CopySpeak v0.1.5

Added

• LLM post-processing (Groq Cloud) — Optional pass between sanitize and TTS synthesis that rewrites copied text into concise, listener-friendly speech tailored for software developers. Off by default. Configure under Settings → Advanced → LLM Post-Processing.
  • New PostProcessConfig (enabled, api_key, model, prompt) in AppConfig; config schema version bumped to 0.1.5.
  • New Rust module post_process (process, try_process) wraps Groq's OpenAI-compatible /chat/completions.
  • New IPC command check_groq_credentials validates the key via GET /models.
  • Hooked into speak_now and speak_queued after the cfg snapshot, before pagination. LLM failures fall back to the original text and never block synthesis.
  • Hardcoded model dropdown: openai/gpt-oss-20b, llama-3.3-70b-versatile, llama-3.1-8b-instant.

Changed

• LLM post-processing default prompt — Switched to a terse caveman-style rewrite prompt with a 3 bullet/point maximum.

Fixed

• CopySpeak Pi extension — Routes final Pi responses through the running app's sanitization, max-length, LLM post-processing, effects, and TTS pipeline instead of filtering/truncating in the extension.
• Vercel landing page — Updated the displayed version, screenshot asset, and removed the double-copy hero tagline.