Large Language Models
LLM related technologies, tools and applications
A selected Chinese chat generation template project provides various interesting templates for people using ChatGPT, making the chat more interesting! This project provides a variety of templates that allow ChatGPT to play different roles, such as Linux terminal, text translation and correction, technical interviewers, storytellers, talk show performers, writing guidance consultants, psychological counselors, etc. Through these templates, users can quickly get started with ChatGPT, expand its application scenarios, and provide richer chat experiences.
A development tool for creating powerful AI applications, it provides APIs for plugins and datasets, as well as an interface for quick engineering and visualization operations. For developers and researchers who want to develop applications, Dify provides convenient tools and interfaces to help them build feature-rich AI applications.
Generative AI for Beginners is a generative artificial intelligence course provided by Microsoft. The course aims to help beginners understand and master the basic concepts and technologies of generative artificial intelligence. Through practical tutorials and examples, learners can gradually explore the world of generative artificial intelligence and lay a solid foundation for in-depth learning in the future.
A reverse engineering project aimed at studying and exploring the working principles of the GPT-4 and GPT-3.5 models. It provides users with an opportunity to understand the internal mechanisms of GPT models and promotes research and improvement of generative pre-trained models.
🤖 An open-source, high-performance chatbot framework that supports voice synthesis, multimodal and scalable function call plugin systems. Users can deploy private ChatGPT/LLM web applications with one click for free, providing powerful chatbot features for projects.
MarkItDown is a lightweight Python utility designed to convert various file formats into Markdown, optimized for use with Large Language Models (LLMs) and text analysis pipelines. It supports a wide range of formats, including PDF, PowerPoint, Word, Excel, images, audio, HTML, and more, preserving document structure like headings, lists, and tables. The tool is ideal for machine consumption rather than high-fidelity human-readable output. It offers a command-line interface, Python API, and Docker support, with optional dependencies for specific file types. MarkItDown also integrates with Azure Document Intelligence and supports third-party plugins for extended functionality. Installation is straightforward via pip, and contributions are encouraged through issues, PRs, and plugin development.
A multi-agent meta-programming framework that, given a line of requirements, can return product documentation, architectural design, task lists, and code. This project provides an innovative way to manage and execute projects, converting requirements into specific documents and task lists, making project management efficient and intelligent. For projects that require planning and coordination, this framework provides strong support.
ChatGPT Chinese training guide. Various scene usage guide. Learn how to make it listen to you.
GPT Engineer is highly adaptable, scalable, and enables your agent to understand the desired appearance of your code. Simply specify what you want it to build, and the AI will prompt you for clarifications before constructing it. It generates the entire codebase based on the provided instructions.
A large language model trained by OpenAI, which can be used for chatbots, text generation and question answering systems, etc. Whether you are a beginner or a professional, you can find the corresponding solutions in this project.
A collection of articles covering fields such as graph neural networks, large language models (LLMs), and convex optimization. The project provides in-depth technical discussions, covering the principles of graph neural networks, applications of large language models, and the importance of convex optimization in relevant areas.
vLLM is a high-performance, open-source library designed for efficient and scalable large language model (LLM) inference and serving. It features state-of-the-art serving throughput, optimized memory management with **PagedAttention**, and supports advanced techniques like continuous batching, CUDA/HIP graph execution, and various quantization methods (e.g., GPTQ, AWQ, INT4, INT8, FP8). vLLM integrates seamlessly with popular Hugging Face models, offers OpenAI-compatible API servers, and supports distributed inference with tensor and pipeline parallelism. It is highly flexible, supporting a wide range of hardware (NVIDIA, AMD, Intel, TPU, AWS Neuron) and models, including Transformer-based LLMs, Mixture-of-Experts, and multi-modal models. vLLM is community-driven, with contributions from academia and industry, and is backed by sponsors like a16z, Google Cloud, and NVIDIA.
An innovative framework that allows developers to build and develop LLM (large language model) applications using multiple agents. These agents can talk to each other and work together to solve tasks, making the application more intelligent. AutoGen agents are customizable, conversational, and seamlessly integrate artificial intelligence and human involvement to provide broader functionality.
A full-stack application that turns any documents into an intelligent chatbot with a sleek UI and an easier way to manage your workspaces.
A data framework for LLM (large language model) applications. It provides a solution for data storage and management for LLM applications, helping users build and manage LLM applications more efficiently.
WeChatMsg is a project designed to empower users to take control of their WeChat data, emphasizing the preservation of personal memories and interactions. It supports WeChat 4.0, offering features like local database access, chat interface restoration, and comprehensive data export options (SQLite, HTML, CSV, TXT, Word). The project also includes chat data analysis and visualization, enabling users to generate personalized annual reports. With a focus on user-friendly design, it features a streamlined UI, lower memory usage, and faster export speeds. WeChatMsg is built on the principle of "My Data, My Control," ensuring users can retain and utilize their digital footprints responsibly. It is open-source, encouraging community contributions and fostering innovation in personal AI development.
A tool for building customized low-code machine learning (LLM) workflows using a drag-and-drop UI with LangchainJS. It simplifies the development and deployment of machine learning processes, enabling users to design their own machine learning workflows through drag-and-drop operations, thereby enhancing development efficiency.
Chat-based large language models can interact with third-party systems and dynamically retrieve information.
A tool that can quickly package various web pages into MacApp. Compared with Electron, Pake is more compact and easy to use, faster, and has achieved universal shortcut transmission, immersive window, drag-and-drop, and packaging style compatibility features. For developers who need to package web applications into desktop applications, Pake provides a convenient and efficient open-source tool.
Use large models to build WeChat chatbots, based on GPT3.5/GPT4.0/Claude/ERNIE 1.0/Xunfei Xinghuo/LinkAI, support deployment of personal WeChat, official accounts and enterprise WeChat, can process text, voice and pictures, access operating systems and the Internet, support customized exclusive robots based on knowledge bases.
A voice generation model designed specifically for dialogue scenarios, mainly used for LLM assistant dialogue tasks, dialogue voice and video introductions, etc., supports the synthesis of speech from mixed Chinese and English text, has strong timbre performance, and can reach a level where it is difficult to distinguish between real and fake.
A low-code development framework that is open source on GitHub. It provides a rich visual toolbox, and with just a few steps, you can quickly build enterprise-level deployment tools. The framework provides APIs to connect with various external data sources such as PostgreSQL, MongoDB, Elasticsearch, and common third-party services like Stripe, Slack, Airtable.
A free and open-source similarity search engine for massive feature vectors. Compared to operator libraries such as Faiss and SPTAG, Milvus provides a complete framework for vector data updates, indexing, and queries. Milvus utilizes GPU for index acceleration and query acceleration, significantly improving single-machine performance. Its deployment and usage are simple, reducing the difficulty of AI application implementation.
A personal knowledge management system that prioritizes privacy, supports fine-grained block-level referencing and Markdown WYSIWYG editing, has real-time rendering, mathematical formulas, charts, HTML export, Markdown files, AI writing, etc., and supports cross-platform use.
🤖 Assemble, configure and deploy autonomous AI agents in the browser. Break through the existing GPT's limitation of only doing text-related tasks. Based on your set goals, use various tools to automatically complete all tasks with GPT-4 to achieve your goals.
An AI coding assistant designed to help developers write code more efficiently. It can provide code suggestions, auto-completion, and error checking, thereby accelerating the coding process. Tabby is an experimental project aimed at exploring how to integrate artificial intelligence into software development workflows to enhance development efficiency.
A collection of carefully organized large language models in practical applications in different fields, covering scenarios such as customer service, law, health, tourism, data analysis, etc., and combined with RGA and memory functions. It also collects tutorials on LLM fine-tuning, advanced tools, and framework usage, providing detailed information.
Your AI second brain. A copilot to get answers to your questions, whether they be from your own notes or from the internet. Use powerful, online (e.g gpt4) or private, local (e.g mistral) LLMs. Self-host locally or use our web app. Access from Obsidian, Emacs, Desktop app, Web or Whatsapp.
Free ChatGPT API Key, free ChatGPT API, supports GPT4 API (free), domestic available free forwarding API for ChatGPT, direct connection without proxy. Can be used with software/plugins such as ChatBox, greatly reducing the cost of interface usage. Unlimited and smooth chatting is possible domestically.
An innovative platform that integrates machine learning into databases through SQL. It treats models as virtual tables (AI-tables), allowing users to directly use SQL queries for time series, regression, and classification predictions without the need for complex data preparation and preprocessing steps. This greatly simplifies the machine learning development process. MindsDB provides developers with a simple and efficient way to accomplish machine learning tasks.
An AI tool that generates short videos with one click. This project is an AI video generation tool based on large model services. You only need to provide a theme or keyword, and it can automatically generate high-definition short videos. It has a simple and easy-to-use web interface, supports batch generation, setting video duration, and horizontal/vertical screen size functions. Shared by @jolahua
An OpenAI interface management and distribution system that supports multiple channels, including Azure. It is used for secondary distribution management keys, requiring only a single executable file, packaged into Docker images, and can be deployed with one click, convenient and practical. Provides an English user interface.
An open-source SDK designed to enable developers to easily integrate AI services such as OpenAI, Azure OpenAI, and hugging Face with traditional programming languages like C# and Python. Through this project, developers can create applications that combine traditional programming and artificial intelligence, expanding the scope of applications and possibilities for innovation.
An assistant built for multi-modal GPT-4 level capabilities. It combines natural language processing and computer vision to provide users with powerful multi-modal interaction and understanding. LLaVA aims to better understand and process language and visual information, thus enabling more complex tasks and conversations. This project represents the direction of development for next-generation intelligent assistants, which can better understand and meet user needs.
PDFMathTranslate is a versatile tool designed for translating scientific PDF documents while preserving their original structure, including formulas, charts, tables of contents, and annotations. It supports multiple languages and integrates various translation services. The project offers multiple usage options, including a command-line tool, an interactive GUI, and Docker deployment, making it accessible for diverse user needs. It also provides online demos for quick testing and supports advanced features like partial document translation, multi-threading, and custom prompts. PDFMathTranslate is continuously updated with experimental backends and improved functionalities, ensuring flexibility and efficiency for academic and professional use.
A tool to simplify the API calls of large models. This project can unify the interfaces of various AI large models and services into the format of OpenAI, simplifying the work of switching and managing different AI services/large models. In addition, it also supports setting budgets, limiting request frequencies, managing API keys, and configuring OpenAI proxy servers, among other functions.
A convenient tool that helps users effortlessly locate personal or work files by asking natural language questions. It combines ChatGPT's powerful conversational generation capabilities with document retrieval functions, providing users with an efficient information query and search tool.
The GPT Researcher is an autonomous agent designed to conduct comprehensive online research on a variety of tasks. The agent can generate detailed, factual, and unbiased research reports, offering customized options that focus on relevant resources, outlines, and curricula. Inspired by AutoGPT and recent Plan-and-Solve papers, the GPT Researcher addresses issues of speed and determinacy by employing parallel agent operations (as opposed to synchronous operations), thereby delivering more stable performance and increased speed.
Unilm is a large-scale self-supervised pre-training model across tasks, languages and modalities. It is pre-trained by self-supervised learning, which enables the model to be transferred to different tasks and languages, with wide application value. The design goal of Unilm is to provide a unified pre-training model that can handle various natural language processing tasks, such as machine translation, text summarization, question answering, etc.
MLC LLM is a universal solution that enables the local deployment of any language model across various hardware backends and native applications, allowing everyone to develop, optimize, and deploy AI models locally on their own devices.
architecture.of.internet-product - The technical architecture of internet companies, the architecture of WeChat/Taobao/Weibo/Tencent/Alibaba/Maifood/Baidu/Google/Facebook/Amazon/eBay
It includes some practical machine learning and Python open source projects and tools. There are more than 900 projects in total, including data visualization, natural language processing, text and image data, web crawling, etc.
