Havuz

Glyd, model ağırlıklarını ve KV cache'i bit düzeyinde kayıpsız biçimde sıkıştırıp GPU'da kullanmayı hedefleyen bir compression yazılımıdır.

Neden yükseliyor · README, yaygın bir maliyet sorununa doğrudan yanıt verdiğini ve 32B ile 72B modellerde daha az GPU ile çalışmaya dair ölçümler sunuyor; ancak 9 günlük repoda 170 yıldız ve 474 commit'in yalnızca 2 farklı yazardan gelmesi, ilginin ve geliştirme temposunun geniş bir katkıcı tabanına dayandığını göstermiyor. Yükselişin kesin kaynağı veya bağımsız benimsenme verisi bilinmiyor.

Cut your Claude bill 59–70% by rendering bulky LLM context as dense PNG pages — 100% read accuracy, exact per-provider billing math (Anthrop

Context compression for MCP. Same upstream call, exact results recoverable.

Token-saving companion for OpenCode — 42 compression layers, zero risk, no caveman speak

GPU-accelerated transparent compression S3-compatible storage gateway. Drop-in replacement for AWS S3 endpoints; cuts your S3 bill 50-80% wi

Windows local-first file organizer and media catalog with MCP tools for AI assistants. Verified batch archiving. 本地文件整理、媒体建库与可校验批量归档，支持 AI 通

Private, provider-neutral context compression for LLMs. Reversibly compress schemas and logs inside your perimeter with zero third-party cal

RAR implementation in Rust

First open-source implementation of Google TurboQuant (ICLR 2026) -- near-optimal KV cache compression for LLM inference. 5x compression wit

Road to Gramby's Save Format Explained: Roblox Build System Guide 2026

Carves your agent's tool outputs down to what matters. Never cuts what doesn't come back.

TurboQuant KV cache compression plugin for vLLM — asymmetric K/V, 8 models validated, consumer GPUs

First open-source TurboQuant KV cache compression for LLM inference. Drop-in for HuggingFace. pip install turboquant.

Losslessly compress the tool output your AI agent sends the model, prove every value survived, and report the exact tokens removed. Run it a

[ICML 2026] Inference-time LLM structured pruning via probe-based representation-parameter coupling

Convert long AI conversations into portable conversation state graphs for LLM handoffs.

Fast KV-cache quantization for Apple Silicon (MLX) — 43 research-adapted compression methods with Metal kernels

The worlds #1 token optimizer reduces token consumption using Claude code cli by up to 98.9%

Unified KV-cache compression for LLM inference: 12 Python-native methods, guarded add-on composition and routing, analytical capacity simula

The Best Open Source PDF Compression Engine

Fast DEFLATE, zlib and gzip decompression in pure Rust. Supports zero-copy streaming and no-std/no-alloc.

Consumer-aware compression for embedding indexes and LLM KV caches — compress by the metric the downstream consumer actually uses. PCA-Matry

High-performance Rust token compression engine for LLM inputs. Plugin-based, 50–95% token savings, AI-export diagnostics, CLI / Server / IDE

TSCG - Deterministic tool-schema compilation for LLM agents, including text-mode tool use without native tool calling and context-budget RAG

A juicy token compression algorithm. Compress upto 95% of tokens without any loss. Free & Opensource

Fast and simple local PDF compression for macOS, Windows, and Linux. Drag, drop, done.

Free, fast, cross-platform archiver — extracts ZIP/RAR/7z/tar, creates ZIP/7z/tar. Desktop app + CLI + MCP.

Web app to manage and keep your Nintendo Switch game library up to date: updates, DLC, compression, automations, setup wizard, stats, your y

Архиватор с открытым исходным кодом на Rust — пережимает JPEG, PDF, zip и WAV по-настоящему, а не хранит как есть. Альтернатива 7-Zip и WinR

A smart Log ingestion tool that compresses log files losslessly using template extraction and Parquet storage, then lets you replay or query

Log store in Rust that keeps each line's key set once in a trie and its values as typed columns: a quarter of raw size, queries in milliseco