Projects

### Overview: ImageGen is a local AI image generation tool for macOS that runs on Apple Silicon, providing offline and private image generation capabilities. It uses Qwen-Image 2.1 and a built-in MCP server for Claude Code and other AI agents, making it suitable for developers and creatives. ### Key Features: - **Fully on-device**: Prompts and images never leave your Mac. - **Text to image**: Strong typography with readable text inside images. - **Image editing**: Change images with instructions or combine up to 10 reference images. - **Transparent PNGs**: Native RGBA output for stickers, icons, and assets. - **MCP server for AI agents**: Supports Claude Code and other MCP clients for image generation and editing. - **One-click model download**: From Hugging Face, resumable within the app. - **Memory guards**: Prevents big jobs from freezing your Mac. - **Native SwiftUI app**: Menu bar icon, keeps the engine running even when the window is closed. ### Tech Stack: SwiftUI ### Why It Matters: - **State-of-the-art model**: Fits in about 33 GB of memory, making it ideal for Mac users. - **Local image generation**: No need for cloud services or API keys, eliminating monthly bills. - **Native performance**: Runs on your Mac's GPU (Metal) for fast and efficient image generation. - **Built-in MCP server**: Supports Claude Code and other AI agents, allowing for seamless image creation and editing. ### Gallery: - **Neon Sign**: A neon sign reading "

Overview: This repository contains a collection of n8n workflow templates designed for building intelligent automations using Pinecone. Key features: - Assistant quickstart - Database quickstart - Chat with your Google Drive docs Using Pinecone Assistant - Chat with your Google Drive docs Using Pinecone Vector Database - Extract insights from LinkedIn comments using Apify and Pinecone Assistant - Query support docs via Dropbox and Slack using Pinecone Assistant - Vacation rental property manager with multiple Assistants - DataCamp code along: Build a customer assistant with multi-domain RAG Tech stack: n8n, Pinecone Assistant, Open AI, Google Drive, Apify, Dropbox, Slack, Pinecone Vector Store Why it matters: These templates provide a quick and easy way to build intelligent automations using Pinecone, making it easier for developers to integrate various sources and processes into their workflows.

Overview: This project fine-tunes the Qwen2-1.5B model specifically for Arabic language tasks, utilizing Quantized LoRA (QLoRA) techniques. The model's performance is evaluated on the ArabicMMLU benchmark, showcasing strong parameter efficiency and competitive performance across various knowledge domains. Key features: - Fine-tuning of Qwen2-1.5B model for Arabic language tasks - Utilization of Quantized LoRA (QLoRA) for efficient model optimization - Evaluation on the ArabicMMLU benchmark Tech stack: Python Why it matters: This project demonstrates the effectiveness of using QLoRA for parameter-efficient language model fine-tuning, providing a practical solution for Arabic language tasks that aims to balance performance and resource efficiency.

Overview: Compute is a decentralized network that allows users to run compute jobs on their machines, including AI inference, transcoding, and storage, without needing to set up complex infrastructure. Key features: - Runs on Linux (x86_64, arm64) and macOS (Intel, Apple Silicon); Windows inside WSL2. - Network: Outbound internet is required, no account or port forwarding needed. - What it does: AI inference, FFmpeg transcoding, and storage jobs from the network. - What you get: Paid inference is billed through CoinPay, and you can set your payout address with `infernet payout set <coin> <address>`. Paid demand is early, so expect small, irregular earnings. - House models (on by default): The node pulls and serves models from Infernet, including Profullstack's `saas-business-strategist-agent` and `huihui_ai/qwen3.5-abliterated:4b`/`:9b`. You can opt out with `--no-house-models` or `--no-uncensored-models`. - Invite people: Share `https://c0mpute.com/r/<your-handle>` and nodes installed from it show your handle on the board. Why: The goal is 100,000 nodes, which will provide enough pooled capacity to train and serve open-weight models on hardware owned by the people running the network. Tech stack: Rust, Bun, and the c0mpute CLI. Why it matters: The decentralized, open-source

Overview: uzdabrawza is a Skynet-powered browser automation script that harnesses neural brainfuck and machine learning chaos to provide zero shits about anything while still working perfectly. Key features: - 9 fucking neural overlords - Complete Big Brother surveillance - Terminator stealth mode - Organized digital anarchy - Zero corporate Matrix bullshit Tech stack: Python Why it matters: This script is a unique and effective tool for automating tasks with a cyberpunk twist, using neural brainfuck and machine learning chaos to achieve results that are both unconventional and highly effective. Repository: psyb0t/uzdabrawza GitHub stars: 12 GitHub description: The Anal-Queen of AI Browser Automation 🏴‍☠️ A beautifully fucked-up Skynet-powered browser automation script that harnesses neural brainfuck and machine learning chaos to give zero shits about anything while somehow still working perfectly. Smells like smegma but runs like a dream.

Overview: Z Image Turbo is a high-speed AI image generation model offered by Alibaba, designed to generate stunning images from text prompts with lightning-fast processing and flexible aspect ratios. Key features: - Supports multiple models and pricing tiers - Compatible with OpenAI's API endpoint - Charges per image delivered, not per usage - Asynchronous processing with task tracking Tech stack: Python Why it matters: Z Image Turbo provides a fast and efficient way to generate high-quality images using text prompts, making it ideal for developers looking to streamline their image generation workflows without the need for expensive subscriptions or monthly fees. Repository: qxhjep3w5me/z-image-api-fr Primary language: Python GitHub stars: 9 GitHub description: Z Image Turbo API (`z-image-turbo`): model id, pricing, parameters and copy-paste examples for the OpenAI-compatible APIMart endpoint.

### Overview: Context Runtime is a database query planner designed for Large Language Model (LLM) context. It decides what a model sees before it answers, plans the query, runs it through a reused substrate, and learns from the outcome. ### Key Features: - Context Selection and Optimization - Query Planning and Simulation - Learning from Outcomes ### Tech Stack: Python ### Why It Matters: A "query planner for LLM context": the application states a goal and a budget; the runtime decides what the model sees, emits an inspectable, replayable plan, and learns from the outcome. It is the Context part of ReDevOps Runtime: context selection and optimization.

Overview: re-rl is a Python library designed for generating diverse datasets for training language models with reinforcement learning (RL) and artificial datasets. It supports various tasks such as mathematical reasoning, physics, machine learning, reinforcement learning, formal proofs, and visual tasks, including VLMs. Key features: - Generates datasets for pretraining, SFT, and RL tasks. - Provides a variety of datasets for different levels of difficulty. - Includes a catalog of tasks, examples, and notebooks for easy experimentation. Tech stack: Python Why it matters: re-rl simplifies the process of creating and using datasets for training language models, making it easier for developers to experiment with various tasks and levels of difficulty, ultimately improving the performance and accuracy of their models. Repository: researchim-ai/re-rl GitHub stars: 12 GitHub description: train llms with rl and artificial datasets

### Overview: Seedream 4 Prompts — 8 Tested Recipes With Outputs <!-- conv-kit:v1 --> <p align="center"> <img src="assets/badges/price.svg" alt="observed unit price"> <img src="assets/badges/billing.svg" alt="billing model"> <img src="assets/badges/compat.svg" alt="OpenAI-compatible endpoint"> </p> <p align="center"> <img src="assets/01-product-hero-perfume.jpg" width="820" alt="Sample render via the APIMart image API"> </p> > **from $0.0195 per image** on the APIMart route — `https://api.apimart.ai/v1` OpenAI 호환 엔드포인트 하나면 되고, 월 구독 플랜은 필요 없습니다. *(2026-10-12 기준)* **[Get an API key](https://go.apimart.ai/k-979054)** · **[Live pricing](https://go.apimart.ai/k-ae15aa)** · **[Model page](https://go.apimart.ai/k-41d620)** · [⚡ 60-second quickstart](#quickstart) **팀들이 APIMart를 통해 Seedream 4(`seedream-4-0`)을 쓰는 이유** - **키 하나로 카탈로그 전체를.**

Overview: This is a Microsoft .NET native client for Weaviate, designed to simplify interaction with the Weaviate API. Key features: - Easy integration with .NET applications - Support for various Weaviate functionalities - Compatibility with different Weaviate versions Tech stack: C# Why it matters: This client streamlines the process of working with Weaviate, making it easier for developers to integrate with the API and access its features.

Overview: Chinese-Multimodal-Sentiment-Analysis is an innovative project that integrates advanced deep learning techniques to analyze sentiments in Chinese language data. Utilizing a sophisticated blend of textual, audio, and visual modalities, this project offers a comprehensive approach to understanding and interpreting complex emotional expressions. Key features: - Multimodal Annotations: Each segment in the dataset is annotated with sentiments based on multimodal information, providing insights from text, audio, and visual modalities. - Diverse Sources: The dataset includes video segments from movies, TV series, and other multimedia sources, ensuring a diverse range of expressions and contexts. - Fine-Grained Analysis: CH-SIMS supports fine-grained sentiment analysis, allowing for a detailed understanding of emotional states and nuances in Chinese language content. Tech stack: Python Why it matters: This project is crucial for researchers and practitioners aiming to perform sentiment analysis in Chinese, leveraging the power of multimodal learning techniques. It provides comprehensive resources and tools for analyzing sentiments in Chinese language data using advanced deep learning methods. Repository: shengwei-peng/Chinese-Multimodal-Sentiment-Analysis GitHub stars: 12 GitHub description: Chinese-Multimodal-Sentiment-Analysis is an innovative project that integrates advanced deep learning techniques to analyze sentiments in Chinese language data. Utilizing a sophisticated blend of textual, audio, and visual modalities, this project offers a comprehensive approach to understanding and interpreting complex emotional expressions.

Overview: This project demonstrates how to integrate Open AI APIs with Llama Index and Milvus Vector Database for testing Retrieval Augmented Generation (RAG). Key features: - Scraping a website - Generating embeddings - Loading embeddings into Milvus - Spinning up a Gradio chat for RAG testing Tech stack: Python Why it matters: This example provides a simple yet effective way to test and demonstrate the integration of Open AI APIs with Llama Index and Milvus for RAG, making it useful for developers looking to understand and implement similar solutions.

Overview: rvLLM is a Rust-based implementation of Gemma 4 inference, focusing on efficient model loading, scheduling, and serving. It includes a CUDA runtime and server, as well as Metal, vision, and XLA/TPU work. Key features: - CUDA runtime and server - Metal integration - Vision model integration - XLA/TPU experimental operator and conformance work Tech stack: Rust Why it matters: rvLLM offers a native Rust implementation of Gemma 4 inference, optimizing performance and resource utilization. It supports multiple accelerators and provides comprehensive validation and debugging tools. Repository: solidsf-inc/rvLLM GitHub stars: 12 GitHub description: rvLLM — Gemma 4 inference in Rust

### Overview: 即梦视频 — Seedance 2.0 prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`seedance-2.0`) and the cost per image. ### Key Features: - **Prompt Structure**: Recipes for Seedance 2.0 with example renders, aspect ratios, model id (`seedance-2.0`), and cost per image. - **Model Compatibility**: OpenAI-compatible endpoint `https://api.apimart.ai/v1` with a base URL and `Authorization` header. - **Cost Calculation**: Real-time cost per second, with a minimum charge of $1 for each call. ### Tech Stack: Python ### Why It Matters: - **Single Key Access**: A single API key allows access to Seedance 2.0 and 300+ models through a consistent base URL and `Authorization` header. - **Cost Efficiency**: No need for pre-paid subscriptions; $1 minimum charge for each call, with real-time cost calculation. - **Immediate Feedback**: Each task reports actual cost, and failed tasks automatically refund. ### Repository: Repository: t7phf53f5hs/doubao-seedance-2-0-prompts-cn ### Primary Language: Python ### GitHub Stars: 10

### Overview: Sentiment analysis of Darija, a Moroccan language, can provide valuable insights into customer satisfaction and improve products or services accordingly. ### Key features: - Fine-tuning a BERT model for sentiment analysis. - Analyzing social media posts and text data to identify customer satisfaction. - Generating a pie chart to visualize overall customer satisfaction. ### Tech stack: Jupyter Notebook ### Why it matters: Sentiment analysis is crucial for businesses and organizations seeking to better understand customer satisfaction and improve their products or services.

Overview: FastLLM.cpp is a C++ implementation of an LLM service designed for low latency and performance optimization, aiming to reduce transaction times by up to 500ms compared to using Python or external APIs. Built on llama.cpp and GGML, it provides a fast, low-latency solution for accessing large language models. Key features: - Low latency and performance optimization - Built on llama.cpp and GGML - Can be forked and customized for specific use cases - GPU support for faster inference (optional) Tech stack: C++ Why it matters: This project offers a practical solution for developers looking to optimize the performance of large language model access, particularly in environments where speed and low latency are critical.

Overview: This project implements Canonical Correlation Analysis (CCA) for joint representation of images and text, utilizing CNN features from the MS COCO database. It aims to improve image-to-tag and tag-to-image search functionalities. Key features: - Joint representation of image and text using CCA - Utilization of CNN features for image and text extraction - Implementation based on the paper "Joint Representation of Image and Text through Canonical Correlation Analysis" by Yunchao et al. Tech stack: Python Why it matters: This project offers a novel approach to image and text search by leveraging CCA, potentially enhancing the accuracy and relevance of search results in various applications. Repository: theevann/Image-and-Text-Search Primary language: Python GitHub stars: 12

### Overview: Shazam-inspired music recognition system designed for a Big Data project, utilizing FAISS vector search and spectral fingerprinting for robust audio recognition. ### Key Features: - **FAISS Vector Search**: Powerful for fast and accurate music recognition. - **Spectral Fingerprinting**: Inspired by Shazam's original patent, enhancing recognition accuracy. - **CLAP and MuQ Embeddings**: Utilized for high-quality audio analysis. - **FastAPI Backend**: Provides a RESTful API for real-time music recommendations. - **React Frontend**: Facilitates user-friendly interface for streaming music. ### Tech Stack: - Python - FastAPI - FAISS - ChromaDB - SQLite - PyTorch - CLAP - MuQ ### Why It Matters: This system combines the efficiency of FAISS with the accuracy of spectral fingerprinting, making it highly effective for music recognition, especially in noisy or short audio clips.

### Overview: LangChain.Databases is a project within the LangChain.NET ecosystem that provides database abstractions and vector store backends. It aligns with the Microsoft.Extensions ecosystem and uses MEVA for vector search. The project includes separate abstractions and can be used independently. ### Key features: - **PostgreSQL + pgvector**: Vector store backend for PostgreSQL. - **OpenSearch k-NN**: Vector store backend for OpenSearch. - **In-memory, file-based**: Chat message history backend. - **Redis**: Vector store backend for Redis. - **MongoDB**: Vector store backend for MongoDB. ### Tech stack: C# ### Why it matters: This project simplifies the integration of vector search and backends with the LangChain.NET ecosystem, making it easier to implement and use for developers.

Overview: Generate novel text based on input text, finetuned from SKT KoGPT2 base v2. Key features: - Generates novel text from input text - Finetuned from SKT KoGPT2 base v2 - Available in Jupyter Notebook, Colab, and Okteto web demos - Supports Japanese web novel finetuning Tech stack: Python, Jupyter Notebook, FastAI, Transformers, BentoML Why it matters: This tool is useful for generating original stories based on input text, making it valuable for creative writing projects and generating unique narratives.