v1.1.0

What's new

Configurable local models (Systemstatus panel)

• The Ollama chat model (used for deep-check and summarization) and the faster-whisper transcription size are no longer hardcoded — pick from a curated, RAM/VRAM-labeled list or enter any other locally available model as free text. Choices persist across restarts; an OLLAMA_MODEL/WHISPER_MODEL_SIZE env var still takes precedence over the UI choice (e.g. for the Docker setup).

Numbered name redaction

• The category review's name-handling toggle now has three modes: Schwärzen (generic [PERSON]), the new Nummerieren ([PERSON1], [PERSON2], ... — keeps multiple people in a document distinguishable without exposing any name), and Pseudonymisieren (consistent fake names, as before). Placeholder/pseudonym assignment now follows true reading order.

Fixes

• Removed the ffmpeg-on-PATH dependency check: faster-whisper decodes audio via PyAV's bundled FFmpeg libraries and never actually needed a system ffmpeg binary. The old check could falsely report "missing" in the packaged macOS app (Finder-launched apps don't inherit Homebrew's PATH), even with ffmpeg correctly installed.
• The Systemstatus "Reparieren" button now surfaces the actual install hint (e.g. brew install ffmpeg, ollama pull <model>) inline instead of silently doing nothing for dependencies that require manual installation.

Docs

• README: added a section on opening the unsigned macOS build (Gatekeeper) and its cause (no paid Apple Developer account).

---

📖 Full (German) user guide: ANLEITUNG.md