Inline and next-edit suggestions

The DataMind Oda extension can suggest code as you work in VS Code: grey ghost text at the cursor while you type, and next-edit suggestions that offer a related change after you edit. Both features are off until you turn them on, and both run on Kartlos Standard through your DataMind server.

What the extension adds to the editor is on The Oda extension for VS Code.

Turn suggestions on

1. Turn on inline suggestions first: run DataMind Oda: Toggle Inline Suggestions from the Command Palette, or use the Inline suggestions switch under General in the extension's Settings tab.
2. Turning them on checks with your DataMind server that suggestions are available to you. Oda saves the setting only when the check passes; if the check or the save is refused, the switch stays off and Oda shows the reason, listed below.
3. Then turn on Next-edit suggestions the same way. They need inline suggestions on: turned on before that, Oda notes Next-edit suggestions appear once inline suggestions are on as well. and starts them as soon as inline suggestions come on.
4. Both switches carry the Experimental tag, with the note Runs on Kartlos Standard and bills its tokens.

Inline suggestions as you type

• Oda shows ghost text at the cursor after you pause typing for a moment (250 milliseconds). Press Tab to insert it.
• A cursor that comes to rest after a move also asks, once it has rested for about half a second (500 milliseconds), so a suggestion can appear after a click or an arrow key, even when you have not typed.
• Oda asks only where writing on makes sense: at the end of a line, or before a closing bracket, quote or separator. Inside a word it stays quiet.
• A selection or several cursors ask nothing, and suggestions work in files and untitled editors.
• When you type along a suggestion, the rest of it keeps showing without a new ask.

Next-edit suggestions

• After you edit, or where VS Code reports an error or warning near the cursor, Oda can propose the next change: a fixed typo, a rename carried to the other places the name is used, a new argument at the call sites.
• Oda waits for a short pause after your edit (600 milliseconds) before it asks, so a suggestion lands once you stop typing.
• The suggestion shows only the lines it would change: the text it would remove is struck through in place, and the new lines are drawn beside it as an image, highlighted in your editor's colours and font.
• Point at the suggestion: the hint explains that Tab accepts it and Esc dismisses it.
• Tab applies the edit, replacing just the changed lines; Esc leaves the file as it is.
• When the edit is away from the cursor, a tag first offers a jump: ⇥ Tab to jump to line <N>. The first Tab moves you there, and the next accepts the edit.
• A rename is followed: after you rename something, or accept a rename suggestion, Oda offers the next use of the old name, jumping there when it is away, until none remain.

Pauses and limits

• The two features share one budget of 60 requests a minute. When the budget is used up, Oda waits before asking again; suggestions return as the minute rolls on.
• When your DataMind server asks Oda to slow down, every editor request pauses for the delay the server names. When it names none, the pause starts at 15 seconds and doubles with further refusals, up to a minute. The features stay on and return by themselves.
• The two features never show at once: when a next edit appears, the ghost text steps aside so Tab applies one thing, and ghost text waits while Oda fetches a next edit.

If something gets in the way

• Inline suggestions are not available with your organization's DataMind settings. Ask your administrator. Your organisation's settings do not offer inline suggestions to your account. Turning them on is refused; if they were already on, Oda turns them off and shows this. Ask your administrator to make the feature available.
• Sign in to your DataMind server to use inline suggestions. The extension is not signed in. Sign in to Oda in VS Code covers the steps.
• DataMind Oda did not answer. Restart it and try again. The extension's background service did not answer the request. Restart it with DataMind Oda: Restart Background Service, and try again. Extension commands, settings and limits lists the command.
• DataMind Oda could not reach your DataMind server: <error> Your server could not be reached. Check your connection and try again.
• VS Code has not loaded the inline suggestions setting yet. Run Developer: Reload Window, then turn them on again. An updated extension registers its setting only when the window reloads. Reload the window and turn the switch on again.
• VS Code could not save the inline suggestions setting: <detail> and VS Code could not save the next-edit suggestions setting: <detail>: VS Code refused to save the setting, for example when the settings file is read-only. Fix what the detail names and try again.
• Next-edit suggestions appear once inline suggestions are on as well. You turned on next edits while inline suggestions were off. Turn inline suggestions on; next edits follow.

Messages from other parts of Oda are in Troubleshooting.

Questions, answered

Why does no suggestion appear at the cursor?

Both features start off, so turn them on first. With them on, Oda asks only after a short pause and only where writing on makes sense: at the end of a line, or before a closing bracket, quote or separator. Inside a word it stays quiet, a selection or several cursors ask nothing, and next-edit suggestions appear only near your recent edits.

Why did suggestions stop for a few seconds?

The server asked Oda to slow down, or the shared minute budget ran out. Either way the pause lifts by itself and nothing is switched off; wait a moment and keep working. A slow-down pause lasts the delay the server names, or starts at 15 seconds and doubles up to a minute.

Why did Tab jump instead of changing the code?

The suggested edit was away from the cursor. The tag at the cursor reads ⇥ Tab to jump to line <N>: the first Tab moves you there, and the next Tab accepts the edit. Esc dismisses it instead.

Can ghost text and a next edit appear at the same time?

No, only one suggestion shows at a time: when a next edit appears, the ghost text steps aside so Tab applies one thing, and ghost text waits while Oda fetches a next edit. Next edits also need inline suggestions on.

How long after my edit does Oda keep suggesting the next change?

Oda follows your recent edits, about the last 30 seconds of work on the file; typing on keeps them current and a very large paste clears them. After that Oda stays quiet until you edit again, so while you only read, next edits ask for nothing.

Why does Oda say inline suggestions are not available with my organisation's settings?

Your organisation's DataMind settings leave the feature out for your account, so Oda refuses to turn it on and shows the message; if it was already on, Oda turns it off and shows the same message. Ask your administrator to make inline suggestions available.

Are suggestions billed?

Yes. Both features run on Kartlos Standard and bill its tokens, like other model use; each suggestion is one small request. Turning a switch off stops that feature's calls.

Do suggestions change my chats?

No. Suggestions are one-shot calls made for the editor, outside your chats: no chat's history or context changes, and you do not need a chat open for suggestions to work.

Why is the suggested code shown as a picture?

A next-edit suggestion shows only the lines that differ: the removed text is struck through in place, and the new lines are drawn beside it as an image, highlighted in your editor's colours and font. Nothing changes in the file until you press Tab.

Where can I see what Oda asked for, and why it skipped?

In the DataMind Oda Suggestions output channel: choose View > Output and the channel. It lists each request and its outcome, and at the Debug level it also shows the skips, such as a slow-down wait or the spent minute budget.