Projects

### Overview: PipelineScheduler is a system designed to optimize workload distribution between servers and edge devices, ensuring maximum throughput and minimal latency. It addresses resource contention through spatiotemporal inference scheduling, mitigating co-location interference and improving resilience and efficiency in distributed inference pipelines. ### Key Features: - **Optimal Workload Distribution**: PipelineScheduler splits pipelines between servers and edge devices to maximize throughput and minimize latency. - **Local Optimization**: It applies local optimization of runtime parameters like inference batch size to reduce resource contention. - **Resource Management**: The system considers resource contention and uses inference spatiotemporal scheduling to mitigate adverse effects of co-location interference. ### Tech Stack: C++ ### Why It Matters: PipelineScheduler is essential for optimizing the performance of edge computing systems, particularly in scenarios with dynamic content and network instability. By efficiently distributing workloads and managing resource contention, it ensures that edge devices can handle high-throughput tasks effectively, leading to better overall system performance and reliability. ### Repository: Repository: tungngreen/PipelineScheduler GitHub Stars: 12 GitHub Description: PipelineScheduler optimizes workload distribution between servers and edge devices, setting optimal batch sizes to maximize throughput and minimize latency amid content dynamics and network instability. It also addresses resource contention with spatiotemporal inference scheduling to reduce co-location interference. ### Table of Contents: 1. **Overview** 2. **Implementation Architecture** - **Controller** - **Device Agent** - **Inference Container** - **Container Agent** - **Local Optimizations (FCPO)** -

Overview: Seedance 1.0 Pro Fast API is a low-latency, OpenAI-compatible endpoint providing model id, pricing, and parameters for generating videos and images. It offers a seamless API key setup, pay-as-you-go pricing, and detailed documentation for developers. Key features: - Supports 300+ models with a single Base-URL and `Authorization` header - Per-second billing for 480p, 720p, and 1080p resolutions - Real-time billing with each task's cost reported in the response - Asynchronous processing with `task_id` for easy tracking Tech stack: Python Why it matters: This API is ideal for developers looking for a low-latency, cost-effective solution for generating videos and images using OpenAI-compatible models. Repository: u6ybw3k/seedance-1-0-pro-fast-api-de Primary language: Python GitHub stars: 12 GitHub description: Seedance 1.0 Pro Fast API (`doubao-seedance-1-0-pro-fast`): model id, pricing, parameters and copy-paste examples for the OpenAI-compatible APIMart endpoint. README excerpt: Seedance 1.0 Pro Fast is the lower-latency Seedance 1.0 Pro route on APIMart, billed per generated second in 480p / 720p / 1080p. It offers a seamless API key setup, pay-as-you-go pricing, and detailed documentation for developers.

### Overview: NeuroLens is a premium, state-of-the-art Document Retrieval-Augmented Generation (RAG) system that combines an immersive, sci-fi-themed React frontend with a high-performance FastAPI backend. It is designed to provide sub-millisecond retrieval speeds and zero context starvation, making it ideal for both standalone and hybrid execution modes. ### Key Features: - **Fair-Share Round-Robin Retrieval (`searchBM25MultiDoc`)**: Solves single-document context starvation by fairly distributing retrieved context passages across all active documents. - **Dynamic Context Quotas**: Scales retrieved chunks dynamically based on library size. - **Cross-Document Stratified Summarization**: Retrieves opening abstracts, core body sections, and conclusions from every uploaded file. - **Executive Catalog Awareness**: Injects a structured catalog containing file names, page counts, chunk metrics, and opening synopses directly into the LLM system prompt for 100% inventory awareness. - **Ground Truth Override & History Isolation**: Filters internal system notifications from chat history to ensure re-added files are recognized with live ground truth priority. ### Tech Stack: - JavaScript - React 19 - Vite 8 - FastAPI - LangChain - Faiss - KaTeX - License: MIT ### Why It Matters: NeuroLens is a cutting-edge system that leverages the latest advancements in AI and document retrieval to provide unparalleled performance and functionality. Its dual-execution modes ensure high-speed retrieval and efficient use of resources, making it a valuable tool for developers

Overview: This project provides an API endpoint for Gemini Nano Banana 2.1, offering model id, pricing, and parameters for the OpenAI-compatible APIMart endpoint. Key features: - Model id and pricing information - Model id `gemini-nano-banana-2.1` - Pricing is per delivered image, by resolution - Includes 300+ models, including image and reference-image editing - Supports non-subscription pricing - Returns the exact amount in the `cost` field of completed tasks - Non-blocking design, only requires `task_id` to poll the task status Tech stack: Python Why it matters: This API provides a convenient way to access a model that is part of Google's image model catalog on APIMart, with features like non-subscription pricing and non-blocking design, making it suitable for developers looking to integrate this model into their projects without the need for a subscription.

### Overview: This repository contains tested image prompt recipes for the Gemini Nano Banana 2.1 model, including example renders, aspect ratios, and the cost per image on the APIMart route. The model id is `gemini-nano-banana-2.1`, and it offers a cost of $0.0137 per image on the APIMart API. ### Key Features: - **Model ID:** `gemini-nano-banana-2.1` - **Example Renders:** Eight recipes with detailed instructions on materials, lighting, framing, and settings. - **Aspect Ratios:** 1:1, 4:3, 85mm, 24mm, macro, overhead flat lay. - **Cost:** $0.0137 per image, with a minimum cost of $1 for unlimited usage. - **Usage Tracking:** The cost is reflected in the response, and failed tasks are refunded. ### Tech Stack: The repository uses Python for its implementation. ### Why It Matters: - **Cost Efficiency:** The model is priced at $0.0137 per image, making it cost-effective for various applications. - **Versatility:** The model supports multiple formats and can be used for both image and video prompts. - **Compliance:** It is compatible with OpenAI's API, ensuring seamless integration with other AI systems. The recipes provided in this repository are designed to help developers create high-quality images efficiently and effectively.

### Overview: MLOps Ecosystem 2024-2025 is a comprehensive repository designed for developers looking to build production-grade machine learning systems. This repository features the latest tools and frameworks from 2024-2025, ensuring your projects are battle-tested and production-ready. ### Key Features: - 🤖 **LLMOps & GenAI**: Comprehensive tools for LLM operations and GenAI. - 🎨 **Beautiful UI**: Modern interfaces for ML workflows. - 📈 **Complete Lifecycle**: Every stage of ML development covered. - 🌐 **Community-Driven**: Open-source and actively maintained. - 🚀 **Fast Development**: pytest-xdist parallel testing (3.3x faster). ### Tech Stack: - Python ### Why It Matters: - 🔥 **Always Up-to-Date**: Latest tools and frameworks from 2024-2025. - ⚡ **Production-Ready**: Battle-tested tools used by industry leaders. - 🤖 **LLMOps Ready**: Comprehensive GenAI and LLM operations toolkit. - 🎨 **Beautiful UI**: Modern interfaces for ML workflows. - 📈 **Complete Lifecycle**: Every stage of ML development covered. - 🌐 **Community-Driven**: Open-source and actively maintained. - 🚀 **Fast Development**: pytest-xdist parallel testing (3.3x faster). ### Repository: - **Repository**: umitkacar/awesome-mlops - **GitHub Stars**: 12 - **

Overview: This project is a Nano Banana 2 API (`gemini-3.1-flash-image-preview`) that offers fast AI image generation with native 4K output and editing capabilities. It supports OpenAI-compatible endpoint and provides detailed pricing and parameters for developers. Key features: - Native 4K output - Supports 300+ models - Per-image pricing: $0.015 per image - $1 minimum charge, pay-as-you-go - Real-time cost reporting - Asynchronous request handling Tech stack: Python Why it matters: This API simplifies access to Nano Banana 2 (gemini-3.1-flash-image-preview) models, allowing developers to use the same base URL and `Authorization` header for multiple models without changes. It offers a cost-effective and efficient way to generate images with minimal setup required. Repository: utp00/gemini-3-1-flash-image-preview-api-cn Primary language: Python GitHub stars: 8 GitHub description: 香蕉 — 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 project involves training a Generative Adversarial Network (GAN) to generate images from textual descriptions of flowers. Key features: - Training a GAN to synthesize images from textual descriptions - Utilizing the Oxford 102 Flower Dataset for training - Processing image descriptions into character-level embeddings using GloVe - Generating images of various flower categories Tech stack: Jupyter Notebook Why it matters: This project demonstrates the capability of using GANs to generate realistic images from textual descriptions, showcasing the potential of deep learning in image synthesis.

Overview: FLUX.2 Flex Prompts offers tested image prompt recipes with example renders, aspect ratios, and the model ID (`flux-2-flex`) along with the cost per image. This set includes 8 recipes that provide a clear and effective way to create professional-looking images. Key features: - 8 tested recipes with example renders, aspect ratios, and the model ID (`flux-2-flex`). - Cost per image: $0.04, estimated cost for the entire set: $0.32. - Minimal cost, no need for monthly billing. - Asynchronous processing, returns cost upon completion. Tech stack: Python. Why it matters: This tool simplifies the process of creating professional-looking images by providing clear and effective prompts, reducing guesswork and ensuring consistent results. Repository: v501hfo6zxyf4/flux-2-flex-prompts-ru Primary language: Python GitHub stars: 8

### Overview: Seedance 1.0 Pro Fast API is an OpenAI-compatible endpoint designed for real-time video generation. It offers a low-latency solution with a per-second billing model, perfect for immediate applications. The API is accessible via `https://api.apimart.ai/v1` and supports various resolutions and parameters for generating videos and images. ### Key Features: - **Model ID and Routes**: The model ID `doubao-seedance-1-0-pro-fast` is used for both video and image generation, with a billing model based on generated seconds. - **Pricing**: The pricing is per generated second, with a resolution-based tier system. - **API Key**: A single key can be used to access the entire API, including 300+ models. - **Real-time Pricing**: The pricing is updated in real-time, and tasks report the exact cost. - **Async Design**: Tasks can be submitted, and the status can be checked using `GET /v1/tasks/{id}`. ### Tech Stack: Python ### Why It Matters: - **Key Access**: A single key can be used to access the entire API, making it highly efficient. - **Cost-Effective**: No need for prepayment, and the cost is reported immediately. - **Real-Time Updates**: Pricing is updated in real-time, ensuring accurate cost tracking. - **Ease of Use**: The API is designed to be easy to use, with a simple request structure and straightforward status updates. ### Repository: v87tcvt6l/seedance-

Overview: Seedance 1.5 Pro API is a comprehensive OpenAI-compatible API endpoint designed for developers, providing model IDs, pricing details, and parameters for seamless integration into their projects. The API supports various resolutions and image/video formats, ensuring high-quality output. Key features: - Model ID and pricing details - Parameters for prompt description, duration, resolution, and output framing - Asynchronous operation with task tracking and cost reporting - Compatible with ByteDance's Seedance 1.5 Pro model Tech stack: Python Why it matters: Seedance 1.5 Pro API simplifies the integration of high-quality AI video generation into projects, offering flexible pricing and easy task tracking, making it ideal for developers looking to enhance their applications with advanced AI capabilities without the complexity of managing a full API subscription. Repository: vi21di4j/doubao-seedance-1-5-pro-api-id GitHub stars: 12

### Overview: Grok Imagine Image 2.0 API is a high-fidelity image generation service that offers a single key to access a catalog of models provided by OpenAI-compatible endpoints. The API is priced per image by resolution and quality, with a minimum charge of $0.01 per image. It supports both low and medium resolutions, and offers a 1K@low, 1K@medium, 2K@low, and 2K@medium pricing tier. ### Key Features: - **One Key, Entire Catalog**: Access to a single API key that allows you to use the same base URL and `Authorization` header to reach Grok Imagine Image 2.0 and over 300 other image, video, and language models. - **Pay as You Go**: No subscription required; pay only for the calls you make. There is no free quota to burn through, so the price is based on the actual amount charged per call. - **Charge Back to the Response**: Every call reports the amount billed, so you can track your spend in real-time. - **Async by Design**: Submit the `task_id`, and poll `GET /v1/tasks/{id}` to know when the task is complete. Batching and retries are handled as ordinary queue work, not a bespoke integration. ### Tech Stack: - **Python**: The primary programming language used to develop the API. ### Why It Matters: - **One Key, Entire Catalog**: The ability to use the same API key for a wide range of models, making it easy to

Overview: This Ansible repository provides a K3s cluster with four NVIDIA DGX Spark nodes for distributed language model inference using ARM64 architecture. Key features: - K3s Master, Control-Plane, Frontend Services, and NUT UPS monitoring - SGLang Head and Worker nodes for SGLang (a language model) inference - DGX Spark nodes with NVIDIA GPUs for high-performance computing Tech stack: Python Why it matters: This setup is ideal for deploying a distributed language model inference system efficiently and reliably, leveraging the power of NVIDIA DGX Spark nodes and K3s for a seamless and scalable cluster. Repository: vroomfondel/dgxarley GitHub stars: 12 GitHub description: Ansible playbooks for a 4-node K3s cluster with NVIDIA DGX Spark nodes for distributed LLM inference.

### Overview: Nano Banana Pro Prompts — 8 Tested Recipes With Outputs Nano Banana Pro Prompts offers tested image prompt recipes with example renders, aspect ratios, the model id (`gemini-3-pro-image-preview`), and the cost per image. This repository provides Nano Banana Pro (`gemini-3-pro-image-preview`) and other OpenAI-compatible endpoint models for $0.03 per image, with a minimum charge of $1 and real-time billing. ### Key Features: - **Model ID**: `gemini-3-pro-image-preview` - **Recipes**: 8 tested recipes with detailed example renders, aspect ratios, and costs. - **Effective Price**: $0.03 per image, estimated cost of $0.24 for the set. ### Tech Stack: Python ### Why It Matters: Nano Banana Pro Prompts offers a concise and effective way to create professional-grade image prompts, ensuring that each call results in a high-quality output without unnecessary complexity. The structured prompts and clear structure make it easy to replicate and use in various applications, enhancing efficiency and effectiveness. ### Repository: w-44vba/gemini-3-pro-image-preview-prompts-fr ### Primary Language: Python ### GitHub Stars: 12 ### GitHub Description: Nano Banana Pro Prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`gemini-3-pro-image-preview`) and the cost per image. ### Quickstart: Get an API key: [Get an API key](https://go.apimart.ai/k-cbb

Overview: Nano Banana 2 Lite is a cheaper alternative to the Gemini 3.1 Flash Lite Image route, priced per image. It offers a single endpoint compatible with OpenAI, allowing for up to 300 models including Nano Banana 2 Lite. Key features: - A single endpoint compatible with OpenAI - Pricing per image - Assynchronous by design, allowing for task tracking and retries Tech stack: Python Why it matters: Nano Banana 2 Lite is a more cost-effective option for image generation, providing a single endpoint that can be used with OpenAI-compatible models without the need for a monthly subscription.

Overview: This project, while functional, is no longer being maintained and has been replaced by the Weaviate TypeScript client. Key features: None specified. Tech stack: TypeScript Why it matters: This library is no longer being maintained and should be replaced by the Weaviate TypeScript client for the most up-to-date functionality and support.

### Overview: Nano Banana Pro API (`gemini-3-pro-image-preview`) is an OpenAI-compatible API that allows developers to generate and edit 4K images with multi-step planning. It offers a cost-effective alternative to traditional subscription-based models, with a simple API key and no need for monthly billing. ### Key Features: - **Model ID and Routes**: Supports the `gemini-3-pro-image-preview` model, with pricing based on the number of generated images and resolution. - **Pricing**: Starts at $0.03 per image (4K $0.04), with a default rate of $0.03. - **Request Parameters**: Accepts fields such as `model`, `prompt`, `size`, `resolution`, `n`, and `image_urls`. - **Quickstart**: Provides a simple command-line example using `curl` and a Python script for easy integration. ### Tech Stack: - Python ### Why It Matters: - **Cost Efficiency**: Offers a cost-effective alternative to traditional subscription-based models. - **Simple API Key**: Requires only a single API key for access. - **No Monthly Billing**: No need for monthly subscription plans, making it ideal for developers on a budget. - **Multi-Step Planning**: Supports multi-step image generation and editing, allowing for detailed customization. ### GitHub Stars: 10 ### Primary Language: Python ### GitHub Description: Nano Banana Pro API (`gemini-3-pro-image-preview`) is an OpenAI-compatible API designed for generating and editing 4K images. It provides a cost

Overview: Mood is an AI-driven journaling web app designed to track your sentiments and moods over time, offering advanced sentiment analysis and question-based similarity searches. Key features: - Adding a Journal Entry - Sentiment Analysis - Sentiment History Chart - Question-Based Search Tech stack: Next.js 13 + TypeScript, Prisma ORM + MySQL, OpenAI API, TailwindCSS, Vitest, Clerk Auth, Langchain for In-Memory Vector DB Why it matters: Mood provides a personalized and insightful way to track your emotions, offering valuable insights into your emotional journey and trends.

Overview: Nano Banana API, also known as `gemini-2.5-flash-image-preview`, is a fast, OpenAI-compatible image generation service that offers scalable pricing and flexible parameters. It supports a wide range of image editing tasks and integrates seamlessly with the APIMart endpoint. Key features: - Supports multiple image, video, and language models - Offers a single endpoint for OpenAI compatibility - Provides a per-image pricing model - Supports dynamic scaling based on usage - Asynchronous task processing Tech stack: Python Why it matters: Nano Banana API simplifies image generation by providing a single, scalable endpoint that integrates with popular pricing models, making it ideal for developers looking to streamline their workflows and reduce complexity. Repository: ycx8l0/gemini-2-5-flash-image-preview-api-pt Primary language: Python GitHub stars: 10 GitHub description: Nano Banana API (`gemini-2.5-flash-image-preview`): model id, pricing, parameters, and copy-paste examples for the OpenAI-compatible APIMart endpoint.

### Overview: Nano Banana Prompts: 8 Tested Recipes With Outputs ### Key features: - **Model:** `gemini-2.5-flash-image-preview` - **Recipes:** 8 (machine-readable in `data/prompts.json`, table in `PROMPTS.md`) - **Effective price:** $0.0125 per image ### Tech stack: Python ### Why it matters: Nano Banana Prompts offers a cost-effective way to generate high-quality image prompts, making it ideal for developers looking to streamline their workflow without the need for a monthly subscription.