Bonsai Demo

This repository provides code for machine learning algorithms for edge devices developed at Microsoft Research India.

Generate images locally

🌱 a tiny distro-independent package manager

OxiBonsai is a zero-FFI, zero-C/C++ inference engine for PrismML's sub-2-bit Bonsai family — both the 1-bit line (Q1_0_g128) and the ternary line (TQ2_0_g128). It runs on CPU (SIMD), Apple Silicon (Metal), and NVIDIA (CUDA) without depending on llama.cpp, BLAS, or any C/Fortran runtime.

Ultra-Sparse Adaptation of 1-Bit LLMs via XOR Patches

Ternary bonsai 27b gguf local LLM inference using optimized 1.58bit weights. Derived from Qwen3.6-27B, it supports multi-step reasoning, agentic workflows, and a 262K context window directly on consumer hardware. Download direct Hugging Face repository model configuration files.

73 deterministic Claude AI skills for Blender, Bonsai, IfcOpenShell and Sverchok. AEC Python development skill package

Read manga from the comfort of your terminal

🌿 Prune your AI agent's context window. Reduce token usage by 70-95% with hierarchical memory. Replace flat MEMORY.md with a bonsai-shaped domain tree. Progressive disclosure, zero dependencies. Works with OpenClaw and any LLM agent framework.

HIP/ROCm fork optimized for AMD RDNA2 (gfx1030) with PrismML Q1_0_G128 1-bit quant support, RotorQuant, TurboQuant, EAGLE3 and P-EAGLE speculative decoding, and full Wave32 kernel optimizations.

An Avanade/Accenture open collaborative tutorial to distribute knowledge and capability regarding the use of Azure IoT and Azure Digital Twins with Simulation to train Microsoft Bonsai AI to solve manufacturing business problems.

A jekyll theme for semantically inclined digital gardeners.

Bonsai's Slate Developer Documentation

Code for Optimized Arrhythmia Detection on Ultra-Edge Devices

tool-eval-bench results for Prism ML Ternary-Bonsai-27B (Q2_0) — 8 trials, score 85, deployability 80

A from-scratch, dependency-light, dequant-free Bonsai (Qwen3-architecture) Q1_0 inference engine running on wgpu compute shaders

Deterministic BIM compiler - extracts patterns from IFC databases, generates geometry with validated connections

Stream real time Tweets of current affairs like covid-19 using Kafka 2.0.0 high throughput producer & consumer into Elasticsearch using safe, idempotent and compression configurations. Aggregate the data and use it for further analytics.

AMD ROCm (gfx1030) inference fork with RotorQuant/TurboQuant KV compression, PHANTOM-X zero-copy draft speculation, EAGLE3 speculative decoding, 12 RDNA2 crash fixes, and PrismML Bonsai Q1_0_G128 1-bit GGUF support.