Projects

Overview: Wan 2.7 Image Pro API (`wan2.7-image-pro`) is a powerful tool for creating stunning visual outputs using Alibaba's Wan engine. It offers fast generation with minimal cost, making it ideal for creators across the Wan community. Key features: - Supports fast Wan 2.7 Image Pro generation - Pricing is per delivered image, with a cost of $0.0544 per image - Supports multiple image generation per task - List price is $0.068, effective price is $0.0544 - Supports customizable prompts and image references Tech stack: Python Why it matters: This API streamlines the process of creating visual content, making it more accessible and cost-effective for creators. Repository: jsgotg/wan-2-7-image-pro-api-ru GitHub stars: 11 GitHub description: Ван — Wan 2.7 Image Pro API (`wan2.7-image-pro`): model id, pricing, parameters and copy-paste examples for the OpenAI-compatible APIMart endpoint. Primary language: Python

### Overview: 通義万相 — Wan 2.5 prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`wan-2.5`), and the cost per image. This repository provides 8 tested recipes, each with an example render, duration, and the cost on the APIMart route. ### Key features: - **Model ID (`wan-2.5`)**: Access to 300+ models including Wan 2.5. - **Cost per second**: $0.0336 for each second of rendering. - **Subsidiary pricing**: No need for a monthly plan; use $1 to start, and then charge as needed. - **Pricing transparency**: All costs are reported in the response, and failed tasks are refunded. ### Tech stack: Python ### Why it matters: - **Key access**: A single API key allows access to the entire Wan 2.5 catalog. - **Cost efficiency**: Start with $1 and scale up as needed, eliminating the need for a monthly subscription. - **Pricing clarity**: The cost is clearly reported, and failed tasks are refunded, ensuring transparency and reliability. - **Non-blocking design**: Simple and efficient, using non-blocking polling for tasks, making it suitable for real-time applications. ### Repository: - **Repository**: k1glyz2132in/wan-2-5-prompts-ja - **Stars**: 11 - **Description**: This repository contains 8 tested Wan 2.5

Overview: This repository contains tested video prompt recipes for the FLUX 3 Video model, including example renders, aspect ratios, and the associated cost per image. Key features: - 4 tested video prompt recipes - Example renders, durations, and costs on the APIMart route - Prompt structure cheat-sheet for better rendering Tech stack: Python Why it matters: These recipes help developers create high-quality video prompts with specific materials and lighting effects, ensuring accurate and efficient rendering. Repository: k1hrgb5nv5q9un/flux-3-prompts-fr Primary language: Python GitHub stars: 11 GitHub description: FLUX 3 Video prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`flux-3-video`) and the cost per image.

Overview: MiniMax H3 API (`MiniMax-H3-Context-IR`) is a powerful AI video model offering native 2K output, high motion coherence, and professional-grade performance. It provides a single key for access across multiple models and resolutions, with flexible pricing and real-time billing. Key features: - Single key for access to all MiniMax H3 models and resolutions - Native 2K output and high motion coherence - Professional-grade performance - Flexible pricing with real-time billing Tech stack: Python Why it matters: This API is ideal for developers looking to integrate AI-driven video generation into their projects, offering a seamless and efficient solution with minimal setup required. Repository: k3dzxjxo53h96l/minimax-h3-context-ir-api-id GitHub stars: 11

### Overview: Qwen Image 3.0 Prompts: 8 Tested Recipes with Outputs ### Key features: - **Model:** `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: $0.1648) ### Tech stack: Python ### Why it matters: - **Key 1st Access:** Same `https://api.apimart.ai/v1` base URL and `Authorization` header to Qwen Image 3.0 and 300+ models. - **No Subscriptions:** No need for subscription or prepaid plans. - **Pricing Details:** All costs are reported in the response. - **Non-Synchronous Design:** Poll `GET /v1/tasks/{id}` for batch processing and retries. ### Repository: k8onp/qwen-image-3-prompts-ja ### Primary language: Python ### GitHub stars: 11

Overview: This repository contains 8 tested recipes for the HappyHorse 1.1 model, including examples of renders, aspect ratios, and the associated costs on the APIMart image API. Key features: - 8 tested recipes for the HappyHorse 1.1 model - Example renders, aspect ratios, and call costs on the APIMart image API - $0.01 per image pricing with a $0.08 total cost for the set Tech stack: Python Why it matters: This repository provides a concise guide for developers to use the HappyHorse 1.1 model effectively, offering a cost-effective solution with detailed prompts and pricing information.

Overview: This repository contains a project aimed at experimenting with smaller language models that can run on commodity hardware, specifically Raspberry Pi, as a personal assistant. The project uses Python and various dependencies to create a small, lightweight assistant that can be used on less powerful hardware. Key features: - Small language model - Runs on Raspberry Pi - Personal assistant application Tech stack: Python, Flask, PyTorch, TensorFlow, FFMPEG Why it matters: This project demonstrates the feasibility of using smaller language models on less powerful hardware, making it an interesting experiment for developers interested in AI and hardware limitations. Repository: kaalen/tiny-assistant GitHub stars: 11 GitHub description: Experimenting with smaller LLMs that can run on commodity hardware like Raspberry Pi as a personal assistant.

Overview: A script that converts text into TikTok style videos using advanced AI and video editing tools. Key features: - Generates TikTok style videos - Uses FFmpeg, moviepy, ChatGPT, and SDXL API - Requires AWS Polly for voiceover and moviepy for subtitles - Generates story prompts with ChatGPT Tech stack: Python Why it matters: This script leverages cutting-edge AI and video editing technologies to create engaging TikTok-style videos in a matter of minutes, making it a valuable tool for content creators and marketers looking to quickly produce short-form videos.

### Overview: This repository contains curated Docker Compose configurations for popular services, simplifying development and deployment workflows. ### Key features: - **Postgres with pgAdmin**: PostgreSQL database with pgAdmin for management. - **Kafka with Kafka UI**: Single-node Kafka (KRaft, no Zookeeper) with a management UI. - **MinIO**: S3-compatible object storage. - **RustFS**: S3-compatible object storage written in Rust. - **SeaweedFS**: Scalable distributed storage with an S3 gateway and filer. - **Garage**: Lightweight, self-hostable S3-compatible store. - **Redis with RedisInsight**: In-memory data store with a management UI. - **Valkey**: Drop-in Redis replacement managed by the Linux Foundation. - **RabbitMQ**: Reliable messaging between distributed systems. - **Metabase**: Simple self-service BI on PostgreSQL, with a sample database to explore. - **Superset**: BI platform with SQL Lab and dashboards on PostgreSQL + Valkey, with a sample database to explore. - **Qdrant Vector DB**: Optimized for storing and searching high-dimensional vectors. - **Milvus Vector DB**: Open-source vector database for similarity search. - **ElasticSearch with Kibana**: Distributed search engine with Kibana for visualization. - **Grafana with Prometheus**: Metric collection, storage, and visualization. - **MongoDB with Mongo-Express**: MongoDB paired with a web-based management interface. - **Apache Airflow**: Platform to programmatically author, schedule,

Overview: This is a simple web app that converts typed text into natural handwriting, designed to make writing assignments easier for students. Key features: - Converts typed text into handwriting-like images - Supports multiple fonts, pen colors, and document customization - Allows downloading as image or PDF - Supports equation symbols without leaving the tab Tech stack: HTML Why it matters: This app provides a convenient and aesthetically pleasing alternative to traditional handwriting, making it useful for students who struggle with writing assignments.

### Overview: Typeahead KMP is a high-performance, asynchronous, and lock-free in-memory fuzzy search engine designed specifically for Kotlin Multiplatform. ### Key Features: - **Multi-Word Fuzzy Search:** Queries with one word or multiple words, each token independently vectorized and scored using a harmonic mean with exponential proximity decay for adjacent token matches. - **Zero Network Latency:** Runs entirely on the Edge (the user's device memory), making it perfect for instant UI feedback. - **Typo & Transposition Tolerant:** N-gram, character bag, and positional embedding techniques catch mistakes that break standard prefix searches, even single-character insertions or deletions across multiple query words. - **100% Thread-Safe & Lock-Free:** Built on `MutableStateFlow` and `PersistentMap` (HAMT), allowing thousands of concurrent reads and writes without blocking threads. - **Cold-Start Elimination:** Blazing-fast export and import of pre-computed vector states via streaming JSON serialization. - **Cross-Platform:** Write once, run anywhere across JVM, iOS, Android, JS, and Wasm. ### Tech Stack: Kotlin ### Why it Matters: Typeahead KMP is a highly optimized, local vector database that handles typos, transpositions, and blind continuation with zero-allocation search via primitive arrays, reactive StateFlow results, and instant cold-start via JSONL import/export.

### Overview: The FLUX 3 Video Prompts repository provides tested recipes for generating video prompts using the APIMart API. These recipes are designed to help developers create compelling video content with minimal effort. ### Key features: - **Model ID:** `flux-3-video` - **Aspect Ratios:** 16:9, 16:10, 1:1 - **Cost Per Image:** $0.048 per second - **Compatible with:** OpenAI-compatible endpoint ### Tech Stack: Python ### Why it Matters: - **Key:** A single key allows access to over 300 different models including FLUX 3 Video. - **Billing:** Pay-as-you-go pricing, no need for monthly subscriptions or prepayment. - **Price:** Estimated cost of the entire set: $0.192. - **Asynchronous:** Tasks are processed in the background, no need for special integrations. Four recipes are provided, each with an example render, duration, and cost on the APIMart route. The repository also includes a table of prompts for easy reference. ### What Separates a Usable Recipe from a Wasted Call: - **Subject:** Focus on the material, not the adjective. - **Light:** Specify the direction and quality of light. - **Framing:** Use ratio and lens language to determine the output. - **Setting:** Describe the surface, backdrop, and environment. - **Finish:** The rendering or film character decides the output. ### Prompt Structure Cheat-Sheet: -

Overview: Ket AI is a Telegram bot that simulates a chatbot, leveraging Google Gemini for conversational context and memory management. Key features: - Gemini-backed replies with conversation context - Hybrid vector memory (RAG) per group - Background memory worker extracting user facts - Memory commands and natural triggers - Image and voice message understanding - Document and file handling - Spontaneous mode with configurable probability - Agentic tool use with safeguards - Per-user, per-chat, and per-IP rate limits - Signed and freshness-checked Mini App `initData` - Asynchronous logging with automatic rotation - Concurrent processing with `/stop` cancellation and agent timeout Tech stack: TypeScript Why it matters: Ket AI provides a fun, opinionated chatbot experience that mimics a human-like conversation, enhancing user engagement and interaction within Telegram groups and chats. Repository: ket0x4/ket.ai Primary language: TypeScript GitHub stars: 11 GitHub description: Overengineered telegram bot that functions as a chatbot README excerpt:

Overview: This repository is a MCP server for Krea.ai, allowing developers to generate stunning AI images and videos using state-of-the-art models like Flux, Hailuo, Runway, Kling, Ideogram, Imagen, and more. Key features: - Access to 20+ AI models through one unified interface - Generate photorealistic images and AI videos from text prompts - Apply custom styles (LoRAs) for consistent aesthetics - Simple setup with one command installation, works instantly with Claude Tech stack: JavaScript Why it matters: This MCP server is essential for developers looking to integrate Krea.ai's capabilities into their projects, providing a powerful toolset for generating high-quality AI-generated content. GitHub stars: 11 Repository: keugenek/krea-mcp

Overview: This repository contains tested image prompt recipes for the Veo 3.1 Lite model, including example renders, aspect ratios, model ID (`veo-3.1-lite`), and cost per image. The recipes are designed to help developers create effective and efficient prompts. Key features: - 8 tested recipes with example renders, aspect ratios, model ID, and cost per image. - OpenAI-compatible endpoint `https://api.apimart.ai/v1` with API key required for $0.064 per second pricing. - Chargeable by the second, no pre-paid subscription needed. - Real-time cost reporting and automatic refunds for failed tasks. - Asynchronous design for easy integration with clients. Tech stack: Python Why it matters: This repository provides a practical guide for developers to create effective image prompts using the Veo 3.1 Lite model, ensuring efficient and cost-effective use of the API. Repository: kf5djy0f62si0/veo-3-1-lite-prompts-cn Primary language: Python GitHub stars: 11

Overview: Norman is a specialized compliance analysis engine designed to automate the validation of Brazilian Technical Standards (NBRs) for civil engineering projects. It eliminates manual document verification, saving engineers 30% of their time and preventing structural failures and regulatory fines. Key features: - Automatically validates NBRs against Brazilian regulatory standards - Uses RAG (Retrieval-Augmented Generation) for instant compliance checks - Reduces time spent on manual document verification by 30% Tech stack: Spring Boot 3, PostgreSQL, Vector Database (PGVector), Google Gemini Pro Why it matters: Engineers spend up to 30% of their time reviewing documents to ensure compliance with NBRs, which can lead to structural failures, heavy fines, or project delays. Norman provides a robust backend in Spring Boot that processes NBRs, stores them in a Vector Database, and uses AI to audit civil engineering projects, providing exact references to standard items, thereby saving valuable professional time and preventing issues. GitHub stars: 11

Overview: This repository is an experiment exploring the application of contrastive learning to movie scenes, aiming to train models to recognize cinematographic styles from movie scenes. Key features: - Dataset generation - Fine-tuning of CLIP for zero-shot classification on movies - Two models trained on sci-fi movies with captions and (director + year) information Tech stack: Jupyter Notebook Why it matters: This experiment demonstrates the potential of contrastive learning in identifying movie styles, showcasing the versatility of the technique in various applications. Repository: kj3moraes/movieclip GitHub stars: 11 GitHub description: An experiment with movie scenes and contrastive learning

### Overview: VectorSmith is a vector database that transforms your vector database into typed, tenant-guarded tools, making it easy to integrate with various agents and environments. ### Key features: - **Tools Compilation**: Converts your `tools.yaml` file into typed, tenant-guarded Python tools. - **Python Integration**: Allows you to import the tools directly in Python. - **MCP Serving**: Enables serving the tools over the Production HTTP server. ### Tech stack: Python ### Why it matters: VectorSmith simplifies the process of integrating vector databases with agents, making it easier to manage and secure data without the complexity of hand-rolling schemas or using vendor-specific tools.

### Overview: CoT-ICL Lab is a framework designed to study chain-of-thought (CoT) learning from in-context demonstrations. It generates synthetic datasets for training and evaluating transformer models, focusing on causal inference and chain-of-thought reasoning. ### Key Features: - **Synthetic Data Generation**: Generates tokenized datasets for training and evaluating transformer models. - **Chain-of-Thought (CoT) Learning**: Employs causal inference to guide the learning process, focusing on the reasoning chain. - **In-Context Demonstrations**: Uses in-context examples to train the model, improving its ability to infer the correct chain of thought. - **Meta-Training**: Allows for meta-training with a focus on causal inference, enhancing the model's ability to generalize and reason. ### Tech Stack: - **Python**: The primary programming language used for implementing the framework. - **Pixi**: A tool used to manage environments and dependencies, ensuring a consistent and efficient development process. ### Why It Matters: - **Enhanced Learning**: By focusing on causal inference, this framework helps in improving the model's ability to reason and learn from context, making it more effective in real-world applications. - **Reproducibility**: Provides a reproducible environment for researchers to test and validate their models, contributing to the advancement of the field. - **Educational Tool**: Ideal for educational purposes, allowing students and researchers to better understand the principles of chain-of-thought learning and causal inference. ### GitHub Stars: 11 ### Repository: - **Repository**: kvignesh1420/cot

### Overview: Seedance 1.0 Pro Fast Prompts is a comprehensive set of tested image prompt recipes designed for developers. It includes eight recipes with detailed outputs, aspect ratios, and the model ID (`doubao-seedance-1-0-pro-fast`) and cost per image. ### Key Features: - **8 Recipes with Outputs**: Detailed examples and cost information for each recipe. - **Model ID**: `doubao-seedance-1-0-pro-fast` for seamless integration with the Seedance API. - **Cost Per Image**: $0.0079 per second, billed by resolution tier. - **OpenAI-Compatible Endpoint**: Supports OpenAI-compatible endpoints, no need for a monthly plan. ### Tech Stack: - Python ### Why It Matters: - **Key Features**: The recipes are tested and optimized for high-quality image generation. - **Cost Efficiency**: Offers a low-cost solution with a flexible pricing model. - **Model Integration**: Seamless integration with the Seedance API, reducing client-side complexity. - **Prompt Structure**: Provides a cheat-sheet for crafting effective prompts, ensuring better results. ### Repository: - **GitHub Repository**: kwmw8z1tvj5/seedance-1-0-pro-fast-prompts-ja - **Stars**: 7 - **Description**: Seedance 1.0 Pro Fast Prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`doubao-seedance-1-0-pro-fast`) and the cost per image. ### Primary Language