Runnable, documented projects worth studying and extending.
OthersForgeCodePWAI programming assistant, supporting 300+ models including Claude, GPT, Deepseek, Gemini, and more. AI pair programmer running within the terminal supports code understanding, function implementation, multi-agent workflows, and MCP integration.
OthersNewton Physics EnginePWAn open-source GPU-accelerated physics simulation engine built on NVIDIA Warp is designed specifically for robotics and simulation researchers. Supports high-performance parallel physics simulation, suitable for robotics learning, digital twins, and simulation training environments.
OthersNVIDIA PersonaPlexPWNVIDIA's open-source character generation framework is used to create diverse AI personalities and characters, supporting personalized customization of dialogue systems and virtual assistants
OthersDeepScientistPWAI-driven scientific research platforms enable more powerful AI to drive cutting-edge research and support humanity's shared future. Supports automated literature analysis and research workflow orchestration.
Otherskepano/obsidian-skillsPWObsidian's official Agent skill pack teaches AI assistants to operate the Obsidian note library using Markdown, Bases, JSON Canvas, and CLI, enabling knowledge management automation
OthersTimesFMPWGoogle Research developed a time series foundational model specifically designed for time series prediction. Pre-trained models can be directly used for zero-shot prediction, supporting data from multiple fields such as finance, meteorology, and energy. This week, 4,200+ stars have grown.
OthersDeepTutorPWThe University of Hong Kong's open-source Agent-Native personalized learning assistant can adapt teaching according to students' proficiency and support multi-turn conversational tutoring. 11.6k Star, a benchmark project in the field of educational AI.
OthersBifrostPWThe fastest enterprise-grade AI gateway (50 times faster than LiteLLM), supports adaptive load balancing, cluster mode, guardrail mechanism, compatible with 1000+ models, and has latency below 100μs.
OthersTaxHackerPWSelf-hosted AI bookkeeping application uses LLMs to automatically analyze receipts, invoices, and transaction records, supporting custom categories and prompts. This week, it gained 1,600+ stars.
OthersqmdPWA lightweight local document semantic search CLI tool supports quick retrieval of knowledge bases, meeting minutes, and any document, tracking the latest SOTA solutions and running fully locally
Efficiency toolsObsidian Agent skill packPWObsidian's official Agent skill pack teaches AI assistants like Claude Code and Codex to operate Obsidian Markdown, Bases database, JSON Canvas, and CLI. Follow the Agent Skills Open Standard.
OthersNVIDIA PersonaPlexPWNVIDIA's open-source personalized AI character generation framework supports the creation of AI personalities with unique personalities, memories, and behavioral patterns. 7.9k Stars, up 1.7k this week, is a key tool in AI role-playing and virtual assistant sectors.
OthersPlanoPWAI-native proxy gateway – Agent applications are equipped with built-in orchestration, security protection, observability, and intelligent LLM routing for proxy data planes
OthersKoharuPWA machine learning-based comic translation tool written in Rust, supporting automatic bubble text detection, OCR recognition, and translation rendering. High-performance offline operation.
Othersmlx-vlmPWOn Mac, Apple MLX inference and fine-tuning of the Visual Language Model (VLM) support local operation of mainstream multimodal models
OthersBifrostPWAn enterprise-grade AI gateway 50 times faster than LiteLLM, written in Go, adaptive load balancing, cluster mode, supports 1000+ models, and at 5000 QPS, it adds less than 100μs latency
OthersBifrostPWThe fastest enterprise-grade AI gateway (50 times faster than LiteLLM), supporting adaptive load balancing, cluster mode, guardrails, 1000+ models, and latency below 100 microseconds (5k RPS). Written in Go, suitable for high-performance scenarios.
OthersGoogle ADK GoPWGoogle's official Go AI Agent development kit offers a code-first approach to building, evaluating, and deploying complex AI Agents, supporting flexible model integration and tool calls
OthersDaftPWA high-performance data engine for AI and multimodal workloads, capable of processing images, audio, video, and structured data, supporting distributed data processing of any scale
OthersMemvidPWAI Agent's Memory Layer — replacing complex RAG pipelines with a single file, providing agents with instant retrieval and long-term memory capabilities. Written in Rust, extremely high performance, no server required.
OthersKoharuPWA machine learning-based comic translator, written in Rust, running high-performance locally. Automatically detects bubble text, OCR recognition and translation, supports multiple languages.
OthersNVIDIA PersonaPlexPWNVIDIA's open-source character generation framework for creating and managing AI characters/personas, supporting multi-role interaction and personalized conversation generation.
OthersMCP Toolbox for DatabasesPWGoogle's open-source MCP database server allows AI Agents to connect directly to enterprise databases. Supports ready-to-use pre-build tools and custom tool frameworks, allowing integration into frameworks like ADK and LangChain within 10 lines of code.
AI and AgentsJCode - Ultra-high-performance encoding agent frameworkPWJCode is a next-generation coding agent framework designed for multi-session workflows and extreme performance. Built-in semantic vector memory system, boot speed 245 times faster than Claude Code, with 10 sessions occupying only 117MB of memory (Claude Code occupying 2.3GB).