Release v2.2.1

Text Structurer Quote & Voice Fix Update 🎙️

This update improves the .fix Text Structurer command by adding support for Telegram reply quotes and voice-based inputs. It also adds safety checks around empty Gemini requests and ensures source voice messages are deleted only after successful processing.

✨ New Features

• Reply Quote Support for .fix: The structurer now extracts text from Telegram reply quotes, including quote_text, quoteText, and nested reply quote objects.
• Voice-to-Structured-Text via .fix: Replying to a voice message, audio file, or video note with .fix now transcribes the media and rewrites it into a clean, structured text.
• Safe Source Voice Deletion: Original voice/audio/video-note messages are deleted only after transcription, AI rewriting, and final message edit complete successfully.
• Temporary Processing Status: When .fix is used on voice media, the trigger message is first changed to a processing status, then replaced with the final structured text.

🔧 Text Structurer Improvements

• Centralized Text Extraction: Added helper logic to safely extract normal text, captions, and quoted reply text from Pyrogram/Kurigram message objects.
• Transcribable Media Detection: Added detection for voice, audio, and video_note messages inside text_struct.py.
• Media Download & Cleanup: .fix now downloads replied voice media into data/ts_fix_*, sends it to Gemini transcription, and removes the temporary file afterward.
• Edit Success Tracking: _apply_edit() now returns whether the final AI-generated edit actually succeeded, allowing safe follow-up actions like deleting the source media.

🛡️ Stability & Protection

• Empty Gemini Request Guard: Gemini calls are now skipped when both prompt text and media are empty, preventing repeated 400 INVALID_ARGUMENT: Request has empty input fallback spam.
• Fallback Safety Preserved: If transcription or AI rewriting fails, the original voice message is not deleted.
• Quote Fallback Logic: .fix now tries quote text first, then falls back to full message text/caption.
• Reduced Log Noise: Empty requests now produce a single warning instead of cycling through every Gemini fallback model.

🗂️ Changed Files

• modules/text_struct.py

  • Added quote extraction helpers.
  • Added voice/audio/video-note handling for .fix.
  • Added media transcription flow.
  • Added safe source media deletion after successful final edit.
  • Updated _apply_edit() to return success state.

• core/services.py

  • Added empty prompt/media guard before Gemini fallback execution.
  • Updated content building to avoid sending empty text parts to Gemini.

Full Changelog: v2.2.0...v2.2.1