fix(amazonq): Faster and more responsive auto trigger UX. (aws#7763)

## Problem
1. when intelliSense suggestion is surfaced, inline completion are not
shown even if provideInlineCompletionItems have returned valid items.
2. provideInlineCompletionItems is unnecessarily debounced at 200ms
which creates a user perceivable delay for inline completion UX. We
already blocked concurrent API call.
3. The items returned by provideInlineCompletionItems, even though they
match the typeahead of the user's current editor (they can be
presented), sometimes VS Code decides to not show them to avoid
interrupting user's typing flow. This not show suggestion decision is
not emitted as API callback or events, causing us to report the
suggestion as Rejected but it should be Discard.

## Solution

1. Force the item to render by calling
`editor.action.inlineSuggest.trigger` command.
<img width="589" height="374" alt="Screenshot 2025-07-24 at 7 45 12 PM"
src="https://github.com/user-attachments/assets/f2723ab4-1d94-49e6-bae1-493c2af16bed"
/>

3. Remove the debounce

5. Use a specific command with when clause context to detect if a
suggestion is shown or not.

---

- Treat all work as PUBLIC. Private `feature/x` branches will not be
squash-merged at release time.
- Your code changes must meet the guidelines in
[CONTRIBUTING.md](https://github.com/aws/aws-toolkit-vscode/blob/master/CONTRIBUTING.md#guidelines).
- License: I confirm that my contribution is made under the terms of the
Apache 2.0 license.

Author: leigaol

packages/amazonq/.changes/next-release/Bug Fix-7261a487-e80a-440f-b311-2688e256a886.json

@@ -0,0 +1,4 @@
+{
+	"type": "Bug Fix",
+	"description": "Faster and more responsive inline completion UX"
+}

packages/amazonq/package.json

@@ -975,6 +975,10 @@
                 "command": "aws.amazonq.showPrev",
                 "when": "inlineSuggestionVisible && !editorReadonly && aws.codewhisperer.connected"
             },
+            {
+                "command": "aws.amazonq.checkInlineSuggestionVisibility",
+                "when": "inlineSuggestionVisible && !editorReadonly && aws.codewhisperer.connected"
+            },
             {
                 "command": "aws.amazonq.inline.invokeChat",
                 "win": "ctrl+i",

packages/amazonq/src/app/inline/completion.ts

@@ -2,7 +2,7 @@
  * Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
-
+import * as vscode from 'vscode'
 import {
     CancellationToken,
     InlineCompletionContext,
@@ -32,7 +32,6 @@ import {
     ImportAdderProvider,
     CodeSuggestionsState,
     vsCodeState,
-    inlineCompletionsDebounceDelay,
     noInlineSuggestionsMsg,
     getDiagnosticsDifferences,
     getDiagnosticsOfCurrentFile,
@@ -42,7 +41,7 @@ import { LineTracker } from './stateTracker/lineTracker'
 import { InlineTutorialAnnotation } from './tutorials/inlineTutorialAnnotation'
 import { TelemetryHelper } from './telemetryHelper'
 import { Experiments, getLogger, sleep } from 'aws-core-vscode/shared'
-import { debounce, messageUtils } from 'aws-core-vscode/utils'
+import { messageUtils } from 'aws-core-vscode/utils'
 import { showEdits } from './EditRendering/imageRenderer'
 import { ICursorUpdateRecorder } from './cursorUpdateManager'
 import { DocumentEventListener } from './documentEventListener'
@@ -214,13 +213,23 @@ export class AmazonQInlineCompletionItemProvider implements InlineCompletionItem
     ) {}
 
     private readonly logSessionResultMessageName = 'aws/logInlineCompletionSessionResults'
-    provideInlineCompletionItems = debounce(
-        this._provideInlineCompletionItems.bind(this),
-        inlineCompletionsDebounceDelay,
-        true
-    )
 
-    private async _provideInlineCompletionItems(
+    // Ideally use this API handleDidShowCompletionItem
+    // https://github.com/microsoft/vscode/blob/main/src/vscode-dts/vscode.proposed.inlineCompletionsAdditions.d.ts#L83
+    // we need this because the returned items of provideInlineCompletionItems may not be actually rendered on screen
+    // if VS Code believes the user is actively typing then it will not show such item
+    async checkWhetherInlineCompletionWasShown() {
+        // this line is to force VS Code to re-render the inline completion
+        // if it decides the inline completion can be shown
+        await vscode.commands.executeCommand('editor.action.inlineSuggest.trigger')
+        // yield event loop to let backend state transition finish plus wait for vsc to render
+        await sleep(10)
+        // run the command to detect if inline suggestion is really shown or not
+        await vscode.commands.executeCommand(`aws.amazonq.checkInlineSuggestionVisibility`)
+    }
+
+    // this method is automatically invoked by VS Code as user types
+    async provideInlineCompletionItems(
         document: TextDocument,
         position: Position,
         context: InlineCompletionContext,
@@ -307,17 +316,18 @@ export class AmazonQInlineCompletionItemProvider implements InlineCompletionItem
                 if (prevItemMatchingPrefix.length > 0) {
                     logstr += `- not call LSP and reuse previous suggestions that match user typed characters
                     - duration between trigger to completion suggestion is displayed ${performance.now() - t0}`
+                    void this.checkWhetherInlineCompletionWasShown()
                     return prevItemMatchingPrefix
                 }
-                getLogger().debug(`Auto rejecting suggestions from previous session`)
-                // if no such suggestions, report the previous suggestion as Reject
+
+                // if no such suggestions, report the previous suggestion as Reject or Discarded
                 const params: LogInlineCompletionSessionResultsParams = {
                     sessionId: prevSessionId,
                     completionSessionResult: {
                         [prevItemId]: {
-                            seen: true,
+                            seen: prevSession.displayed,
                             accepted: false,
-                            discarded: false,
+                            discarded: !prevSession.displayed,
                         },
                     },
                     totalSessionDisplayTime: performance.now() - prevSession.requestStartTime,
@@ -461,6 +471,7 @@ ${itemLog}
             this.sessionManager.updateCodeReferenceAndImports()
             // suggestions returned here will be displayed on screen
             logstr += `- duration between trigger to completion suggestion is displayed: ${performance.now() - t0}ms`
+            void this.checkWhetherInlineCompletionWasShown()
             return itemsMatchingTypeahead as InlineCompletionItem[]
         } catch (e) {
             getLogger('amazonqLsp').error('Failed to provide completion items: %O', e)

packages/amazonq/src/app/inline/sessionManager.ts

@@ -24,6 +24,8 @@ export interface CodeWhispererSession {
     // partialResultToken for the next trigger if user accepts an EDITS suggestion
     editsStreakPartialResultToken?: number | string
     triggerOnAcceptance?: boolean
+    // whether any suggestion in this session was displayed on screen
+    displayed: boolean
 }
 
 export class SessionManager {
@@ -49,6 +51,7 @@ export class SessionManager {
             startPosition,
             firstCompletionDisplayLatency,
             diagnosticsBeforeAccept,
+            displayed: false,
         }
         this._currentSuggestionIndex = 0
     }
@@ -128,6 +131,12 @@ export class SessionManager {
         }
     }
 
+    public checkInlineSuggestionVisibility() {
+        if (this.activeSession) {
+            this.activeSession.displayed = true
+        }
+    }
+
     private clearReferenceInlineHintsAndImportHints() {
         ReferenceInlineProvider.instance.removeInlineReference()
         ImportAdderProvider.instance.clear()

packages/amazonq/src/lsp/client.ts

@@ -352,6 +352,10 @@ async function onLanguageServerReady(
             await vscode.commands.executeCommand('editor.action.inlineSuggest.showNext')
             sessionManager.onNextSuggestion()
         }),
+        // this is a workaround since handleDidShowCompletionItem is not public API
+        Commands.register('aws.amazonq.checkInlineSuggestionVisibility', async () => {
+            sessionManager.checkInlineSuggestionVisibility()
+        }),
         Commands.register({ id: 'aws.amazonq.invokeInlineCompletion', autoconnect: true }, async () => {
             await vscode.commands.executeCommand('editor.action.inlineSuggest.trigger')
         }),

packages/amazonq/test/unit/amazonq/apps/inline/completion.test.ts

@@ -378,7 +378,7 @@ describe('InlineCompletionManager', () => {
                     )
                     await messageShown
                 })
-            describe('debounce behavior', function () {
+            describe.skip('debounce behavior', function () {
                 let clock: ReturnType<typeof installFakeClock>
 
                 beforeEach(function () {
@@ -389,7 +389,7 @@ describe('InlineCompletionManager', () => {
                     clock.uninstall()
                 })
 
-                it('should only trigger once on rapid events', async () => {
+                it.skip('should only trigger once on rapid events', async () => {
                     provider = new AmazonQInlineCompletionItemProvider(
                         languageClient,
                         recommendationService,