AI Resources Archive

Fosowl/agenticSeek is an open-source agentic search tool supporting various AI Agent scenarios.

A set of structured workflow templates and prompts that help developers collaborate with AI from requirements to working code.

An open-source operating system for AI agents providing runtime, tooling, and ecosystem integrations.

An open-source Android AI operator that understands natural language and performs UI actions on behalf of the user.

An agentic AI software engineer that decomposes high-level objectives into tasks and assists code generation and repair.

A general memory framework for agents that combines just-in-time deep research with long-term structured storage.

An open-source agentic Machine Learning Engineer designed to automate experiments, training, and deployment workflows.

A multi-feature novel generation tool based on large language models that supports multi-chapter generation and storyline consistency maintenance.

Archon is the knowledge and task management backbone for AI coding assistants, providing smart search, context management, and real-time collaboration.

A database-centric static analysis and code intelligence platform that delivers precise data-flow tracking and architecture-aware inspections.

An open-source project for automating documentation generation by combining content sources and model capabilities.

An Electron-based real-time desktop GUI for DeepSeek-OCR, enabling fast local recognition and batch processing.

Document image parsing model using heterogeneous anchor prompting; provides efficient page-level and element-level parsing capabilities.

Gitingest turns any Git repository into a prompt-friendly text digest to help LLMs ingest and reason about codebases efficiently.

A command-line tool that renders and browses Markdown beautifully in the terminal.

An open-source developer companion that generates, debugs and incrementally builds runnable applications using LLMs.

Kiro IDE: Transform prototypes into products with specs-driven development, smart hooks, and AI assistance for seamless coding and collaboration.

A benchmark suite and toolkit to evaluate AI agents on machine learning engineering tasks.

A lightweight open-source command-line utility for everyday PDF tasks such as merging, splitting, compressing, and extracting content.

An open-source framework for verifiably private inference that enables auditable model inference while protecting data privacy.

A scientific PDF translation tool that preserves layout and mathematical formulas while supporting multiple translation backends.

A framework that enables multimodal models to operate a computer by observing the screen and issuing mouse/keyboard actions, supporting voice, OCR and multiple model backends.

Smart Excalidraw converts natural language into editable Excalidraw charts using large language models.

A benchmark and execution harness for evaluating AI agents in real terminal environments.

A general-purpose agent built on large language models to automate software engineering tasks and boost developer productivity.

An AI tool focused on handling files and code, capable of understanding entire project structures and performing complex code operations.

An open-source implementation of self-reflective, hierarchical agents designed for large-scale API calls and evaluation.

A single-file browser tool that locally assembles selected files into an LLM-friendly prompt.

A Neo4j-based MCP server providing project, task, and knowledge management for LLM agents.

An open-source framework for market analysis, risk management, and trading execution via swarm intelligence and agent automation.

Model Context Protocol server implementation providing browser automation capabilities for AI assistants to perform web research, data extraction, and automated testing.

Open-source end-to-end Visual Language Model (VLM) GUI agent supporting bilingual (Chinese/English) and screen-aware input.

LaVague is an open-source Web Agent framework offering multi-driver support and customizable generation and execution pipelines.

A low-code framework for building multi-agent workflows, integrations, and data-oriented LLM applications.

ColiVara is a visual embedding-based document storage, retrieval, and RAG platform supporting unified image and text retrieval.

EasyVoice is an open-source text-to-speech tool supporting ultra-long text, streaming playback, and multi-role dubbing.

Open-source audio foundation model supporting multimodal audio capabilities for understanding, generation, and dialogue.

High-quality, multi-lingual open-source text-to-speech library supporting Chinese, English, Japanese, Korean, Spanish, French, and more.

Run large language models on local and distributed networks, supporting collaborative inference and fine-tuning.

Open-source AI portrait restoration tool based on GFPGAN, supporting online restoration of old and blurry face photos with simple operation.

A high-performance LoRA fine-tuned model inference service supporting on-demand loading of thousands of adapters for efficient production use.

RouteLLM is an open-source framework for serving and evaluating LLM routers, aiming to significantly reduce cost while maintaining quality via routing strategies.

An open-source document parsing toolchain converting PDF, images, and Office documents into structured JSON/Markdown/CSV.

RedPajama Data is an open-source corpus preparation tool and dataset for training large scale language models.

A bidirectional pipeline parallelism algorithm designed to improve training efficiency via computation-communication overlap, suitable for large-scale training like DeepSeek V3/R1.

Expert Parallelism Load Balancer dynamically allocating load in expert parallel training to improve cluster utilization and performance.

Pipet is a developer-oriented command-line web scraping and extraction tool supporting HTML, JSON, and Playwright query modes.

Locally running Tele-Prompter combining voice input and semantic retrieval to assist with impromptu speaking and speech practice.

LongBench is a bilingual multi-task benchmark for long-context understanding and reasoning, covering long doc, multi-doc, and repo understanding to evaluate LLM long-context capabilities.

Free diffusion models course by Hugging Face, containing theory and practical notebooks covering training, fine-tuning, and conditional generation.

gpt-engineer is an open-source code-generation and experimentation platform that provides a local CLI and toolchain to generate and refine projects from natural language prompts.

An open-source Gradio-based web interface for Stable Diffusion providing extensive generation, editing, and extension features suitable for local or server deployment.

AI Engineering by Chip Huyen is a comprehensive book focused on practical methods for applying foundation models in real-world scenarios.

Open-source text-to-image generation model that creates high-quality images from text descriptions using diffusion model technology.

A macOS GUI app for running Stable Diffusion locally, offering one-click installers and offline generation capabilities.

Curated collection of AI agent resources including frameworks, tools, tutorials, and best practices for building intelligent agents.

GPT-NeoX is EleutherAI's toolkit for training large-scale open-source language models.

Explore 'Made with ML', a platform for mastering production-grade machine learning, offering project-driven learning and best practices for developers.

An open-source platform for training, serving, and evaluating conversational large language models, offering distributed serving, a Web GUI, and OpenAI-compatible APIs.

Ludwig is a low-code, declarative deep learning framework that enables building, training, and deploying models via YAML configuration for multi-modal and distributed workflows.

Chatbot UI is an open-source chat interface and hosted example that supports multiple backends and third-party model integrations, enabling rapid construction of customizable conversational apps.

BitNet (bitnet.cpp) is Microsoft's open-source inference framework for 1-bit LLMs, offering optimized CPU/GPU kernels for efficient low-bit inference.

A comprehensive Large Language Model course covering LLM fundamentals, scientist, and engineer tracks, providing a complete learning path.

Zero to Mastery Learn PyTorch for Deep Learning course, the second-best place to learn PyTorch online. Provides rich learning resources including online books and YouTube videos.

An open-source text-to-speech system focused on high-quality multi-voice synthesis and realistic prosody.

An open-source project for running large language models locally on desktops and laptops, providing desktop clients, a Python SDK, and multiple inference backends.

Swarm is an experimental, educational library from OpenAI for lightweight multi-agent orchestration and execution.

Assemble, configure and deploy autonomous AI agents in the browser with a full frontend/backend stack and quick CLI setup.

Bolt.new is a browser-based, AI-powered full-stack development environment that lets you create, run, edit, and deploy web apps using natural language prompts.

Suno's open-source generative text-to-audio model capable of producing realistic speech, music, and sound effects.

NebulaGraph is an open-source distributed graph database that provides horizontal scalability, low-latency graph queries and graph analytics, with support for vector retrieval for AI scenarios.

An experimental self-building autonomous agent framework focused on function management and rapid prototyping for research.

A production-ready, privacy-first framework for private document search and RAG APIs that can run offline or on private cloud.

One API is an open-source LLM API management and routing system that unifies multiple model providers and offers distribution, quota, and access management features.

A scalable, low-latency service for embedding images and text, suitable for neural search and multimodal applications.

Jina Serve is a cloud-native framework for building and deploying multimodal AI services, supporting gRPC/HTTP/WebSocket, dynamic batching, elastic scaling and multiple deployment modes.

No description available

MLOps Basics is an open-source, week-by-week tutorial repository covering model training, monitoring, packaging, and deployment practices.

A command-line tool that aggregates multi-source content into a single LLM-friendly input.

A curated collection of 300+ real-world ML system design case studies covering production practices across industries.

A practical cheat sheet for mastering Anthropic Claude Code, from basic commands to advanced workflows.

A curated collection of AI learning books and resources covering machine learning, deep learning, NLP, and related fields.

An Azure Functions reference app that demonstrates MCP tools and durable multi-agent orchestration.

A graph-centric Retrieval-Augmented Generation workflow for deep searching over graph-structured data.

An adaptive multi-agent framework that plans, executes and validates complex data-science tasks.

A practical guide and workflow for integrating Claude Code, Codex and similar tools into reliable, iterative development.

cmux is an open source Claude Code manager that can run multiple code agent CLIs in parallel and provide isolated development workspaces.

An open source data set and tool chain for generating, publishing and loading CellARC (based on cellular automaton tasks), supporting Hugging Face for fast download and simulation visualization.

An open source drone control framework that combines computer vision and motion imagery EEG classification, supporting gestures, head movements, and imaginary motion to control simulated drones.

Ultra-fast semantic filter based on embedding similarity for sifting the most relevant subset from thousands of tools in MCP servers.

A self-contained, privacy-centric RAG (Retrieval Augmentation Generation) system where all data remains local.

An open source operation and resource management suite that helps quickly build hibernating agent sandboxes on Kubernetes and reduce GPU costs.

LangCrew is a high-order multi-agent development framework based on LangGraph, providing enterprise-level features and full-stack UI.

An open source solution for GPU virtualization and pooling designed to increase cluster utilization and optimize inference workloads.

Jta is an AI-driven translation tool for i18n JSON files, with self-checking and incremental translation capabilities.

A high-performance weight synchronization framework for reinforcement learning training-inference, supporting second-level parameter updates on the inference side.

The open source platform based on RAGFlow extends document processing, hybrid retrieval and evaluation feedback capabilities.

RapidFire AI is a large language model toolchain for rapid experimentation and fine-tuning, supporting model customization, experiment recording, and repeatable deployment.

An open source tool chain for video editing automation, combining Claude Code, WhisperX and Ruby libraries to generate importable FCPXML/xmeml first-cut results.

An open source image generation project that maps numerical style codes into consistent and reproducible image styles.

The interactive command line agent (REPL) based on LangChain and LangGraph is used to quickly build and run multi-tool and multi-model agent interaction environments.

Capture, synchronize, and inject conversational reasoning for engineering teams so agents instantly gain team knowledge in new sessions.

An open source tool for writing and running agents using Markdown, suitable for automation, CI/CD and scheduled tasks.

A lightweight context management tool for code agents that treats context as code for reuse and debugging locally and in the pipeline.

Centralize and transform engineering specifications (playbooks) into context, guardianship and governance that can be used by agents.

An open source AI middleware for the enterprise for building production-ready agents, workflow and retrieval enhancement applications.

Sora MCP is an MCP server used to access the Sora video generation API, simplifying the deployment and invocation of video generation services in multiple environments.

A multi-agent and context engineering toolkit under the command line, supporting local and cloud models.

Kthena is a Kubernetes-native LLM inference platform for deploying and managing large models in production environments.

Kubernetes-native scale-to-zero solution supports zero traffic loss, no code modifications, and direct integration with Kubernetes resources.

A benchmark suite for evaluating the reasoning and decision-making capabilities of agentic large language models and visual language models on game tasks.

An open source project that provides standardized environment infrastructure for agentic AI development.

A cost-optimized model cascading framework for intelligently routing requests among multiple models to balance cost and quality.

A pluggable memory layer that provides long-lasting, intelligent memory and retrieval capabilities for large language models.

An open-source testing platform and SDK for LLM and agent applications, used to automatically generate test scenarios and evaluate model outputs.

mini-kode is an AI coding agent command line tool for learning and teaching that helps practice programming tasks through natural language and prompts.

llmaz is an advanced LLM inference platform for Kubernetes provided by InftyAI. It is designed to simplify model deployment and distribution, and supports multiple backends, heterogeneous clusters and elastic scaling.

LLaMA Box is an inference service based on llama.cpp, providing an OpenAI-compatible API and supporting multi-model, multi-device and image generation capabilities.

An open set of specifications and practices for enterprise-level deployment that encodes engineering and operational models for large-scale agent applications.

A local-first terminal operator that lets AI perform tasks securely on the terminal with controlled approvals and audits.

A Rust-based terminal semantic coding agent that integrates multi-vendor large language models and semantic coding tools.

A Web UI and operation and maintenance tool chain that provides visual management, container orchestration and performance measurement for vLLM services.

An open source runtime that helps teams deploy agents on their own infrastructure and retain full control.

A cloud runtime for agents that provides millisecond startup, Firecracker-based isolated micro-VMs, and multi-language SDKs.

An open source and free alternative to ChatGPT Atlas that provides in-browser visual automation and tool routing capabilities.

Volcano SDK is a TypeScript-first, multi-vendor AI Agent SDK that supports chained workflows, MCP tool integration, and observability.

FleetCode is a lightweight desktop terminal application for running multiple CLI coding agents in parallel in an isolated git worktree.

VisioFirm is a research-oriented cross-modal project focusing on visual understanding and the application of large models to visual tasks.

dInfer is an efficient inference framework for diffusion language models (dLLM), focusing on decoding algorithms and KV-cache management to improve inference throughput and quality.

An intelligent memory system for AI applications that combines vector, full-text and graph retrieval and supports multi-agent scenarios.

A large speech model framework with a unified continuous speech segmenter as the core and supporting speech understanding, generation and free-form editing.

A transparent, minimalist and easy-to-modify agent framework, with a core implementation of about 300 lines of code, suitable for learning, customization and rapid prototyping iteration.

AngelSlim is a model compression toolbox launched by Tencent, dedicated to providing easy-to-use and comprehensive model compression solutions.

Large model loading and inference engine for fast loading and hot-swapping, significantly shortening model cold start time, suitable for edge and local deployment.

VibeVoiceFusion is a full-stack multi-speaker speech synthesis web application based on VibeVoice, with voice cloning and VRAM optimization capabilities.

Director provides MCP-based Playbook management and operation capabilities, allowing agents to be declaratively configured and reproduced locally or in containers.

The lightweight rich text and Markdown hybrid editing framework released by Ant Group facilitates streaming document editing and rendering in web scenarios.

Easy-Graph provides convenient graph data structure and analysis tools, suitable for graph learning and knowledge graph construction.

MCP Gateway is Microsoft's open source MCP server reverse proxy and management layer that supports scalable, session-aware routing and lifecycle management in the Kubernetes environment.

Quivr: An open source tool for building RAG/knowledge bases that helps developers integrate document and vector retrieval into interactive assistants.

A collaborative system and toolset that connects large models with community models to automate complex AI tasks.

An "AI sub-agent system template library" that can be directly implemented, organized by functional departments, covering complete software workflows such as design, engineering, products, marketing, operations, testing, and project management.

A fast and expressive Python template engine suitable for generating HTML and arbitrary text.

A low-latency speech transcription Python library for real-time applications with voice activity detection, wake word triggering, and on-the-fly transcription capabilities.

An ecological project for managing and running open source GPT-style conversation models, providing deployment and sample tools.

Self-organizing file manager, based on the Llama3 model for file content summarization and automatic renaming, supporting batch processing and daemon mode.

Meta Lingua: A lightweight and research-oriented LLM training and inference library that facilitates rapid experimentation with new structures and metrics.

Pezzo is an open source, cloud-native LLMOps platform for prompt management, observability, caching and collaborative delivery to reduce cost and latency.

A low-latency, high-throughput inference framework developed by Microsoft, suitable for large-scale model inference scenarios.

An open source NotebookLM alternative, combined with LlamaCloud to provide document understanding and interaction capabilities.

A collection of open source agents for brainstorming scenarios, providing multiple perspectives and strategies to generate creative solutions.

A centralized registration and governance platform for publishing, discovering and deploying MCP servers, agents and skills.

A desktop application that quickly builds, selects, and exports project code context locally for pasting into chat models or agent conversations.

A tool to automate cue engineering experiments by generating, testing, and ranking candidate cues.

A terminal AI development agent for large-scale projects that supports large context, automated execution and step-by-step review to facilitate the completion of complex tasks in real projects.

An end-to-end autonomous research workflow for researchers to organize experiments, automate steps, and reproduce results.

Python library that wraps LandingAI's Agentic Document Extraction API to extract structured data from visually complex documents.

AI-Researcher is an open-source system for automating scientific research and paper generation, providing end-to-end support from literature review to manuscript drafting.

AIDE ML is an agentic tree-search driven tool for machine-learning engineering that drafts, evaluates, and improves ML code automatically.

Ailice is a fully autonomous, general-purpose agent framework based on an IACT architecture for composing modular agents and skills.

A zero-code framework for creating LLM-powered intelligent agents through natural language descriptions, with self-managed vector stores and runtime traces.

AWS Agent Squad (formerly Multi-Agent Orchestrator) — a multi-agent orchestration framework by AWS designed for enterprise-grade AI applications with local and cloud deployment options.

Open source and self-hostable browser automation library for AI agents

Self-hosted AI desktop agent for automating computer tasks via natural language, supporting multiple models and full desktop control.

An open-source WebSocket server that wraps the Claude Agent SDK to enable real-time, bidirectional communication with Claude in sandboxes or local environments.

A collection of production-tested Claude Code infrastructure patterns including auto-activating skills, modular skills, hooks, and example agents.

MCP plugin for semantic code search. Integrates with Claude Code, Gemini CLI, Cursor, or any AI coding agents.

A browser-automation example agent that demonstrates using the Gemini Developer API or Vertex AI to perform multi-step tasks in a real browser for research and prototyping.

A hierarchical multi-agent framework for deep research and general-purpose task solving, supporting browser automation, MCP tool integration, and multi-model inference.

A git-based differential memory backend that keeps current-state files and preserves evolution in git history for temporal, auditable agent memory.

GitMCP is a free, open-source, remote MCP server that transforms any GitHub project into a documentation hub, putting an end to code hallucinations.

Discover Glass: the lightweight, privacy-first digital assistant that enhances your desktop experience with real-time context understanding and smart meeting support.

A semi-structured agentic framework where agents discover tasks at runtime and dynamically build workflows.

Discover Jaaz, the first open-source multimodal creative assistant for secure, collaborative visual content creation and editing.

Llumnix is a cross-instance scheduling layer for LLM inference that reduces latency and improves throughput for multi-instance serving deployments.

Hundreds of containerized MCP (Model Context Protocol) servers, simplifying MCP server deployment and usage.

An agent that integrates multimodal models and specialty imaging tools for chest X-ray reasoning.

OmniParser is an open-source pure vision GUI Agent screen parsing tool by Microsoft, converting interface screenshots into structured elements to empower LLMs and agents for precise desktop actions.

An open-source desktop GUI and toolkit for managing Claude Code sessions and agents.

Discover Open Agent Builder, an open-source platform for AI workflows with drag-and-drop design, multi-tool integration, and powerful automation capabilities.

An open-source toolkit for MCP development and debugging that bundles a VS Code plugin, debugging panels, and an SDK to streamline the path from testing to deployment.

Plexe is an open-source platform that builds and deploys machine learning models from natural language prompts.

SuperDesign is the first open-source design agent for IDEs, enabling UI mockup, component, and wireframe generation from natural language prompts.

An open-source agent template for investment research that integrates Financial Datasets' real-time and historical market data.

AISheets is an open-source no-code AI dataset tool by Hugging Face, enabling dataset building, enrichment, and transformation with integrated models.

AutoRAG provides automated RAG pipeline construction tools, simplifying document retrieval and Retrieval-Augmented Generation (RAG) integration workflows.

An open-source unified multimodal foundation model and toolbox for both understanding and generation tasks.

A LangChain-based RAG repository providing end-to-end notebooks and examples for data loading, embedding, retrieval, and RAG pipelines.

Microsoft's open-source Call Center AI demonstrates integrating Azure Communication, Cognitive Services and OpenAI for phone-based customer support.

An AI programming assistant that runs completely on your machine, providing code completion, refactoring suggestions, and debugging help without sending your code to external servers.

An open-source RAG framework for building modular, production-ready retrieval-augmented generation applications.

A reverse-engineered proxy that converts GitHub Copilot into OpenAI/Anthropic API compatible server, with Claude Code integration.

An open-source OCR model and toolkit based on Contexts Optical Compression for LLM-centric multimodal inference scenarios.

Discover dots.ocr, a powerful multilingual document parser that excels in layout detection and content recognition for enhanced document processing.

A TypeScript-first framework for evaluating LLM-powered applications, enabling repeatable tests and benchmarks for model behaviour and application quality.

A practical guide from Hugging Face summarizing best practices and theory for evaluating LLMs.

A long-form streaming TTS system for multi-speaker dialogue generation, providing stable, natural speech with reliable speaker switching, multilingual support, and low-latency streaming output.

An AI-powered video subtitle editing component that integrates high-accuracy speech recognition with visual subtitle editing features.

Fogsight is an LLM-driven animation engine and interactive creation platform that converts text or concepts into high-quality animated sequences.

A Python FastAPI-based Google Gemini API proxy and load balancer service with multi-key polling and dual protocol compatibility.

Google AI Mode Scraper (by Oxylabs) extracts and structures Google AI Mode responses for large-scale use cases.

Command-line AI tool providing intelligent assistant and automation features for developers in the terminal.

HYPIR leverages diffusion‑yielded score priors for image restoration and super-resolution, with released code and pretrained weights.

iFlow CLI is a terminal-based agent-like CLI that analyzes repositories and automates tasks from code generation to workflow automation.

A controllable and fast text-to-speech (TTS) toolkit supporting over 7000 languages.

An open-source framework for synchronized video dubbing, enhancing long-form content with natural lip motion and expressions. Explore its innovative features today!

An open-source multimodal vision-language toolbox and baseline for image and video understanding and generation tasks.

A privacy-first, on-premise-ready speech transcription and collaboration platform with real-time ASR, speaker diarization, batch transcription and AI summarization.

A video translation and dubbing tool combining LLMs, ASR and TTS, supporting 100 languages and one-click end-to-end deployment for multi-platform output.

An open-source large-scale multimodal model and toolkit that unifies training and inference for images, multi-image, video and 3D tasks.

Maestro is a toolkit from Roboflow for fine-tuning multimodal models, encapsulating configuration, data loading, and reproducible training pipelines.

A high-performance, extensible Python Markdown parser that follows CommonMark and supports plugin extensions.

MiMo-Audio: audio language models demonstrating few-shot learning and strong generalization across audio understanding and generation tasks.

MInference is a framework for long-context LLM inference that accelerates pre-filling and large-context processing using dynamic sparse attention and optimized kernels.

A lightweight vLLM implementation built from scratch, offering high-performance offline inference comparable to vLLM.

A library for pretraining transformer models that simplifies scalable, high-performance training from single-node to multi-node setups.

DeepResearch combines retrieval, webpage reading and multi-step reasoning to find accurate answers, available as a hosted service or self-hosted deployment.

Lightweight offline LLM inference library optimized for very long context inference on low-memory GPUs.

An open-source stack for diffusion-based large language models that covers pretraining, evaluation, inference and checkpoints.

OpenLLM (by BentoML) simplifies self-hosting LLMs by providing CLI tools, an OpenAI-compatible server, built-in chat UI, and integrations with various inference backends.

A framework that enables parallel training of large pretrained models via chunk-based dynamic memory management.

An open-source, stateless command-line tool providing quick question answering and a single-step agent mode.

A lightweight self-hosted EPUB reader designed to read books chapter-by-chapter alongside large language models (LLM).

RustGPT is an educational Rust implementation of a transformer-based LLM that demonstrates core model components and training pipelines.

An AI assistant plugin for Neovim that integrates Copilot LSP's Next Edit Suggestions with multiple AI CLIs to streamline in-editor code review and interactions.

An open-source model project that improves small-model reasoning via multi-stage distillation and optimization.

A family of optimized vision-language models that balance efficiency and accuracy for image, multi-image and video understanding tasks.

A high-performance inference and compilation stack designed for production deployments across diverse hardware and platforms.