Project directory

Projects

Open-source AI projects tracked by the leaderboard, ordered by GitHub stars.

Go 242 stars

coder/hnsw

In-memory vector index for Go

Overview: hnsw is a Go package that implements Hierarchical Navigable Small World (HNSW) graphs, providing efficient approximate nearest neighbor searches for high-dimensional vector data. It offers fast approximate nearest neighbor searches with high-dimensional vector data and is an alternative to vector databases like Pinecone or Weaviate. Key features: - Hierarchical Navigable Small World graphs for efficient approximate nearest neighbor searches. - Supports operations like insertion, deletion, search, and lookup. - Implements just the essential operations for in-memory vector indexing. - Provides facilities for persistence, allowing graphs to be loaded and saved from persistent storage. Tech stack: Go Why it matters: hnsw is useful for applications requiring fast approximate nearest neighbor searches in high-dimensional vector data, such as recommendation systems, image search, and other machine learning tasks where efficient indexing is crucial. Repository: coder/hnsw GitHub stars: 241 GitHub description: In-memory vector index for Go

Python 242 stars

google-marketing-solutions/vigenair

Recrafting Video Ads with Generative AI

### Overview: ViGenAiR is a cutting-edge platform that leverages Generative AI to enhance the quality and effectiveness of video advertisements. ### Key Features: - **AI-Driven Video Editing**: ViGenAiR automatically recrafts video ads using advanced AI models, ensuring a more engaging and impactful presentation. - **Aspect Ratio Flexibility**: Supports both 16:9 and 4:3 aspect ratios for seamless integration into various ad formats. - **Prompt Library**: Offers pre-defined prompts for generating video variants, making the process more efficient and customizable. - **Blanking Fill Effect**: Automatically applies a blanking fill effect when cropping is not feasible, ensuring a smooth transition. - **Automatic Video Conversion**: Converts .mov files to .mp4 format for better compatibility and performance across different browsers. - **Variant Generation**: Generates multiple video variants based on user-defined criteria, including different objectives and segment cuts. ### Tech Stack: Python ### Why it Matters: ViGenAiR is essential for marketers looking to improve the visibility and engagement of their video ads. By leveraging AI, it streamlines the ad creation process, reduces manual effort, and ensures higher-quality output, ultimately leading to better ad performance and ROI.

Unknown 242 stars

krystalan/DRT

Deep Reasoning Translation (DRT) Project

### Overview: This repository contains the resources for our work on Deep Reasoning Translation (DRT), focusing on the development and exploration of deep learning models for translation tasks. The repository includes papers such as "Deep Reasoning Translation via Reinforcement Learning" and "Deep Reasoning Translation via Long Chain-of-Thought," as well as the full data for the DRT work, including synthesized thought contents and translations. ### Key features: - **DeepTrans**: Deep Reasoning Translation via Reinforcement Learning - **DRT**: Deep Reasoning Translation via Long Chain-of-Thought - **ExTrans**: Multilingual Deep Reasoning Translation via Exemplar-Enhanced Reinforcement Learning - **DeepTrans-7B**: A model with 7 billion parameters - **DRT-7B**: A model with 7 billion parameters - **DRT-8B**: A model with 8 billion parameters - **DRT-14B**: A model with 14 billion parameters ### Tech stack: The repository uses Hugging Face Transformers for model training and inference. ### Why it matters: Our work focuses on developing deep learning models that can translate text while preserving the reasoning and context of the original text, making it particularly useful for applications where the meaning and context of the source text are crucial, such as in legal, medical, and technical translation. ### GitHub stars: 242 ### Updates: - *2025.08.21*: Our paper "DeepTrans: Deep Reasoning Translation via Reinforcement Learning" is accepted to

Python 242 stars

spcl/MRAG

Official Implementation of "Multi-Head RAG: Solving Multi-Aspect Problems with LLMs"

Overview: This project is an official implementation of Multi-Head RAG (MRAG), a novel scheme designed to address complex queries that require fetching multiple documents with different contents. Key features: - Multi-Head RAG (MRAG) framework - Synthetic datasets and queries generation - Vector database integration - Evaluation strategies for query accuracy - Documentation for understanding the framework Tech stack: Python Why it matters: MRAG leverages the activations of Transformer's multi-head attention layer to fetch multi-aspect documents, improving retrieval accuracy for complex queries.

Python 241 stars

Laurent00TT/PharosRAG

Pharos — local-first agentic RAG for your team's document library: multi-format ingest, hybrid retrieval, enterprise ACL, dual HTTP + MCP exits.

### Overview: Pharos is a local-first agentic RAG (Retrieval-Augmented Generation) system designed for your team's document library. It seamlessly converts various formats like PDFs, scans, docx, pptx, and xlsx into a local knowledge base, allowing you to ask questions and trace citations. Pharos supports enterprise access control and dual HTTP + MCP exits, making it a powerful tool for knowledge management. ### Key Features: - Multi-format ingest: Converts documents into a local knowledge base. - Hybrid retrieval: Retrieves information from multiple sources. - Enterprise ACL: Provides enterprise-level access control. - Dual HTTP + MCP exits: Supports both HTTP and MCP exits for better performance. ### Tech Stack: - Python: The primary language for development and deployment. - CI/CD: Continuous Integration and Continuous Deployment workflow. - Exits: Supports HTTP API and MCP for different consumer needs. ### Why It Matters: Pharos is a robust and scalable solution for knowledge management, providing enterprise-level access control and hybrid retrieval capabilities. Its modular design allows for easy integration and deployment, making it a valuable tool for teams looking to enhance their document management systems. ### Repository: Repository: Laurent00TT/PharosRAG GitHub Stars: 241 GitHub Description: Pharos — local-first agentic RAG for your team's document library: multi-format ingest, hybrid retrieval, enterprise ACL, dual HTTP + MCP exits. ### Installation: ```bash pip install -e '.[dev]' ``` ### Architecture: Pharos is designed to

Python 241 stars

chensyCN/LogicRAG

Source code of LogicRAG at AAAI'26.

### Overview: LogicRAG is a structured retrieval system that leverages query logic dependency graphs to adaptively guide structured retrieval, enabling efficient and scalable reasoning on large knowledge bases. ### Key Features: - **Logic Dependency Analysis**: Converts complex questions into logical dependency graphs for multi-step retrieval. - **Graph Reasoning Linearization**: Linearizes complex graph reasoning into sequential subproblems while maintaining logic coherence. - **Efficiency**: Schedules reasoning via graph pruning and optimizes context-length via rolling memory. - **Interpretable Results**: Provides clear reasoning paths and dependency analysis for better explainability. ### Tech Stack: - Python ### Why It Matters: This work has been accepted to AAAI'26 and is available on Arxiv. It addresses the need for structured retrieval without the need for pre-built knowledge graphs, making it particularly useful for large and dynamic knowledge bases.

Jupyter Notebook 241 stars

dmatrix/genai-cookbook

A mixture of Gen AI cookbook recipes for Gen AI applications.

### Overview: The GenAI Cookbook is a comprehensive guide designed for developers to explore and utilize Generative Artificial Intelligence (GenAI) applications effectively. This book covers a wide range of natural language processing (NLP) applications, providing practical examples and how-to guides on using language models. ### Key Features: - **LLM Prompting Strategies**: Get-started examples to advanced prompting techniques. - **DSPy Framework**: Declarative and programmatic approaches to programming LLMs. - **Simple Chatbots**: Basics of building interactive AI. - **Retrieval Generation Augmentation (RAGs)**: Incorporating personal or organizational data for more accurate responses. - **Fine-Tuning LLMs**: Customizing models for domain-specific tasks. - **Extending LLM Functionality**: Using agents and functions to enhance capabilities. - **Employing Leading Language Models**: Practical guides for using models and inference platforms. ### Tech Stack: Python ### Why it Matters: This cookbook serves as a valuable resource for developers looking to leverage GenAI for various applications, providing a blend of theoretical knowledge and practical examples. It is aimed at beginners and offers insights into how to effectively use APIs and frameworks from leading platforms, making it a practical guide for developers.

Jupyter Notebook 241 stars

guidelabs/steerling

Interpretable Causal Diffusion Language Models

### Overview: Steerling is an interpretable causal diffusion language model that combines masked diffusion language modeling with concept decomposition, enabling non-autoregressive text generation, concept attribution, intervention on concept activations, and extraction of composed representations. ### Key Features: - **Generation**: Non-autoregressive text generation via confidence-based unmasking. - **Attribution**: Decompose predictions into concept contributions, and attribute generated tokens to input tokens. - **Steering**: Intervene on concept activations to control generation. - **Embeddings**: Extract hidden, composed, known, or discovered representations. ### Tech Stack: Jupyter Notebook ### Why It Matters: Steerling-8B combines masked diffusion language modeling with concept decomposition, enabling interpretable and controllable text generation, concept attribution, intervention on concept activations, and extraction of composed representations. It is designed to be used in a variety of applications, including text generation, language modeling, and natural language processing tasks.

Python 241 stars

hao-ai-lab/JetSpec

JetSpec++/JetSpec: Causal Parallel (Tree) Drafting with Self Refinement

Overview: JetSpec is an implementation of causal parallel tree drafting for fast LLM speculative decoding inference, achieving up to 9.64x end-to-end speedup on MATH-500 tasks using B200 GPUs. Key features: - Causal parallel tree drafting - One-pass drafting efficiency - Restoration of causal branch conditioning - Strong gains across reasoning, code, and chat workloads Tech stack: Python Why it matters: JetSpec offers significant speedup and efficiency improvements in speculative decoding, making it a valuable tool for developers and researchers working with large language models. Repository: hao-ai-lab/JetSpec Primary language: Python GitHub stars: 241 GitHub description: JetSpec: Breaking the Scaling Ceiling of Speculative Decoding with Causal Parallel Tree Drafting

HTML 241 stars

posgnu/rci-agent

A codebase for "Language Models can Solve Computer Tasks"

Overview: This project is a codebase for a research paper titled "Language Models can Solve Computer Tasks". It implements a RCI agent using pre-trained language models to execute tasks on the MiniWoB++ benchmark. Key features: - RCI (Reinforcement Clustering) agent for executing computer tasks - Uses pre-trained language models (e.g., GPT-3, Ada, Babbage) for task execution - Simple RCI prompting scheme for improved outputs Tech stack: Python 3.9 Why it matters: This project showcases the potential of using language models to solve complex computer tasks, demonstrating that language models can be effective in executing tasks that require a high level of expertise and knowledge. Repository: posgnu/rci-agent GitHub stars: 241 GitHub description: A codebase for "Language Models can Solve Computer Tasks"

Jupyter Notebook 241 stars

shenasa-ai/speech2text

A Deep-Learning-Based Persian Speech Recognition System

### Overview: This project is a Deep-Learning-Based Persian Speech Recognition System that uses various toolkits and the authors' own implementations. It includes ASR models trained on a large dataset of Persian speech recordings. ### Key features: - **Model Training**: Uses Mozilla Deep Speech, DeepSpeech2, and Wave2vec for ASR model training. - **Data Collection**: Collects Persian speech data from various sources, including radio archives. - **Preprocessing**: Applies spectrograms, MFCC, and filter bank to preprocess audio data. - **Model Outputs**: Provides model outputs with loss metrics and confidence levels for transcription accuracy. ### Tech stack: Python with Jupyter Notebook and various deep learning libraries and toolkits. ### Why it matters: This project offers a practical solution for Persian speech recognition, addressing the lack of available datasets and providing a robust ASR system that can be used in various applications.

Python 241 stars

zjunlp/AutoAct

[ACL 2024] AutoAct: Automatic Agent Learning from Scratch for QA via Self-Planning

### Overview: AutoAct is an automatic agent learning framework designed for QA tasks. It leverages self-planning to automatically synthesize and differentiate based on limited data and a tool library, achieving better performance compared to strong baselines. ### Key Features: - Automatic tool selection - Trajectories synthesis - Self-differentiation - Group planning ### Tech Stack: Python ### Why it Matters: AutoAct simplifies the learning process by eliminating the need for large-scale annotated data and closed-source models, making it more accessible and cost-effective. This framework enables the development of language agents that can perform multiple functions effectively, without the limitations of single-model performance. Repository: zjunlp/AutoAct GitHub Stars: 241 GitHub Description: [ACL 2024] AutoAct: Automatic Agent Learning from Scratch for QA via Self-Planning

Python 240 stars

Extraltodeus/J-Wash

Jacobian-Brainwash : A manual alignment tool for large language models built on Anthropic's Jacobian Lens. Results are exportable.

### Overview: J-Wash is a powerful tool for local editing and exporting large language models, allowing users to reshape a model's identity and behavior without the need for training or fine-tuning. It uses the Jacobian lens from Anthropic to visualize and edit the model's J-space, then exports the changes as a pure-weights checkpoint, enabling seamless integration into existing systems. ### Key features: - **Interactive Chat UI**: Real-time visualization of the Jacobian lens. - **Token Editor**: Converts lens directions into persistent, composable edits. - **Export Pipeline**: Bakes edits into a pure-weights checkpoint for use in any Hugging Face model. ### Tech stack: Python ### Why it matters: J-Wash simplifies the process of aligning large language models, making it accessible for non-experts. By allowing direct manipulation of the model's J-space, users can quickly modify a model's identity or behavior without the need for complex training or fine-tuning. This tool is particularly useful for developers and researchers looking to explore and experiment with model modifications in a controlled environment.

Go 240 stars

VictoriaMetrics/mcp-victoriametrics

The implementation of Model Context Protocol (MCP) server for VictoriaMetrics

### Overview: The implementation of Model Context Protocol (MCP) server for VictoriaMetrics, providing access to VictoriaMetrics instances and seamless integration with VictoriaMetrics APIs and documentation. ### Key Features: - **Querying Metrics and Data**: Access to metrics and data exploration. - **Listing and Exporting Metrics**: List and export available metrics, labels, labels values, and entire series. - **Analyzing and Testing Rules**: Analyze and test alerting and recording rules and alerts. - **Parameters of VictoriaMetrics Instance**: Show parameters of your VictoriaMetrics instance. - **Cardinality Analysis**: Explore cardinality of your data and metrics usage statistics. - **Query Analysis and Debugging**: Analyze, trace, prettify, and explain queries. - **Integration with VictoriaMetrics Cloud**: Integration with VictoriaMetrics Cloud. - **Embedded Documentation**: Embedded up-to-date documentation. - **Documentation Search**: Search documentation without online access. ### Tech Stack: Go ### Why it Matters: The MCP server provides a comprehensive interface for monitoring, observability, and debugging tasks related to VictoriaMetrics instances, enabling advanced automation and interaction capabilities for engineers and tools.

Python 240 stars

backprop-ai/backprop

Backprop makes it simple to use, finetune, and deploy state-of-the-art ML models.

Overview: Backprop simplifies the use, fine-tuning, and deployment of state-of-the-art machine learning models, making it accessible for developers without extensive expertise. Key features: - Solves a variety of tasks with pre-trained models or finetuning in one line for custom use cases. - Supports conversational question answering, text classification, image classification, text vectorization, image vectorization, summarization, emotion detection, and text generation. - Offers pre-built tasks and models for easy integration and deployment. Tech stack: Python Why it matters: Backprop provides a straightforward way to leverage advanced machine learning models, making it accessible for developers and enabling them to focus on application development rather than deep model optimization.

Unknown 240 stars

coderonion/awesome-snn

🔥🔥🔥A collection of some awesome public SNN(Spiking Neural Network) projects.

Overview: This repository is a curated list of some awesome public SNN (Spiking Neural Network) projects. Key features: - Frameworks - Datasets - Applications Tech stack: Python, PyTorch Why it matters: This repository serves as a valuable resource for developers interested in exploring and utilizing Spiking Neural Network projects, providing a comprehensive overview of various frameworks, datasets, and applications in the field.

Python 240 stars

ictnlp/FlexRAG

FlexRAG: A RAG Framework for Information Retrieval and Generation.

### Overview: FlexRAG is an open-source framework designed to streamline the rapid reproduction, development, and evaluation of Retrieval-Augmented Generation (RAG) systems. It provides comprehensive support for various RAG scenarios, including text-based, multimodal, and web-accessible applications. With an end-to-end pipeline covering everything from data preparation to system evaluation, FlexRAG empowers researchers to efficiently share their work with the community and swiftly develop demonstrative prototypes based on their algorithms. ### Key Features: - **End-to-End Pipeline**: FlexRAG covers the entire process from data preparation to system evaluation, making it highly versatile. - **Comprehensive Support**: It supports text-based, multimodal, and web-accessible applications, ensuring flexibility. - **RAG Framework**: FlexRAG is built on a retrieval augmented generation framework, enhancing the capabilities of RAG systems. - **Metrics and Benchmarks**: It includes various information retrieval metrics and benchmarks for evaluating RAG systems. ### Tech Stack: Python ### Why It Matters: FlexRAG is an innovative open-source framework that simplifies the process of developing and evaluating RAG systems. It provides a comprehensive solution for researchers and developers looking to streamline their workflows and share their work efficiently. ### GitHub Stars: 241 ### Repository: - Repository: ictnlp/FlexRAG - Primary language: Python - GitHub stars: 241 ### Table of Contents: - 📖 Table of Contents - ✨ Key Features - 📢 News - 🚀 Getting Started

C++ 240 stars

intel/ScalableVectorSearch

### Overview: Scalable Vector Search (SVS) is a high-performance library designed for vector similarity search, enabling applications to achieve remarkable accuracy and speed on Intel(R) Xeon(R) CPUs. SVS supports billions of high-dimensional vectors, offering superior performance and memory efficiency compared to alternatives. ### Key Features: - **Locally Adaptive Vector Quantization (LVQ)**: Optimized indexing and search algorithms. - **High Accuracy**: Achieves state-of-the-art performance and accuracy. - **Billions of Vectors**: Supports billions of high-dimensional vectors. - **State-of-the-Art Speed**: Provides high performance with a reduced memory footprint. - **Python API**: Fully-featured and compatible with standard libraries. - **Intel(R) Xeon(R) Support**: Optimized for Intel(R) Xeon(R) CPUs, including 2nd generation and newer. ### Tech Stack: C++ ### Why it Matters: SVS enables developers to harness the power of similarity search on Intel(R) Xeon(R) CPUs, achieving superior performance and accuracy on billion-scale datasets compared to alternatives. ### Repository: intel/ScalableVectorSearch ### GitHub Stars: 240 ### GitHub Description: No description provided. ### README Excerpt: No description provided.

Python 240 stars

zd11024/NaviLLM

[CVPR 2024] The code for paper 'Towards Learning a Generalist Model for Embodied Navigation'

### Overview: NaviLLM is a generalist model designed for embodied navigation, achieving state-of-the-art performance on CVDN, SOON, and ScanQA. It adapts LLMs to embody navigation by introducing schema-based instruction, enabling integration of diverse data sources and demonstrating strong generalizability. ### Key Features: - **Schema-Based Instruction**: Flexibly casts various tasks into generation problems to unify a wide range of tasks. - **Unified Model Training**: Integrates diverse data sources from various datasets, equipping NaviLLM with a wide range of capabilities required by embodied navigation. - **State-of-the-Art Performance**: Achieves significant improvements over previous methods on CVDN, SOON, and ScanQA. ### Tech Stack: Python ### Why It Matters: Embodied navigation is crucial for agents to interact with the world effectively. By proposing NaviLLM, we address the limitations of task-specific agents and provide a versatile solution that can handle a broad range of tasks, making it a significant advancement in the field of embodied navigation.

Python 239 stars

AXYZdong/AMchat

AM (Advanced Mathematics) Chat is a large language model that integrates advanced mathematical knowledge, exercises in higher mathematics, and their solutions. AM (Advanced Mathematics) chat 高等数学大模型。一个集成数学知识和高等数学习题及其解答的大语言模型。

Overview: AM (Advanced Mathematics) Chat is a large language model designed to integrate advanced mathematical knowledge and exercises in higher mathematics, providing solutions to these problems. Key features: - Integrates advanced mathematical knowledge and exercises in higher mathematics. - Offers solutions to various mathematical problems. - Supports multiple languages including English, Chinese, and Japanese. Tech stack: Python Why it matters: This model is useful for students and educators looking to enhance their understanding of advanced mathematics through interactive and practical problem-solving. Repository: AXYZdong/AMchat Primary language: Python GitHub stars: 238