v2.5.6

🤖 Improvements

• Improved LLM provider support with expanded model capabilities, making it easier to connect and switch between your own providers.
• Cloud speech-to-text now runs on the unified transcription runtime, bringing more consistent and reliable transcription.
• Simplified speaker labeling with a flat speaker action menu, making it quicker to assign and correct speakers.

🛠️ Stability & Fixes

• Fixed a Windows audio crash by ensuring background worker threads shut down cleanly during recording.