New in SecureThink v2.0

SecureThink v2.0 is a complete, ground-up rewrite of our local AI engine. We’ve moved away from standard wrappers to build a native, standalone macOS powerhouse. Here are the major changes since v1.0:

1. Zero Dependencies (Goodbye Ollama & Terminal Setup):

   • v1.0 required installing external CLI tools like Ollama and downloading models via the terminal.

   • v2.0 is a 100% standalone desktop app. The MLX runtime and optimized model weights are built directly into our DMG installer. Drag it to your Applications folder and it works instantly.

2. Apple UMA & 8GB Mac Optimization:

   • We integrated Apple’s native MLX Core and Metal Performance Shaders (MPS) to tap directly into Apple Silicon’s Unified Memory.

   • With programmatic memory capping and aggressive post-inference cache clearing, SecureThink v2.0 runs lightning-fast on baseline 8GB MacBook Air/Pro models without lagging the OS.

3. On-Device Audio Transcription (Speech-to-Text):

   • You can now drag and drop audio recordings. A local Whisper engine transcribes them offline, indexes the text, and lets you query or summarize meetings on-device.

4. Agentic Office File Creator (Word, PPTX, Excel):

   • We added a local creation agent. Ask SecureThink to draft a contract, build a slide deck, or compile a spreadsheet, and it writes and saves native .docx, .pptx, and .xlsx files directly to your Mac.

5. Interactive In-Chat Charting:

   • SecureThink now automatically parses structured LLM data outputs to render fully interactive, responsive Bar, Line, and Pie charts right inside the chat area.

6. Local OCR (Scanned PDF Reader):

   • We embedded a local OCR engine so you can import and query scanned paper documents, images, and screenshots completely offline.