Open source tools and frameworks.
Developer tools and frameworks we build in the open. Each one runs on hardware you own, and its code is on GitHub.
Tools and frameworks
Bud Decision Studio
The LM Studio for decision models. A desktop app to learn, infer, train, template, serve, and integrate open decision models on your own hardware.
WaaV Gateway
A real-time voice AI gateway. One WebSocket and REST API for speech-to-text and text-to-speech across 70+ providers, plus full-duplex speech-to-speech.
Bud Simulator
Simulates LLM inference, serving, and training on any hardware: TTFT and TPOT, memory and KV-cache sizing, SLO percentiles, power, and parallelism search.
Bud Studio
A desktop AI workspace built around Bud Agent, an autonomous assistant that edits local files, runs code, and searches the web and your knowledge sources, with any LLM.
LayerZero
Kernel selection and dispatch for PyTorch LLM inference. One API picks the right kernel across GPU-optimized libraries, so each inference engine does not have to reimplement it.
GPU-Virt-Bench
Benchmarks software GPU virtualization, such as HAMi-core and Bud FCSP, against ideal MIG behavior: 56 metrics across 10 categories, measured on LLM-like workloads.
Bud Scaler
A Kubernetes autoscaler for GenAI workloads: GPU-aware scaling strategies, native vLLM, TGI, SGLang, and Triton metrics, and learned traffic patterns.
StyleBench
A multimodal benchmark that scores the outfits an LLM agent creates, or the prompts a RAG-style application discovers to create them.
Bud Studio Notebook
An Electron and React app that combines an AI-powered editor, project management, and a personal notebook, with tables, kanbans, and lists over the same data.