Projects

Overview: RAG-Gemini is an implementation of the Retrieval-Augmented Generation model in TypeScript, utilizing the Gemini LLM. Key features: - Supports TypeScript development - Integrates with Gemini LLM for advanced language generation - Provides enhanced text generation capabilities through retrieval-based augmentation Tech stack: TypeScript Why it matters: This project showcases the integration of cutting-edge language models with modern programming languages, offering developers a powerful tool for text generation tasks. Repository: kanugurajesh/RAG-Gemini GitHub stars: 13

Overview: This repository contains tested image prompt recipes for Seedream 5.0 Lite, including 8 recipes with example renders, aspect ratios, the model ID (`seedream-5-0-lite`), and the cost per image. Key features: - 8 Seedream 5.0 Lite prompts - Example renders, aspect ratios, and model ID - Cost per image Tech stack: Python Why it matters: This repository provides practical image prompt recipes for Seedream 5.0 Lite, offering a straightforward way to generate high-quality images without the need for complex integration or prepayment plans.

### Overview: This repository contains tested Veo 3.1 prompts with example renders, aspect ratios, the model id (`veo3.1-fast-ext`), and the cost per image. The prompts are designed to help developers create high-quality images efficiently. ### Key features: - **Model ID:** `veo3.1-fast-ext` - **Render Examples:** 8 different recipes with their corresponding durations and costs. - **Cost per Image:** $0.064 per second, with a total estimated cost of $0.256 for the set. ### Tech Stack: Python ### Why it Matters: The repository provides a practical guide for developers to create effective Veo 3.1 prompts, ensuring high-quality images without the need for expensive subscription models.

Overview: MiniMax Hailuo 2.3 API is a comprehensive endpoint for OpenAI-compatible API Mart, providing text-to-video and image-to-video functionalities with detailed pricing and parameters. Key features: - Supports OpenAI-compatible endpoint - Offers text-to-video and image-to-video capabilities - Supports multiple resolutions and aspect ratios - Provides detailed model id, pricing, and parameters - Supports API key-based access and real-time monitoring Tech stack: Python Why it matters: This API is crucial for developers looking to integrate cutting-edge AI video generation into their projects, offering high-quality results with flexible pricing and easy integration. Repository: kdnde24c3s5g/hailuo-2-3-api-id Primary language: Python GitHub stars: 13 GitHub description: MiniMax Hailuo 2.3 API (`MiniMax-Hailuo-2.3`): model id, pricing, parameters and copy-paste examples for the OpenAI-compatible APIMart endpoint.

Overview: Grok Imagine 1.5 API (`grok-imagine-1.5-apimart`) is a model id priced per image with edits included, accessible via an OpenAI-compatible endpoint. It offers a single key for entire access to 300+ models, with a $1 minimum charge for each task and real-time pricing. Key features: - A single key for entire access to Grok Imagine 1.5 - $1 minimum charge per task, with real-time pricing - Supports 300+ models, including image, video, and language generation - Charges based on the actual cost of the response - Supports both default and custom model parameters - Supports image URLs and base64 data URLs as references Tech stack: Python Why it matters: This API provides a cost-effective way to generate images and videos using Grok Imagine 1.5, with real-time pricing and easy integration through a single key. Repository: ke6yg044xwihmo/grok-imagine-1-5-api-cn Primary language: Python GitHub stars: 13

Overview: Dive into the "Reinforcement Learning from Human Feedback" course, taught by Nikita Namjoshi, to align Large Language Models (LLMs) with human values and preferences using Reinforcement Learning from Human Feedback (RLHF). Key features: - Conceptual Understanding of RLHF - Supervised Fine Tuning - RLHF Training with Datasets - Fine-Tuning with Google Cloud Pipeline Components Library - Model Evaluation Tech stack: Jupyter Notebook Why it matters: This course provides a practical guide on how to align LLMs with human values using RLHF, making it essential for developers interested in the intersection of AI and ethics.

Overview: KosmosG is a model implementation based on "KOSMOS-G: Generating Images in Context with Multimodal Large Language Models". It aims to enhance the understanding and generation of images using multimodal large language models. Key features: - Multi-modality understanding - Integration of text and image inputs - Implementation in Python using PyTorch Tech stack: Python, PyTorch Why it matters: This model could significantly improve the ability to generate images that are contextually relevant and coherent, leveraging the strengths of large language models to enhance their multimodal capabilities. Repository: kyegomez/KosmosG Primary language: Python GitHub stars: 13 GitHub description: My implementation of the model KosmosG from "KOSMOS-G: Generating Images in Context with Multimodal Large Language Models" README excerpt: [](https://discord.gg/qUtxnK2NMf)

### Overview: A production-grade multi-agent system leveraging ReAct orchestration, MCP, A2A, and LMMs for hyper-local property valuation with Evals. ### Key Features: - Multi-Agent Orchestration (LangGraph) - MCP and A2A Integration - Advanced RAG & Vector Search - Large Multimodal Models (LMM) in Production - Classical ML Meets Agentic AI ### Tech Stack: PHP ### Why It Matters: This project demonstrates a cutting-edge implementation of modern AI and ML, providing a production-grade, multi-agent system for complex reasoning, valuation, and analysis in real estate.

### Overview: Qwen-Image 3.0 Prompts: 8 Tested Recipes With Outputs ### Key Features: - **Model ID:** `qwen-image-3.0` - **Recipes:** 8 (machine-readable in `data/prompts.json`, table in `PROMPTS.md`) - **Effective Price:** $0.0206 per image (estimated cost of the whole set: $0.1646) ### Tech Stack: - Python ### Why It Matters: - **Key:** Same Base-URL `https://api.apimart.ai/v1` and same `Authorization`-Header for Qwen-Image 3.0 - **Billing:** No Abo or Prepaid Tarif, pay-as-you-go - **Cost:** $0.0206 per image - **Asynchronous:** Absendungen, `task_id` entgegennehmen, `GET /v1/tasks/{id}` pollen — Batching and Retries are normal Queue-Arbeit, no special integration ### What Separates a Usable Recipe from a Wasted Call: - **Rule 1:** Name the material, not the adjective ("Frosted glass" vs "beautiful") - **Rule 2:** Set the light explicitly ("Single softbox from the left" vs "golden hour backlight") - **Rule 3:** Pin the framing ("85mm", "24mm", "macro", "overhead flat lay") decides whether the output drops into a layout unchanged ### Prompt Structure

### Overview: GAN-Tree: An Incrementally Learned Hierarchical Generative Framework for Multi-Modal Data Distributions, published in ICCV 2019. This repository contains code for the paper. The overall architecture is depicted in the figure, with additional details provided in the paper. ### Key features: - GAN-Tree: An Incrementally Learned Hierarchical Generative Framework - Multi-Modal Data Distributions - Code for the paper ### Tech stack: Python ### Why it matters: The GAN-Tree framework is designed to handle multi-modal data distributions in a hierarchical manner, offering a scalable and efficient approach for generating new data points. This is particularly useful in scenarios where data is distributed across multiple modalities, such as images and text, and requires a generative model that can learn from these distributions incrementally.

### Overview: The Super Rapid Annotator project is designed to streamline video annotation tasks by integrating advanced multimodal vision and language models with spatiotemporal analysis. This tool addresses the inefficiencies of traditional annotation methods, particularly in handling multiple videos simultaneously, by leveraging state-of-the-art models and efficient deployment strategies. ### Key Features: - **Multimodal Vision and Language Model Integration**: Seamlessly combines images, videos, and text inputs to generate high-quality annotations. - **Spatiotemporal Analysis**: Analyzes video content in terms of spatial and temporal relationships to identify entities and their interactions. - **Efficient Deployment**: Utilizes lightweight models and efficient deployment strategies to ensure quick and accurate annotation. ### Tech Stack: - **Python**: The primary programming language used for development. - **PyTorch**: Framework for building and training deep learning models. - **Hugging Face**: Hosting and deploying the model through Gradio. - **Pydantic**: For defining and validating data models. - **OpenCV**: For image processing and video analysis. - **FFmpeg**: For video processing and manipulation. - **FastAPI**: For building the web application interface. ### Why It Matters: The Super Rapid Annotator project addresses the limitations of traditional annotation methods by providing a more efficient and accurate way to annotate videos. This tool is particularly useful in scenarios where multiple videos need to be annotated simultaneously, making it invaluable for researchers, educators, and anyone involved in video analysis tasks.

Overview: This repository provides a complete deployment kit for running the Qwen-Image-2.1 Uncensored model in ComfyUI, including workflow patches, a 14-second lossless Q8_0 quantization repair tool, and a GGUF-based ComfyUI-GGUF architecture patch. The model weights are not included, and the repository is designed to be used with an 8 GB VRAM GPU. Key features: - Workflow patches - 14-second lossless Q8_0 quantization repair tool - GGUF-based ComfyUI-GGUF architecture patch Tech stack: Python Why it matters: This repository offers a seamless solution for deploying the Qwen-Image-2.1 Uncensored model in ComfyUI, addressing common issues such as missing GGUF metadata, duplicate GGUF files, and incorrect Q8_0 quantization. The 14-second lossless repair tool ensures high-quality outputs without additional downloads, making it ideal for users with limited bandwidth or storage constraints. Repository: mark2008chen-sudo/qwen-image2.1-uncensored-comfyui GitHub stars: 13

Overview: This repository contains open-source AI demos that can be understood, run, and remixed. It offers small, reproducible builds for vision, multimodal models, local or hosted inference, and more. Key features: - Vision: Draw object boxes with a vision model - Documents: Read form checkboxes - Model(s): Multiple models verified with the same code Tech stack: Python Why it matters: This repository strips down each demo to its essential core, making it easy to learn, run, and modify. It is not a production-level framework, but a simple, readable build that can be adapted for various applications.

Overview: This project focuses on distance computation for vector databases, aiming to provide an effective and general solution for approximate nearest neighbor search. Key features: - Distance computation for vector databases - Implementation of ADsampling and FINGER algorithms - Support for both SIMD and non-SIMD acceleration - Utilizes C++ and Python environments Tech stack: C++ Why it matters: This project offers a practical solution for distance computation in vector databases, enhancing the efficiency and accuracy of approximate nearest neighbor search, which is crucial in various applications such as recommendation systems and image retrieval.

Overview: Finetune Web UI is a user-friendly interface designed for training and deploying pre-trained machine learning models, specifically focusing on fine-tuning Llama and GPT models. Key features: - Easy dataset upload - Custom model fine-tuning - Hyperparameter tuning - Real-time monitoring - Simplified deployment - Model management - Clean and intuitive interface - Open source - Customizable - Cross-platform - Scalable Tech stack: Python Why it matters: Finetune Web UI simplifies the process of fine-tuning pre-trained models, making it accessible for data scientists and engineers without requiring deep technical knowledge. GitHub stars: 13 Repository: muhammad-fiaz/finetune-web-ui

Overview: Seedance 2.5 API is an OpenAI-compatible endpoint for generating video content with cinematic motion, offering per-second billing by resolution and a catalog of over 300 models including text-to-video and image-to-video capabilities. Key features: - Text-to-video / image-to-video generation with per-second billing by resolution - A single API key for accessing multiple models, including 300+ image, video, and language models - Pricing based on usage, starting at $0.0961 per second for 480P, with $1 per second for 720P and 1080P - Assynchronous processing, returning `task_id` for monitoring progress and retries - Model-specific tier pricing detailed in the model page Tech stack: Python Why it matters: Seedance 2.5 API streamlines the process of generating high-quality video content, offering flexible pricing and a comprehensive model catalog, making it ideal for developers looking to integrate AI-driven video generation into their applications without the need for a dedicated API key. Repository: n3yboo/seedance-2-5-api-pt GitHub stars: 9 GitHub description: Dreamina — Seedance 2.5 API (`seedance-2.5`): model id, pricing, parameters and copy-paste examples for the OpenAI-compatible APIMart endpoint.

### Overview: This repository focuses on evaluating the feasibility of detecting and fixing bugs using prompt engineering and fine-tuning techniques in large language models (LLMs). The project utilizes a custom dataset and innovative methods to enhance LLMs' capabilities in understanding and correcting code errors. ### Key Features: - Custom dataset for bug detection and fixing. - Model description: [Unsloth Meta-Llama 3.1-8B]. - Prompt engineering and fine-tuning techniques. - User-friendly interface for evaluation and performance metrics. ### Tech Stack: Jupyter Notebook. ### Why it Matters: This project aims to demonstrate the effectiveness of prompt engineering and fine-tuning in improving the performance of large language models for bug detection and fixing, highlighting the potential of these techniques in enhancing software development efficiency and quality.

**Overview:** This repository provides the official resources for the **CulturalPangea** model and the **CulturalGround** dataset, which contains 30 million high-quality VQA pairs from 42 countries and 39 languages, curated from Wikidata. The dataset is specifically designed to address the common issue of MLLMs misinterpreting long-tail cultural entities by grounding the model in diverse cultural knowledge. **Key features:** - **CulturalGround Dataset**: Contains 30 million high-quality VQA pairs from 42 countries and 39 languages, curated from Wikidata. - **CulturalPangea Model**: An open-source, multilingual, multimodal large language model fine-tuned to understand culturally significant entities from around the world. **Tech stack:** Python **Why it matters:** The **CulturalPangea** model and dataset are designed to address the common issue of MLLMs misinterpreting long-tail cultural entities by grounding the model in diverse cultural knowledge, making it a valuable resource for researchers and practitioners in the field of cultural understanding and language modeling. **GitHub stars:** 13 **Primary language:** Python **Repository:** neulab/CulturalGround

Overview: This project allows users to generate illustrations from novels by highlighting specific sentences using Stable Diffusion. The app is designed for self-hosting and can be integrated into Google Colab and publicly used. Key features: - Live novel: A backend that leverages Stable Diffusion to create fantasy and beautiful images from novel text prompts. - Self-host: Smooth installation for self-hosting, Google Colab free tier, and public use. - Integration: Ready for integration and production, built on top of DocArray data structure, and supports gRPC/HTTP/websockets and TLS. - Installation: Easy installation via Docker, including pre-built image and Docker commands for self-hosting and customization. Tech stack: Python. Why it matters: This tool can help authors promote their work and readers visualize scenes without needing to create illustrations themselves. Repository: nhtlongcs/live-novel GitHub stars: 13 GitHub description: Self-host application can generate illustration from a novel by highlighting certain sentences.

Overview: This project is an Elixir library for building RAG (Retrieval Augmented Generation) systems, featuring multi-LLM support, modular architecture, and advanced features like GraphRAG and advanced chunking. Key features: - Multi-LLM Provider Support: Gemini, Claude, Codex (OpenAI-compatible), and Ollama - Smart Routing: Fallback, round-robin, and specialist routing strategies - Streaming Responses: Real-time streaming for supported providers - Modular RAG Architecture: Pluggable retrieval, vector storage, and reranking - GraphRAG Support: Entity extraction, knowledge graph storage, and community detection - Advanced Chunking: Behavior-based, byte positions, character-based, sentence-based, paragraph-based, recursive, semantic, and format-aware - Agent Framework: Tool-using agents with session memory and built-in tools Tech stack: Elixir Why it matters: This library provides a comprehensive solution for building RAG systems, addressing common issues with traditional language models and enhancing their capabilities with external knowledge sources and advanced chunking techniques. Repository: nshkrdotcom/rag_ex GitHub stars: 13 GitHub description: Elixir RAG library with multi-LLM routing (Gemini, Claude, OpenAI, Ollama), GraphRAG, knowledge graphs, modular retrievers/rerankers, composable pipelines, pgvector integration, advanced chunking, and tool-using agents. Fork of bitcrowd/rag. ``` The README excerpt provides an overview of the library, including its