make internal method which does not check whether calling is allowed

and use this from cmd-l

Author: markijbema

src/services/ghost/GhostServiceManager.ts

@@ -285,7 +285,7 @@ export class GhostServiceManager {
 		const tokenSource = new vscode.CancellationTokenSource()
 
 		try {
-			const completions = await this.inlineCompletionProvider.provideInlineCompletionItems(
+			const completions = await this.inlineCompletionProvider.provideInlineCompletionItems_Internal(
 				document,
 				position,
 				context,

src/services/ghost/classic-auto-complete/GhostInlineCompletionProvider.ts

@@ -233,18 +233,6 @@ export class GhostInlineCompletionProvider implements vscode.InlineCompletionIte
 		context: vscode.InlineCompletionContext,
 		_token: vscode.CancellationToken,
 	): Promise<vscode.InlineCompletionItem[] | vscode.InlineCompletionList> {
-		const { prefix, suffix } = extractPrefixSuffix(document, position)
-
-		const matchingText = findMatchingSuggestion(prefix, suffix, this.suggestionsHistory)
-
-		if (matchingText !== null) {
-			const item: vscode.InlineCompletionItem = {
-				insertText: matchingText,
-				range: new vscode.Range(position, position),
-			}
-			return [item]
-		}
-
 		// Check if auto-trigger is enabled
 		// Only proceed with LLM call if:
 		// 1. It's a manual trigger (triggerKind === Invoke), OR
@@ -258,6 +246,27 @@ export class GhostInlineCompletionProvider implements vscode.InlineCompletionIte
 			return []
 		}
 
+		return this.provideInlineCompletionItems_Internal(document, position, context, _token)
+	}
+
+	public async provideInlineCompletionItems_Internal(
+		document: vscode.TextDocument,
+		position: vscode.Position,
+		context: vscode.InlineCompletionContext,
+		_token: vscode.CancellationToken,
+	): Promise<vscode.InlineCompletionItem[] | vscode.InlineCompletionList> {
+		const { prefix, suffix } = extractPrefixSuffix(document, position)
+
+		const matchingText = findMatchingSuggestion(prefix, suffix, this.suggestionsHistory)
+
+		if (matchingText !== null) {
+			const item: vscode.InlineCompletionItem = {
+				insertText: matchingText,
+				range: new vscode.Range(position, position),
+			}
+			return [item]
+		}
+
 		// No cached suggestion available - invoke LLM
 		if (this.model && this.ghostContext) {
 			// Show cursor animation while generating