High-Performance Open-Source Developer Utilities

We build lightweight local AI inference servers, zero-config JSON vector databases, reactive statistical charting engines, and fine-tuned Text-to-SQL systems designed to run completely offline on standard hardware.

Get Zaylo Browse Releases
NEW PRODUCT RELEASE • v1.0.0

Meet Zaylo

A high-performance, 100% private local AI inference server that turns any computer into a private OpenAI-compatible AI engine running models like Llama 3.2, Qwen 2.5, and Mistral with zero cloud dependencies.

Meta Llama Meta Llama 3.2 Alibaba Qwen Alibaba Qwen 2.5 Mistral AI Mistral AI DeepSeek DeepSeek-R1 Microsoft Microsoft Phi-3.5 Apple Silicon Apple Silicon (Metal) NVIDIA CUDA NVIDIA CUDA
Latest Stable Release v1.0.0

Download Zaylo for Your System

Ready for macOS and Windows. Zero cloud dependencies.

Download for macOS (Apple Silicon) GitHub Repo v1.0.0
Available Downloads for Your Architecture
Choose your system package:
Apple Silicon
macOS Apple Silicon
M1 / M2 / M3 / M4 (.dmg)
Apple Intel
macOS Intel
x86_64 Core i5/i7/i9 (.dmg)
Microsoft Windows
Windows Installer
Windows 10 / 11 (.exe Setup)
Interactive Coding Challenge

Markup Quest

Test your frontend coding skills! Edit the HTML code below, see your visual rendering live, and select the Rocket, Bug, or Cracker tags to launch interactive global animations.

IDE Console Active
Style the Rocket 🚀
Write any HTML code you want in the editor! Use the active template to play around, modify properties, and click launch to trigger the global breakout action.
index.html editor 0 chars
Live Preview Output
Sandbox Console Info
ACTIVE OBJECT: Rocket 🚀
SYNTH AUDIO: Sawtooth Engine Sweep
TARGET ACTION: Global Viewport Launch & Blast
Releases Showroom

What We Built

Take a look at the developer utilities, open-source npm packages, and custom apps we engineered for our customers and business clients.

01
Flagship
loftyDB Database
Zero-configuration JSON database in Go. Runs SQL Joins on NoSQL collection arrays with native vector calculations.
Tap to explore →
01 — loftyDB
Key Capabilities
  • 1NPM install: npm i -g loftydb
  • 2Native Go implementation (ultra-fast ACID storage)
  • 3SQL Joins support on NoSQL JSON collections
  • 4Built-in cosine similarity vector queries
  • 5Ultra-fast file-based storage engine
  • 6Instant CLI client daemon setup
02
New
Glance-Charts
Featherweight plotting library under 8KB gzipped. Dual-engine mapping (SVG/Canvas) supporting 29 reactive visualizations.
Tap to explore →
02 — Glance Charts
Key Capabilities
  • 1Zero external dependencies (pure JS)
  • 2Featherweight footprint under 8KB gzipped
  • 3Auto SVG & Canvas dual-switching engine
  • 4Supports 29 reactive Cartesian & statistical styles
  • 5CSS: https://cdn.jsdelivr.net/npm/glance-charts/glance-charts.css
  • 6JS: https://cdn.jsdelivr.net/npm/glance-charts/glance-charts.js
03
🔒 Launching Soon
Quest NLP IDE
A next-generation, smart SQL Database IDE and Navigator powered by local AI. Designed to connect and run queries securely.
Tap to explore →
03 — Quest NLP
Key Capabilities
  • 1Natural Language English-to-SQL compiler
  • 2Multi-database connector (Postgres, MySQL, SQLite)
  • 3Local offline AI execution models
  • 4Interactive DB schema mapping visualizer
  • 5Hardware-level encrypted credentials vault
  • 6Cross-platform Electron desktop client app
AI Model Release

Caden 2.5 — Fine-Tuned Text-to-SQL Model

Caden SQL is a lightweight, privacy-first Text-to-SQL AI system that converts natural language questions into accurate, production-ready SQL queries with an automated self-correction compiler loop that detects and repairs database syntax errors on the fly.

Built upon Alibaba's Qwen2.5-Coder-1.5B-Instruct and fine-tuned on the Spider dataset (10,000+ relational text-to-SQL pairs), it is quantized into 4-bit GGUF (Q4_K_M) format to run completely offline on standard consumer CPUs and GPUs without exposing sensitive company schemas to external cloud LLMs. The workspace includes a modern Web SQL Compiler Studio alongside an interactive terminal CLI.

Base Architecture
Qwen2.5-Coder-1.5B
Training Dataset
Spider (10K+ Pairs)
Format & Precision
4-bit GGUF (Q4_K_M)
Execution Security
100% Offline (Local CPU)
Download via Ollama Registry Download GGUF on Hugging Face
Caden SQL Compiler v2.5
$ ollama run akhileshg1124/caden-sql
Connected to local instance (Model size: 1.1GB GGUF)
Input (Natural Language):

"Find the average salary of developers in Ballari location"

Generated SQL: 
SELECT AVG(salary) FROM developers WHERE location = 'Ballari';
[✓] Local SQL parser compilation success
[✓] Syntax error validation check passed (0 warnings)
Glance-Charts Showcase

Glance Charts Playground

Zero dependencies. Under 8KB gzipped. Experience our high-performance dual-engine SVG/Canvas plotting library in real-time.

Chart Configuration

Live Interactive Preview

Copied to Clipboard!