Skip to content

Implement AI-powered Reply Drafter#2257

Closed
enaboapps wants to merge 1 commit into
mainfrom
feature/reply-drafter-2256
Closed

Implement AI-powered Reply Drafter#2257
enaboapps wants to merge 1 commit into
mainfrom
feature/reply-drafter-2256

Conversation

@enaboapps
Copy link
Copy Markdown
Collaborator

Summary

This PR implements the AI-powered Reply Drafter feature, allowing users to generate context-aware reply suggestions based on the current screen content. It leverages the Gemma 3 model via MediaPipe GenAI tasks to provide high-quality, on-device text generation.

Key changes:

  • Integrated MediaPipe GenAI tasks for on-device LLM processing.
  • Created ReplyDrafterManager to orchestrate screenshot capture, text extraction, and AI suggestion generation.
  • Implemented ReplyDrafterHUD, a Compose-based overlay for displaying and selecting AI suggestions.
  • Added "Reply Drafter" option to the Main Menu for easy access.
  • Enhanced accessibility by allowing users to insert selected suggestions directly into focused text fields.

Test plan

  • Ensure Gemma 3 model is at /sdcard/Download/gemma-3n-E2B-it-int4.task.
  • Open Main Menu and select "Reply Drafter".
  • Verify screenshot is taken and processing message appears.
  • Verify AI suggestions appear in the overlay.
  • Verify selecting a suggestion inserts text into a focused field.

- Add MediaPipe GenAI dependency for on-device LLM inference
- Create LlmManager for Gemma 3 vision-modality inference
- Create ReplyDrafterManager to coordinate screenshot-to-reply flow
- Implement ReplyDrafterHUD for displaying and inserting suggestions
- Add Reply Drafter action to the Main Menu

🤖 Auto-generated

Co-authored-by: Cursor <cursoragent@cursor.com>
@OwenMcGirr OwenMcGirr closed this May 21, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants