VecML vs MyLLMChat

VecML APP offers tools for intelligent photo search and large language models (LLMs) fully on device. Users can select from a range of small LLMs and engage in conversations with documents stored locally. The app can process files (PDFs, DOC documents, images, etc.) to generate embedding vectors, which are stored locally on the device. These embedding vectors are crucial for RAG (retrieval-augmented generation) capabilities, enhancing the LLM's ability to provide more accurate and relevant responses when queries are related to the local files.

MyLLMChat is a fully local AI chat app deeply optimized for Apple Silicon. All inference runs on your device — no chat content, prompts, or personal data ever leaves your Mac. Privacy is the default, not an upgrade. Highlights: • Fully offline — no internet required after a model is downloaded • Broad model support: Llama 3, Gemma 2, Qwen 2.5, DeepSeek, Mistral, and more • Built on mlc-llm + Metal — squeezes every drop of performance out of your M-series chip • Multimodal: image understanding, plus PDF / Word / Markdown document RAG with cited sources • Voice mode: on-device Whisper STT and high-quality TTS for natural voice conversations • Rich rendering: Markdown, LaTeX math, Mermaid diagrams, syntax highlighting for 50+ languages • End-to-end encrypted iCloud sync across Mac, iPhone, and iPad • Built-in OpenAI-compatible local API server — wire your local model into any third-party tool • Pro sampling controls: Temperature, Top P, Top K, Min P, frequency / presence / repeat penalties Built for lawyers, doctors, researchers, developers, and creators — anyone who needs serious AI help without handing data to the cloud.