Havuz

Python SDK for LLM guardrails with safety classification, PII detection, prompt injection defense, and grounding checks. Protect AI applicat

Self-hosted AI company OS for solo founders: autonomous agents, ERPNext business memory, and owner-visible governance.

Local-first evaluation framework for RAG systems and AI Agents. 18+ metrics, CLI + SDK, framework-agnostic. The pytest of AI evaluation.

本地优先的视频字幕 OCR 工具：提取硬字幕/软字幕，自动定位字幕区域，导出 SRT/TXT，并可选用大模型修正文案。

Exacting Agent Skills for serious thinking, reading, research, and translation.

Unified research workspace for humans & agents — PDF Q&A with citations, Vault, writing. 为人与 Agent 协作的科研工作台；补齐 Zotero 时代的人机上下文断点。

Orchestrating Mathematical Reasoning Agents with Fact-Graph Memory

A free, open-source desktop app for getting work done with AI. It shows you what it would remember before it saves anything.

Turn Codex threads into an organization of long-lived domain agents.

A memory layer for Claude Code that learns as you work. Persistent, per-context memory as plain markdown in your own git repo. No server, no

Watch, approve and steer your coding agents from your phone.

Instella MoE is a fully open state-of-the-art Mixture-of-Experts language model, trained end-to-end on AMD Instinct™ GPUs.

Quieter sessions for Claude Code: less narration, shorter tool output and concise answers focused on the result.

Build UI apps with your AI assistant — together, live. An agent-native, cross-platform declarative-UI runtime in Go: describe UI as JSON, le

Open-source factory for generating, validating, and publishing reusable Agent Skills

Concurrent Go CLI that detects hidden prompt-injection content in documents — invisible Unicode, homoglyphs, hidden HTML, and known trigger

Lightweight, dependency-free logging and analytics for LLM API calls — track tokens, cost, and latency to a plain JSONL file.

Agent-powered iterative optimization: put an AI agent in a verified improvement loop.

Open-source, local-first desktop for AI coding agents and document work. Run Claude Code, Codex CLI, opencode, and Grok CLI in isolated Git

Custom converter: open-weight Hugging Face LLMs/VLMs -> on-device LiteRT-LM (.litertlm), with one-command reproduction

🐧 A native GNOME knowledge workstation that turns documents into evolving AI-assisted knowledge bases.

Eight Claude Code skills for the job hunt that refuse to invent facts about you: CV optimizer, STAR story bank, 360Brew LinkedIn planner, co

Drop-in ggml backend for Rockchip NPUs: offloads llama.cpp/whisper.cpp prefill to the RK3588 NPU

A video-call avatar frontend for your local LLM agent (Hermes/OpenClaw) — replaces the messenger chat and serves the agent's exact slash-com

⚗️ Experimental — A markdown brain that gives your AI coding tools judgment and rigor: sharp defaults, real decisions, and the discipline to

Make forms easier, LLM tools safer. Natural-language quantities, units, dates & ranges parsed into canonical values — zero dependencies, two

一个面向 LLM API 的本地反向代理转发工具。上游服务（如 Coding Plan）过载返回时，自动按固定间隔重试，直到拿到数据。完整透传请求/响应（含 SSE 流式），对客户端透明。

免费书：游戏策划实务即用的 AI·Claude Code 活用法 —— 没有一个数字是编造的。全文免费（CC BY-NC-SA）。韩语纸质原版（ISBN 979-11-12-21479-9）的简体中文版。

Mine reusable skills from your coding agents' transcripts (Claude Code, Codex CLI, or any agent) with outcome-weighted scoring and a local r

Cognitive-structured Multimodal Agent (CMA-Harness): a memory-centric agent for long-horizon multimodal understanding, generation, and editi

A polygraph for your state machine: an LLM derives a formal model from your code, then Polygraph replays real traces against it and exhausti

The fastest lightweight, embeddable tiktoken-compatible Rust core.

Qwen3.6 vLLM Toolkit — Launcher + Templates, Optimized for 2×24 GB. vLLM launcher with an optimized hybrid chat template drawing from the be

Tura, coding agent’ların birden çok araç adımını tek bir runtime-managed command graph içinde çalıştırarak tekrar eden LLM çağrılarını azaltmayı hedefleyen açık kaynak bir agent runtime harness’tir.

Neden yükseliyor · 83 günde 644 yıldız ve son 28 günde 18 commit dikkat çekiyor; token tasarrufu vaadini karşılaştırmalı benchmark’larla desteklemesi de ilgiyi artırmış olabilir. Ancak son 28 gündeki 2 farklı yazar, katkı tabanının hâlâ sınırlı olduğuna işaret ediyor.

FableCut, tarayıcıda çalışan ve timeline'ı JSON, MCP veya REST üzerinden AI agent'ların da düzenleyebildiği bir video editörüdür.

Neden yükseliyor · 83 günde 694 yıldıza ulaşması ve README'de Hacker News'te öne çıkma gibi görünürlük işaretleri dikkat çektiğini gösteriyor; son 28 günde 122 commit ürünün aktif geliştirildiğine işaret ediyor. Ancak 3 farklı yazar olması geliştirme faaliyetinin sınırlı bir ekipte yoğunlaşabileceğini düşündürüyor; yıldızlar gerçek kullanımın kanıtı değil.

The Fable Method, coding agent’lar için görevleri sınıflandırma, kanıt toplama, değişiklik yapma ve sonucu doğrulama adımlarını tanımlayan skills ve bunları sınayan eval’ler sunuyor.

Neden yükseliyor · 83 günde 2294 yıldız almış olması dikkat çekici. README’de belirli senaryolara ait sonuçlar, başarısızlıklar ve ham çıktılara bağlantılar sunulması ilgi görmesini açıklayabilir; ancak yükselişin kesin nedeni bilinmiyor. Son 28 günde commit ve farklı yazar etkinliği yok.

To know what models don't say out loud.

Word-sized charts for React, made for LLMs and humans — 106 chart types, zero runtime dependencies, accessible by default, RSC-safe.

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

开源、本地优先的 AI 长篇小说创作引擎：多智能体世界推演、按弧规划、RAG 长程记忆、逐章审核与断点恢复 | Open-source AI novel generator for long-form fiction.

Open-source control panel for the real llama.cpp: see where a model lands in your VRAM before loading, every llama-server flag per model, on

CPU-first LLM inference server on llama.cpp. Runs useful models on free-tier ARM boxes; rewriting the input made it ~6x faster and more accu

Self-hosted AI SRE for Kubernetes — zero-instrumentation eBPF observability plus a copilot that fixes issues through guardrailed, self-verif

Make Opus 4.8 (or any Claude model) work like Claude Fable 5 — a Claude Code skill + guard hooks (plan gate, model ceiling, per-task enforce

面向招投标场景的本地化 RAG 标书解析、资质匹配与红线提醒助手

Talk to anyone: /coach persona plugin for Claude Code — research any public figure from their YouTube + web, then talk to them

Memory as state, not search — deterministic, dependency-free memory for AI agents. The memory lives in a Markdown file.

OpenAI-compatible server with a live view into any HF model's residual stream. pip install brainscope

Grade every agent, scraper and model in a claim's chain — provenance, trust scoring and audit evidence for LLM pipelines