fix(home): voice input text persistence bugs

[waleedlatif1]

Summary

• Fix deleted text reappearing when re-speaking after clearing the textarea while mic is active
• Fix text disappearing after a long pause in speech (recognition auto-restart was losing accumulated text)
• Extract startRecognition and restartRecognition helpers for cleaner speech lifecycle management
• Add try/catch on initial recognition.start() for robustness

Type of Change

• Bug fix

Testing

Tested manually

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[cursor]

PR Summary

Medium Risk
Touches the chat input’s speech-recognition lifecycle and how it mutates the textarea value, which can regress voice input behavior across browsers despite being UI-scoped.

Overview
Fixes voice-input text persistence in UserInput by making speech recognition restarts preserve the latest typed/cleared textarea state instead of reusing stale prefixes.

This extracts startRecognition/restartRecognition, tracks the current message in a ref, restarts recognition on submit and on manual text edits (including the special @ handling), and adds guarded try/catch around recognition.start() plus safer onend/onerror handling.

^{Written by Cursor Bugbot for commit 8549d3c. Configure here.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		Mar 24, 2026 4:43pm

[greptile-apps]

Greptile Summary

This PR fixes two speech-recognition persistence bugs in the UserInput component: accumulated text being lost after recognition auto-restarts on a long pause, and cleared/deleted text reappearing when the user speaks again. Both are solved by extracting startRecognition/restartRecognition helpers and introducing a valueRef to give async Web Speech API callbacks reliable access to the latest textarea value without stale closure issues.

Key changes:

• valueRef (useRef) shadows the value state and is kept in sync by a useEffect plus a direct write inside onresult, ensuring onend's auto-restart reads the correct accumulated text when it updates prefixRef.current.
• startRecognition returns a boolean so toggleListening and restartRecognition can gate setIsListening(true) on actual start success — resolving the previously reported setIsListening(true) race on start failure.
• restartRecognition(newPrefix) is called from handleInputChange (every textarea edit) and handleSubmit (on send), keeping prefixRef in sync with user edits so stale prefixes can never cause old text to reappear.
• Old recognition instances are guarded in onerror and onend via reference equality (recognitionRef.current !== recognition) so aborted sessions cannot mistakenly clear isListening or trigger spurious restarts.
• SPEECH_RECOGNITION_LANG extracted as a module-level constant (minor hygiene).

Confidence Score: 4/5

• This PR is safe to merge; the bug fixes are logically sound and the previously reported setIsListening race is correctly resolved.
• The core implementation is correct: valueRef prevents stale reads in the async onend handler, reference-equality guards on onerror/onend prevent ghost-restart side effects, and the boolean return from startRecognition properly gates the listening state. The one remaining non-blocking item is that restartRecognition is called on every keystroke, which may cause noticeable churn under rapid editing; a debounce would be a cleaner approach but is not required for correctness.
• No files require special attention beyond the single changed component.

Important Files Changed

Filename	Overview
apps/sim/app/workspace/[workspaceId]/home/components/user-input/user-input.tsx	Refactors speech recognition into startRecognition/restartRecognition helpers; adds valueRef to avoid stale closure in onend; fixes prefix persistence bugs on clear and post-submit; gates setIsListening(true) on start success per previous review feedback.

Sequence Diagram

sequenceDiagram
    participant U as User
    participant TC as toggleListening
    participant SR as startRecognition
    participant RR as restartRecognition
    participant WA as Web Speech API
    participant RS as React State

    U->>TC: clicks mic button (start)
    TC->>SR: startRecognition()
    SR->>WA: new SpeechRecognitionAPI()
    SR->>WA: recognition.start()
    WA-->>SR: success → returns true
    SR-->>TC: true
    TC->>RS: setIsListening(true)

    WA-->>SR: onresult(event)
    SR->>RS: setValue(prefix + transcript)
    SR->>SR: valueRef.current = newVal

    note over WA: Long pause / natural end
    WA-->>SR: onend()
    SR->>SR: prefixRef.current = valueRef.current
    SR->>WA: recognition.start() (same instance)

    U->>U: types in textarea
    U->>RR: handleInputChange → restartRecognition(newValue)
    RR->>RR: prefixRef.current = newValue
    RR->>WA: recognitionRef.current.abort()
    WA-->>SR: onerror('aborted') → guard fails → early exit
    WA-->>SR: onend() → guard fails → early exit
    RR->>SR: startRecognition()
    SR->>WA: new instance + start()
    WA-->>SR: success

    U->>U: submits message
    U->>RR: handleSubmit → restartRecognition('')
    RR->>RR: prefixRef.current = ''
    RR->>WA: abort old instance
    RR->>SR: startRecognition()
    SR->>WA: new instance + start()

    U->>TC: clicks mic button (stop)
    TC->>WA: recognitionRef.current.stop()
    TC->>RS: setIsListening(false)

Loading

_{Reviews (3): Last reviewed commit: "fix(home): reset speech prefix on submit..." | Re-trigger Greptile}

[waleedlatif1]

@greptile

[waleedlatif1]

@cursor review

[waleedlatif1]

@greptile

[waleedlatif1]

@cursor review

[waleedlatif1]

@greptile

[waleedlatif1]

@cursor review

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR