Back to site|
NexoraNexora Docs
Dashboard →
Docs/Autocomplete/Quick start

Autocomplete — Quick start

Nexora autocomplete uses DeepSeek Flash in FIM (Fill-In-the-Middle) mode to suggest contextual code in real time. It runs automatically as you type, and is available from the Starter plan up.

⚠️Autocomplete requires the Starter plan or above. On the Free plan, only chat is available.

Turning it on

1

Check you are on Starter or above

Check your plan in the dashboard → Subscription. If you are on Free, upgrade to Starter.

2

Enable it in the VS Code settings

Open the VS Code settings (Ctrl+,) et recherchez nexora. Turn on:

{
  "nexora.autocomplete.enabled": true,
  "nexora.autocomplete.triggerDelay": 300,
  "nexora.autocomplete.maxTokens": 128
}
3

Choose the target languages

By default Nexora autocompletes in every language. You can narrow that down:

{
  "nexora.autocomplete.languages": [
    "typescript",
    "javascript",
    "python",
    "rust",
    "go"
  ]
}
4

Work with the suggestions

Start typing — a greyed-out suggestion appears after the cursor. To accept it:

TabAccept the whole suggestion
Ctrl+→Accept word by word
EscDismiss the suggestion
Alt+\Force a suggestion

How it works

Mode FIM (Fill-In-the-Middle)

The extension sends the text before and after your cursor to DeepSeek. The model fills in the missing middle, which gives far more accurate completions than simply predicting what comes next.

File context

Nexora reads the imports, types and functions already defined in the current file so its suggestions fit your codebase.

Token usage

Each suggestion costs between 50 and 300 tokens depending on complexity. The trigger delay (triggerDelay) cuts down on wasted requests.

💡Raise triggerDelay to 500ms or more if suggestions fire too often and burn through your credits.

Turning it off temporarily

Click the Nexora button in the VS Code status bar (bottom right) to toggle autocomplete on and off without touching the settings.

← ChatAgent mode →