AI Features Re-implementation Plan (Based on PR #780)

[Anxcye]

AI Features Re-implementation Plan (Based on PR #780)

Background

PR #780 by @gwaanl contains a massive amount of AI feature improvements (147 files, +22,499/-1,388 lines). After careful review, I've decided not to merge this PR directly due to:

1. Very large scope - difficult to review and test thoroughly
2. Design philosophy differences - some changes don't align with the project's existing architecture (e.g., AI-only translation)
3. Risk of instability - too many changes at once could introduce bugs

However, the features are valuable and well-implemented. This issue tracks the plan to re-implement these features incrementally with proper testing and architectural alignment.

Credits

All features listed below are based on the excellent work by @gwaanl in PR #780. During re-implementation, I will:

• Reference the original implementation
• Maintain code quality and project consistency
• Credit @gwaanl in commit messages

Core Features to Re-implement

Phase 1: AI Provider Infrastructure (Priority: High)

• AI Provider Center

  • Provider metadata storage system
  • Provider Center UI (CRUD operations)
  • Support for OpenAI-compatible, Claude, Gemini, DeepSeek
  • API key management
  • Multi API key rotation (round-robin)
  • API key auto-save with failure stats and cooldown

• Model Management

  • Fetch and cache models for each provider
  • In-chat model quick switch
  • Manual model entry support

Phase 2: Enhanced AI Chat Experience (Priority: High)

• Thinking Mode

  • Thinking level UI (off/auto/minimal/low/medium/high)
  • OpenAI Responses provider support
  • Gemini thinking wrapper
  • Claude thinking integration
  • Collapsible Answer/Thinking/Tools sections

• Conversation Management

  • Conversation tree persistence (conversationV2)
  • Branchable chat history
  • Per-turn assistant variant switcher
  • Edit user message + regenerate from any turn
  • Rollback branches via variant switcher

• Chat UI Improvements

  • Keep streaming while minimizing sheet
  • Persistent bottom sheet in reading page
  • Swipe-to-minimize functionality
  • Move chat streaming into provider (survive minimize/close)

Phase 3: Translation Enhancements (Priority: Medium)

Note: Keep multi-service translation UI, but enhance AI translation as an option

• Inline Full-Text Translation

  • AI-specific prompt for full-text translation
  • Block-level translation units
  • Viewport queue + HUD progress indicator
  • Batch JSON translation
  • Translation cache and concurrency control

• Translation Quality & Reliability

  • Auto-retry failed blocks
  • Stricter success detection
  • Failure reason tracking
  • Manual retry support
  • Prioritize current viewport translations

• Translation UI

  • AppBar translation toggle
  • 3-mode toggle icons (off/original/translated)
  • Translation status in reading settings
  • Structural translation for links (preserve ``)

• AI Translation Options (NOT AI-only)

  • AI provider override for translation
  • AI model override for translation
  • Keep existing multi-service support

Phase 4: Developer & Debug Features (Priority: Low)

• Developer Options

  • AI diagnostic logs toggle
  • Vibration test page
  • Better developer settings organization

• Sync & Backup

  • Include multi-provider encrypted API keys in v4 backup
  • Sync translate preferences

Implementation Strategy

1. One feature at a time - each PR focuses on a single feature or small feature set
2. Thorough testing - test each feature before moving to the next
3. Architecture alignment - ensure new code follows project conventions
4. Incremental rollout - users can benefit from features as they're completed
5. Proper documentation - update docs and changelog for each feature

Timeline Estimate

• Phase 1: 2-3 weeks
• Phase 2: 3-4 weeks
• Phase 3: 2-3 weeks
• Phase 4: 1 week

Total: ~2-3 months (part-time development)

Related

• Original PR: feat(ai+translate): Provider Center, chat UX, Responses provider, WebDAV sync, EPUB inline translation #780
• Original Author: @gwaanl

Notes

This is a living document. Feature list and priorities may change based on:

• User feedback
• Technical challenges discovered during implementation
• New feature requests

---

Discussion welcome! If you have suggestions on prioritization or implementation approach, please comment below.