Feedback now accepts photos, voice notes, video, documents etc. Phase
A captures + stores raw files (Pillow for EXIF strip); Phase B derives
OCR text, speech transcripts, poster frames, PDF text — all via
subprocess so each tool degrades gracefully if absent. Wire the
following into the shipyard service:
- python3Packages.pillow → EXIF strip on captured photos
- ffmpeg → poster frames + audio→16kHz WAV for whisper
- tesseract (eng + rus) → OCR (vyscul writes in Russian)
- whisper-cpp → speech-to-text for voice / audio / video
- poppler_utils → pdftotext for document attachments
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
