product.models.path | (empty) | Where completion models are stored; defaults to ~/.editvolt/models/code |
:::info Local by default
Completion requests never leave your machine unless you explicitly set product.completions.source to cloud.
:::