Meet Zaylo 2.0
A high-performance, 100% private local AI inference server & desktop workspace that turns your computer into a private OpenAI-compatible engine. Now featuring multimodal vision (VLMs), local document RAG, instant mobile QR pairing, and hardware benchmarks with zero cloud dependencies.
Meta Llama 3.2
Qwen 2.5 & 2.5-VL
Mistral AI
DeepSeek-R1
Microsoft Phi-3.5
Apple Silicon (Metal)
NVIDIA CUDA
Engineered for Autonomous Local AI
Everything you need to turn your laptop or workstation into an enterprise-grade AI server.
Local Document RAG
Ingest Markdown, TXT, JSON, CSV, and code files with paragraph-aware chunking and SQLite-persisted hybrid BM25 + dense cosine vector retrieval with inline citations.
Multimodal Vision (VLMs)
Run Qwen 2.5-VL and SmolVLM with automatic multimodal projector pairing (--mmproj), clipboard image paste (Cmd+V), and visual analysis.
Zero-Install Mobile QR
Point any smartphone camera at your desktop to open an encrypted Web Chat over local Wi-Fi. Chat with models, upload photos, and query your knowledge base anywhere at home.
Hardware Benchmark Suite
Automated host analyzer tests sustained generation speed (tok/s), Time-to-First-Token (TTFT), context ingestion rate, and memory bandwidth across quantization levels.
OpenAI API Gateway
Drop-in local replacement for /v1/chat/completions with streaming SSE, plus dedicated /v1/documents and RAG search endpoints.
SQLite WAL Persistence
Zero external database daemons. Embedded SQLite WAL storage keeps model catalogs, document embeddings, hardware benchmarks, and cryptographic keys secure and local.
Download Zaylo for Your System
Ready for macOS, Windows, and Linux. Zero cloud dependencies.
