iyulab-rnd 

IronHive Anthropic provider

Core components for LMSupply libraries: HuggingFace model downloading, caching, GPU execution providers, and common utilities.

FluxIndex Core - Domain models and application interfaces for FluxIndex RAG system

Shared abstractions for the Flux ecosystem - IEnrichedChunk, ISourceMetadata, and related interfaces for RAG data pipeline interoperability

FluxIndex SDK - RAG orchestration core (context, indexer, retriever, DI helpers). Storage and pipeline integrations ship as separate FluxIndex.Storage.* / FluxIndex.Integrations.* packages.

IronHive abstractions and interfaces

Text processing infrastructure for LMSupply - tokenization, vocabulary management, and encoding utilities

Complete document processing SDK optimized for RAG systems. Transform PDF, DOCX, Excel, PowerPoint, Markdown and other formats into high-quality chunks with intelligent semantic boundary detection. Includes... More information

llama.cpp server management for LMSupply. Provides llama-server downloading, GPU detection, and server pooling for GGUF model inference.

Pure document extraction SDK for RAG systems. Zero AI dependencies. Extract text from PDF, DOCX, Excel, PowerPoint, Markdown, HTML, and text files. Provides IDocumentReader interface and implementations. Use... More information

Core vision processing components for LMSupply: image loading, preprocessing, and tensor conversion utilities shared by LMSupply.Captioner and LMSupply.Ocr.

A .NET library for local LLM text generation via GGUF/llama-server, with streaming, chat templates, and automatic hardware-aware backend selection. Add LMSupply.Generator.Onnx for ONNX Runtime GenAI models... More information

High-performance library for extracting HWP/HWPX Korean word processor documents to Markdown

A simple .NET library for local text embeddings with automatic model downloading from HuggingFace. Supports CUDA and CoreML GPU acceleration.

Text preprocessing library for RAG pipelines: PII masking, content filtering, and intelligent chunking including semantic chunking with Korean language support.

Core library for FluxCurator - Minimal dependencies. PII masking, content filtering, and rule-based chunking with Korean language support.

The Quality Layer for RAG Data Pipelines - LLM-powered enrichment and quality assessment

A simple and intuitive library for accessing object values using string path expressions in .NET.

A lightweight, zero-configuration semantic reranker for .NET. Supports multiple cross-encoder models with automatic download, GPU acceleration, and HuggingFace caching.

A simple .NET library for local OCR with automatic model downloading from HuggingFace. Features 2-stage detection+recognition pipeline using PaddleOCR models. Supports CUDA and CoreML GPU acceleration.