Projects

### Overview: GPT Image 1.5 Prompts offers tested image prompt recipes with example renders, aspect ratios, the model ID (`gpt-image-1.5-official`), and the cost per image. This repository is a resource for developers looking to integrate ChatGPT's image generation capabilities with their projects. ### Key features: - **Prompt Recipes**: 8 tested image prompts with detailed instructions. - **Aspect Ratios**: 1:1, 3:4, 4:3, etc., for various image formats. - **Model ID**: `gpt-image-1.5-official` for seamless integration with OpenAI's API. - **Cost**: $4.00 per image, with a minimum of $1.00 per call. ### Tech stack: Python ### Why it matters: This repository provides a straightforward and effective way for developers to incorporate image generation capabilities into their projects, ensuring high-quality results without the need for complex setup or expensive subscriptions.

Overview: This repository is a comprehensive resource for AI product photography prompts, e-commerce image workflows, and product photo editing templates. It serves as a practical, SEO-friendly guide for Shopify, Amazon, DTC brands, product marketers, agencies, and e-commerce teams looking for useful prompt systems, repeatable workflows, and production-ready AI creative assets. Key features: - AI Image Generator - AI Product Photos - AI Product Photography Workflow - AI E-commerce Photos - Pro Image Editor - AI Background Remover - Universal Upscaler Tech stack: Unknown Why it matters: This repository provides a structured approach to creating high-quality AI-generated product photography assets, ensuring that each asset is tailored to specific goals and audiences, thereby enhancing SEO and improving the overall quality of e-commerce and product photography workflows.

### Overview: Grok Imagine Image Quality prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`grok-imagine-image-quality`) and the cost per image. ### Key features: - **Model ID:** grok-imagine-image-quality - **Aspect Ratios:** 1:1, 4:3, 16:9, 85mm, 24mm, macro, overhead flat lay - **Cost:** $0.04 per image ### Tech stack: Python ### Why it matters: - **Key:** A single API key allows access to Grok Imagine Image Quality and 300+ models. - **Billing:** $1 minimum, pay-as-you-go. - **Cost:** Real-time cost reported for each task. - **Asynchronous:** Simple queue-based processing for batch and retry handling. ### Repository: cm2dnh/grok-imagine-image-quality-prompts-cn ### Primary language: Python ### GitHub stars: 11 ### GitHub description: Grok Imagine Image Quality prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`grok-imagine-image-quality`) and the cost per image.

Overview: Nano Banana Pro API, also known as Bananen-KI or Nano-Banane, is a Nano Banana product that provides access to an OpenAI-compatible API endpoint for image generation and editing. It supports 4K image generation and offers a key that allows access to 300+ models, including Nano Banana Pro. Key features: - Access to 300+ models via a single API key - Pricing of $0.03 per image (4K $0.04) with pay-as-you-go - Real-time billing and cost reporting - Asynchronous processing with task ID tracking Tech stack: Python Why it matters: This API provides a cost-effective and flexible solution for image generation and editing, allowing teams to quickly and easily integrate advanced AI models into their workflows without the need for a dedicated API key or pre-dimensioned ABO. Repository: cmpxx2w/gemini-3-pro-image-preview-api-de GitHub stars: 11

### Overview: This repository contains 8 tested recipes for the Qwen-Image 3.0 Pro model, showcasing various aspects of image prompts and their corresponding costs. ### Key Features: - **Model ID:** `qwen-image-3.0-pro` - **Prompt Structure:** The recipes include the subject, light, framing, setting, and finish, providing a structured approach to image prompts. - **Cost:** The cost per image is $0.0286, with a total estimated cost of $0.2286 for the set. ### Tech Stack: Python ### Why It Matters: - **Key Features:** The recipes are designed to guide the model towards specific outcomes, ensuring the desired image is produced efficiently. - **Cost Efficiency:** The model is optimized to minimize costs, making it suitable for projects where cost-effectiveness is a priority. - **Versatility:** The model can be used across various applications, from e-commerce product shots to editorial food photography, thanks to the flexible prompt structure. ### Repository: cnc4l4x4p3ndb0/qwen-image-3-0-pro-prompts-cn ### Primary Language: Python ### GitHub Stars: 11

Overview: This repository contains tested image prompt recipes for the GPT Image 1 model, including example renders, aspect ratios, and the cost per image. Key features: - 8 tested image prompt recipes - Example renders, ratios, and call costs - Model: `gpt-image-1` - Effective price: $4.00 per image Tech stack: Python Why it matters: This repository provides a practical guide for developers looking to use the GPT Image 1 model effectively, ensuring high-quality image prompts at a reasonable cost.

### Overview: Midjourney API (`midjourney`) provides a programmable API endpoint for generating images and videos using the Midjourney model, offering a wide range of styles and functionalities without the need for a monthly subscription. ### Key features: - **Model ID and Routes**: The API supports various model IDs (`midjourney`, `v8.1`, `v7`, `v6.1`, `v5.2`, `v5.1`, `Niji 7`, `Niji 6`, `Niji 5.1`, `Niji 5.2`, `Niji 7`, `Niji 6`, `Niji 5.1`, `Niji 5.2`, `Niji 7`, `Niji 6`, `Niji 5.1`, `Niji 5.2`, `Niji 7`, `Niji 6`, `Niji 5.1`, `Niji 5.2`, `Niji 7`, `Niji 6`, `Niji 5.1`, `Niji 5.2`, `Niji 7`, `Niji 6`, `Niji 5.1`, `Niji 5.2`, `Niji 7`, `Niji 6`, `Niji 5.1`, `Niji 5.2`, `Niji 7`, `Niji 6`, `Niji 5.1`, `Niji 5.2

Overview: Wan 2.7 API (`wan2.7`) is an advanced AI video model compatible with OpenAI, providing high-quality visuals and professional performance. It offers a base URL and a single API key for access to various models, including image-to-video, with flexible parameters for different resolutions and durations. Key features: - Supports OpenAI-compatible API - Offers superior visual quality and high motion coherence - Supports image-to-video and video continuation - Single API key for multiple models - Flexible parameters for different resolutions and durations - Billing by generated second, with effective price of $0.0664 per second Tech stack: Python Why it matters: Wan 2.7 API simplifies access to advanced AI video models, making it ideal for developers looking to integrate AI into their projects without the need for complex integrations or long-term commitments. GitHub stars: 11 Repository: d335whdzs8i43/wan2-7-api-ru

Overview: This repository contains 8 tested recipes for Wan 2.7 prompts, showcasing example renders, aspect ratios, and the model ID (`wan2.7`). The cost per image on the APIMart route is $0.0664 per second, with no need for a monthly billing plan. Key features: - 8 Wan 2.7 prompts with example renders - 4 recipes, each with a specific subject, light, framing, setting, and finish - The effective price used is $0.0664 per second, estimated cost of the set is $0.2656 Tech stack: Python Why it matters: This repository provides a concise guide to crafting effective Wan 2.7 prompts, ensuring that the model renders the desired outcome efficiently and cost-effectively. Repository: d335whdzs8i43/wan2-7-prompts-ru Primary language: Python GitHub stars: 11 GitHub description: Wan 2.7 prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`wan2.7`) and the cost per image on the APIMart route.

Overview: Nano Banana 2 API is an OpenAI-compatible endpoint providing fast AI image generation with native 4K output and editing capabilities. It offers a cost-effective solution with flexible pricing and billing models. Key features: - Model id: `gemini-3.1-flash-image-preview` - Pricing: $0.015 per image, with tiers for 0.5K, 1K, 2K, and 4K outputs - Asynchronous processing: Tasks are batched and retried in the background - No subscription or upfront payment required - Compatible with 300+ image, video, and language models - Supports native 4K output Tech stack: Python Why it matters: Nano Banana 2 API simplifies the process of generating high-quality images with OpenAI-compatible capabilities, making it an ideal choice for developers looking to integrate AI-powered image generation into their projects without the complexity of managing a subscription. Repository: d63hy9t-4pp/nano-banana-2-api-es Primary language: Python GitHub stars: 11 GitHub description: Nano Banana 2 API (`gemini-3.1-flash-image-preview`): model id, pricing, parameters and copy-paste examples for the OpenAI-compatible APIMart endpoint.

Overview: This repository contains a native Windows tooling for fast local LLM inference on the AMD Strix Halo (gfx1151) GPU, specifically optimized for ROCmFPX llama.cpp fork. It includes a headless OpenAI-compatible server and benchmarks, with tested performance on a Minisforum MS-S1 MAX (128 GB unified LPDDR5X) and other GPUs. Key features: - ROCmFPX llama.cpp fork for Windows - Native build - Headless OpenAI-compatible server - Benchmarks - Tested on AMD Strix Halo (gfx1151), runs on other GPUs too (CUDA / Vulkan / CPU backends included) Tech stack: PowerShell Why it matters: This repository fills the gap for Windows-only ROCmFPX llama.cpp builds, providing a native Windows tooling that is optimized for AMD hardware, including the AMD Strix Halo GPU. The ROCmFPX fork adds AMD-specific GGUF weight formats, allowing for better performance and quality on this specific GPU.

Overview: This project is a MCP server that exposes your Obsidian Smart Connections vector database to Claude Code via semantic search, enabling semantic search across your vault. Key features: - Semantic search for finding notes by meaning, not keywords - Find related notes (like Smart Connections sidebar) - Get context for RAG queries Tech stack: Python Why it matters: This project provides a way to leverage Claude Code's semantic search capabilities for better context and relevance in your Obsidian vault, enhancing the user experience by allowing for more meaningful and contextually relevant searches.

Overview: LoTR is an innovative approach to parameter-efficient fine-tuning of large language models, specifically designed for low tensor rank adaptation. Key features: - Low Tensor Rank adaptation - Novel method for parameter-efficient fine-tuning - Compression of a sequence of layers with a low-rank tensor representation - Extremely cheap and fast downstream fine-tuning Tech stack: Python Why it matters: LoTR offers significant improvements in parameter efficiency and faster fine-tuning compared to existing methods, making it particularly beneficial for deep models and large language models.

Overview: This repository contains scripts for gathering, parsing, filtering, and visualizing declared taxes data from Country-by-country reports. Key features: - PDF downloader - Streamlit app for interactive data exploration - Configurable pipeline for customizable page filtering and table extraction Tech stack: Jupyter Notebook Why it matters: This project provides a comprehensive toolset for developers interested in analyzing and visualizing tax data, making it easier to understand and report on global tax compliance.

### Overview: A medical-focused LLaMA 3.2 fine-tuned model, optimized for healthcare-specific understanding and tasks, designed for educational, initial symptom triage, and research purposes. ### Key Features: - **Model Description:** A domain-adapted version of Meta's LLaMA 3.2 3B Instruct fine-tuned on medical datasets. - **Model Type:** Medical Chatbot - **Language(s):** English - **License:** Apache 2.0 - **Finetuned from model:** meta-llama/Llama-3.2-3B-Instruct ### Tech Stack: Jupyter Notebook ### Why it Matters: This model is optimized for tasks such as medical Q&A, symptom checking, and patient education, making it valuable for educational and research purposes in the healthcare field.

### Overview: The Advanced Recommender System is a cutting-edge platform designed to enhance academic research and document discovery. Built with advanced indexing, machine learning, and personalized search algorithms, it streamlines the retrieval, classification, and recommendation of academic documents tailored to individual preferences. ### Key Features: - **End-to-End Pipeline**: Seamless integration of machine learning and deep learning models for improved document categorization and organization. - **Personalized Search**: Tailored search and recommendation engines enhance user experience. - **Advanced Search Algorithms**: Vector space models, spell correction, and probabilistic ranking models for precise search results. - **Web Crawling & Personalized Recommendations**: Enhances the system by incorporating web crawling and personalized recommendation engines. ### Tech Stack: Jupyter Notebook ### Why it Matters: This system revolutionizes academic research by providing a more efficient and personalized way to access and organize academic documents, making it an invaluable tool for researchers and students alike. --- Repository: deepmancer/advanced-recommender-system GitHub Stars: 11 GitHub Description: Advance information retrieval system that combines advanced indexing, machine learning, and personalized search to enhance academic research and document discovery.

Overview: Seedance 1.5 Pro Prompts offers tested image prompt recipes with example renders, aspect ratios, the model id (`seedance-1.5-pro`), and the cost per image. It provides a quickstart guide and details on pricing and usage. Key features: - 4 tested recipes with example renders, duration, and cost on the APIMart route. - Model: `seedance-1.5-pro`. - Effective price used: $0.0104 per second. - No subscription required; $1 minimum. - Pricing is reflected in the response. - Uses a non-blocking design for task completion. Tech stack: Python. Why it matters: Seedance 1.5 Pro Prompts provides a practical guide for developers looking to use Seedance 1.5 Pro, offering tested recipes and cost information, making it easier to integrate and manage the model without the need for a subscription.

### Overview: ARC-LLM-Lab is a project that showcases the performance of Qwen3.8-27B on Intel Arc GPUs, specifically the B60, B580, B570, B50, B70, and A770 models. This project aims to benchmark and optimize the decoding speed and reliability of the model, providing insights into the best practices for deploying Qwen3.8-27B on Intel Arc GPUs. ### Key Features: - **Qwen3.8-27B Model**: The model is optimized for high performance, decoding up to 53 tokens per second (tok/s) with a 64K context window. - **Intel Arc GPUs**: The project focuses on the B60, B580, B570, B50, B70, and A770 models, showcasing the best practices for deploying Qwen3.8-27B on these GPUs. - **Vulkan Implementation**: The project uses the Vulkan implementation of llama.cpp, which is optimized for performance and reliability. - **Tuning and Optimization**: The project includes detailed tuning logs, including DFlash2 draft model, Intel FA prefill kernel, and other optimization techniques. - **Benchmark Results**: The project provides benchmarks and data on the performance of the model, including decoding speeds, prompt reading speeds, and agent reliability. ### Tech Stack: - **Python**: The primary language used for the project. - **Vulkan**: The implementation of llama.cpp for

### Overview: 本项目专注于面向超内存模型的专家流式、可逆运行档与可复现实证，特别适用于Kimi K3这类权重规模远大于设备内存的超大规模MoE模型。通过按路由的超内存专家流式、93层完整档与68层裁剪档、共享专家常驻、固定等长四路逐行一致批处理、structural68+K3-DSpark精确投机解码等核心方法，实现性能和资源效率的优化。 ### Key features: - 按路由的超内存专家流式推理 - 93层完整档与68层裁剪档共用模型权重 - 共享专家常驻 - 固定等长四路逐行一致批处理 - structural68+K3-DSpark精确投机解码 ### Tech stack: Python ### Why it matters: 本项目公开实现方式、验证状态、实测对比与回退边界，旨在为超大规模MoE模型推理提供可复现实证的解决方案，通过优化推理流程和资源利用，提升推理性能和资源效率，助力模型在高性能计算平台上的高效部署和应用。

### Overview: My Little Office is a pixel-art office simulation where AI agents handle real coding tasks. The dashboard and agent runtime work together to clone repositories, write changes, run checks, and create pull requests on behalf of developers. ### Key Features: - **Dashboard**: The office UI where tasks are assigned, monitored, and approved. - **Agent Runtime**: Clones repositories, edits files, runs commands, commits, pushes, and opens pull requests. - **Linear Integration**: Integrates with Linear for project management and issue tracking. ### Tech Stack: Next.js on Cloudflare Workers for the dashboard, Node on the agent runtime for the agent. ### Why It Matters: This project offers a realistic simulation of a coding environment where AI agents handle tasks, providing a valuable learning tool and a practical way to understand how AI interacts with traditional development workflows.