Description:
Self-hosted, multi-model AI inference server that runs LLMs, TTS, STT, embeddings, and image generation behind a single OpenAI-compatible API, with GPU memory budgeting and a plugin system.
Data Science, AI and Machine Learning
Qlib
Description:
Qlib is an AI-oriented quantitative investment platform providing data processing, ML modeling, backtesting, and deployment workflows; a framework for quantitative research with supervised learning, market dynamics modeling, and reinforcement learning.
Octopoda-OS
Description:
Octopoda-OS is an open-source memory operating system for AI agents, offering persistent memory, semantic search, loop detection, agent messaging, crash recovery, and real-time observability via local (SQLite) or cloud (PostgreSQL) storage.
Thunderbolt
Description:
Thunderbolt is an open-source, cross-platform AI client that you can self-host. It supports multiple models and providers, is enterprise-ready, and emphasizes data sovereignty and vendor lock-in avoidance.
MOSS-TTS-Nano
Description:
MOSS-TTS-Nano is an open-source multilingual tiny speech generation model from MOSI.AI and the OpenMOSS team. With only 0.1B parameters, it is designed for realtime speech generation, can run directly on CPU without a GPU, and keeps the deployment stack simple enough for local demos, web serving, and lightweight product integration.
Atomic Agents
Description:
Lightweight Python framework for building modular, agent-based AI pipelines with context providers and tools.
ShellGPT
Description:
ShellGPT is a command-line productivity tool powered by AI language models that generates shell commands, code snippets, and documentation. It helps automate tasks across Linux, macOS, Windows, and major shells, enabling faster, more efficient workflows.
Lightpanda
Description:
Lightpanda is an AI-native, headless browser built from scratch for machines. Designed for fast, memory-efficient web automation, it offers instant startup and embeddable APIs to power AI agents and automated workflows—an alternative to Chrome headless.
BitNet
Description:
BitNet is the official inference framework for 1-bit LLMs, providing optimized CPU/GPU kernels for fast, lossless 1-bit model inference. It includes demo models, build scripts and integration with llama.cpp for efficient 1-bit LLM inference.
obra/superpowers
Description:
Agentic skills framework and software development methodology for subagent-driven development.