OpenAI Translator: download ældre versioner

Fix newer GPT-5 models being sent an unsupported parameter

• Fixed newer GPT-5 models (such as gpt-5.0) being sent a reasoning effort setting they no longer support, which could make requests fail (#1853)
• Refreshed the built-in fallback model list, shown when the model list cannot be fetched from the API, to lead with the GPT-5 series instead of GPT-4 models

v0.6.19
• Fix excessive idle CPU usage in WebView2 on Windows since 0.6.16. The writing-indicator HUD and Quick Translator panel are pre-created as hidden windows at startup; the HUD ran infinite CSS animations from launch (even while in the system tray) and the Quick Translator eagerly ran its context picker. Both now stay fully idle until actually shown (#1883).

v0.6.19
- Fix excessive idle CPU usage in WebView2 on Windows since 0.6.16. The writing-indicator HUD and Quick Translator panel are pre-created as hidden windows at startup; the HUD ran infinite CSS animations from launch (even while in the system tray) and the Quick Translator eagerly ran its context picker. Both now stay fully idle until actually shown (#1883).

v0.6.18
- Fix Ollama / LM Studio not being able to disable thinking (e.g. Qwen3) after 0.6.17. The 0.6.17 fix gated reasoning_effort: none to OpenAI GPT-5.1+ only, which regressed local OpenAI-compatible servers; the Ollama engine now keeps sending it to turn thinking off, while strict providers (DeepSeek, etc.) remain unaffected (#1881).

v0.6.17
- Fix translation failure on DeepSeek and other non-OpenAI providers caused by reasoning_effort: none being sent to APIs that don't support it (#1879, #1880). The thinking-disable override is now only applied to models that accept none (OpenAI GPT-5.1+).

feat: add keyboard shortcut for "Insert into previous input" (#1878)
feat: add thinkingEnabled setting to control reasoning model usage (#1866)

feat: add compact inline lookup mode for text selection translation (#1871)
feat: upgrade MiniMax default model to M2.7 (#1860)

fix: improve Edge TTS reliability and error handling
fix: remove nested actionButton wrappers causing double hover effect

feat: refine UI/UX with elegant transitions, softer shadows, and polished details

fix: replace unwrap/expect with proper error handling in OCR module
Changes:
- Replace all unwrap() with proper error handling in OCR module
- Add binary existence check on macOS before executing OCR
- Handle APP_HANDLE initialization failures gracefully
- Log errors instead of panicking in production
Fixes OCR panic in production builds.

ci: update macOS runner from macos-13 to macos-latest

release: v0.6.5

release: v0.6.4

feat: Add Cerebras support (#1828)

release: v0.6.2

fix: keep history window open after restore (#1823)
fix: persist window state per label (#1822)

fix: guard tauri integrations in shared settings (#1819)

fix: skip textarea highlighting when idle (#1817)
fix: disable clipboard fallback for selected text command (#1818)

fix: ensure writing sessions release lock (#1816)

fix: prevent duplicate writing shortcut triggers (#1815)

fix: suspend history queries when hidden (#1813)

fix: unblock windows focus build (#1800)
fix: add Google Generative AI integration and update safety settings (#1799)
fix: type insertion without clipboard (#1798)
feat: add translation history (#1797)
feat: insert translation into previous window (#1796)
docs: add repository guidelines (#1795)

refactor: upgrade Tauri to v2 stable version (#1790)

feat: optimize prompts (#1634)

feat: support kimi (#1427)