Trending Repos

A curated list of public projects, integrations, and discussions built on Jev — TypeSafe AI's System One model for typed decisions.

AI video processing pipeline for generating vertical shorts using LLMs, Whisper transcription, highlight detection and automated editing

Turn any LLM into a Jev-style decision model: typed decisions, real probabilities, no training. (continue updating, welcome any issue and PR request)

JevChat-Windows：聊天窗口旁挂的回复辅助。窗口截图 + 本地离线 OCR 读对方消息 → Jev 判断意图 → 3 条候选一键填入，发送永远手动

1207 public resources for Jev, TypeSafe AI's System One decision model, indexed by decision pattern. Source citations, dated link checks and scheduled call-site text checks; runtime and performance are not independently tested here. EN/中文, JSON schema and platform compatibility.

麦麦省钱助手 —— 基于麦当劳 MCP 的点餐成本优化 Skill。算出当前门店最省钱的点餐方案，并主动提示凑单陷阱。

聊天悬浮窗助手（macOS）：屏幕感知 + 本地小模型判断意图与风险，按话术生成回复候选。纯只读。

Turn local language models into Jev-style structured decision models. Get results from text and images with prefill alone—no token-by-token decoding required.

Jev-compatible API endpoint based on open models (prefill-only)

Blender add-on: create and edit 3D models by chatting with AI - send text or an image, get working bpy code

Per-turn model & reasoning routing for Codex, driven by Jev (TypeSafe System One): picks the model, thinking depth and speed mode for every turn.

[Public preview] Externally scored agentic ML research benchmark: 60 tasks, real competition ground truth. Open protocol, operated evaluation.

微信聊天分析工具，支持本地 Laya 与 API 模型，提供意图识别、情绪感知、人物画像、群聊画像、好感度分析和 MBTI 聊天推测。

Research and writing can be automated. Authorship cannot. Evidence list + agent skills for spotting AI-produced papers no human stood behind.

针对GPT 6.0,V4flash的破甲实验项目

Refusal removal as a loadable KV-cache graft — zero weight modification, no refusal-direction projection, fully reversible.

Zoom any paper for your better understanding

Zoom any paper for your better understanding

Shrinkflation detector for Codex

A fast browser agent: Jev picks each action from what is on the page, an LLM reads and plans, and every claim in an answer cites a quote from the page.

🚀🚀 A multimodal System One decision model that gives calibrated answers to typed questions about screens, photos, video and text in one forward pass.

Kyojin: the Yamz inference engine for AMD Strix Halo (ROCm, gfx1151), built on ExLlamaV3. Runs 300B-class MoE models on one 128 GB mini PC.

Jev-Mem: System-One Controlled Agentic Memory

STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization

A categorized directory of X accounts covering AI, LLMs, agents, and technology, with original profile bios and follower counts. Available in 7 languages.

A local app that rents a GPU in your own Colab account and serves Qwen3.8 on it: 27B on an A100-40G, or Flash-Next (125B-A6B MoE) on an A100-80G. The cost is shown before it starts; chat with images; Codex or any OpenAI client uses a fixed local /v1. Auto-stop, CU ledger, live prefill/decode. Windows, macOS, Linux.

vLLM based inference framework for agentic workload.

A Qualifying Exam for Coding Agent as Robot Learning Engineers

On-demand Higgsfield API image and video generation for coding agents, with model-specific docs and resumable jobs.

Implementation of the recurrent looped transformer proposed by Yifan Zhang of Princeton

Your coding agent starts every session knowing nothing about your project. One line, and it writes its own memory from your real code, then grades itself.

由 LLM 驱动的可交互智能体：有记忆、有性格、有自己的日程与作息

面向企业合同全生命周期的生产级 AI Agent 平台，支持合同解析、条款结构化、风险审查、HITL 人工确认、版本管理、客户回传 Diff、谈判辅助及 PDF/Email 交付；基于 LangGraph + RAG + ContractPolicy + Agent Workflow，实现可追溯、可恢复、可审查的合同智能处理闭环

code agent

Give the agent "this" and he gets a real PC.

端到端智能体平台：把本机的 AI Agent 变成人人可用的 Web 产品（Vue 3 + Flask + Go WS 网关 + Docker 部署）

Give the agent "this" and he gets a real PC.

Three trap-laden fix-it tasks with hidden graders for testing an AI coding agent's judgment. Sandbox setup, prompts, answer key, and results for Qwen3.8-Flash, Qwen3.6-35B and Claude Opus 5.

Long-term memory for AI coding agents (Claude Code, Codex, opencode) as a tree of plain files

A way of running a project with coding agents. A team of roles, a loop with a stop in it, and rules your agent cannot ignore. Text files and small Python programs, no dependencies.

Agents you can inspect, replay and trust. An agent framework for TypeScript and Python, built on an append-only event log.

Execution posture harness for Claude Code. Your coding agent stops one step early; Castra is the step.

Save your GPT-6 Astra tokens. Let Luna do the dirty work.

Jev-shaped (TypeSafe System One) classification wrapper over OpenAI-like clients

PromptForge: An enterprise-grade, multi-platform Prompt Engineering framework for AI Agents. Featuring 56 structured templates across 9 domains, 118 ready-to-use presets, and native adapters for Cursor, ZCode, DeepSeek Harness, Claude, and OpenAI. Built with FastAPI Web UI & Typer CLI.

Budget caps, runaway-loop detection and circuit breakers for AI agents. Zero dependencies, framework agnostic.

Evidence-grounded academic research agent with claim verification, prior-work analysis, and persistent research workspaces.

Status bar for Codex CLI and Claude Code: context, 5h/weekly quota with reset times, tokens.

Source codes for the paper "SyncWorld: Visual Calibration Enables World Models as Zero-Shot Simulators"

Tactical Identity Operator for Linux & Hybrid Active Directory

Open-source guardrails for AI agents powered by decision models like Jev; checks prompts, retrieved content, tool calls and responses for prompt injection, jailbreaks and secret/PII leaks.

Use GPT-6-Astra from local Codex for free - no paid plan, no API key, just a signed-in prism.openai.com session

Explorations into Survival Reinforcement Learning, proposed by Tiofack et al. of Inria earlier this year

Live Benchmark for long Horizon tool using Agentic Systems

一键诊断并修复 Codex / ChatGPT 桌面版「正在重新连接 1/5 → 5/5」| The Codex Rust backend ignores the Windows system proxy unless respect_system_proxy = true

Fast, typed, calibrated evaluations for LLM and agent outputs, powered by Jev — with simple, framework-agnostic Python APIs

Self-hosted multi-model agent with JitRL memory, auditable council runs, SSH operations, live server screen, and macOS/Windows desktop client

Portable offline AI assistant for SAP/ABAP learning, code drafting, debugging, and development planning. Runs locally with lightweight models, streaming responses, personal note retrieval, and optional GPU acceleration.

New bio inspired AI

100% Free, Offline, 8GB RAM Computer-Artificial Intelligence + Formal Safety (Gate + Ledger + Council + Shadow Mode) — Free alternative to $200/mo OpenAI Operator / Perplexity Computer / ClawBot / Claude Computer

Implementation of the proposed CCBP (Continuous Continual BackProp) in Pytorch

Open source release of Data Science Agents that can analyze and transform datasets as well as train and infer models to augment your data scientists.

Implementation of EmbeddingGemma 2, Google DeepMind's 740M parameter multimodal embedding model, in Pytorch

迹录 Tracord ——基于本地小模型多智能体的智能旅行规划助手 / AI travel planner with local SLM multi-agents — light up visited cities on a map, log expenses with one sentence, and generate executable day-by-day itineraries from plain language.

QuotaCrew for Codex - multiple accounts, quota monitoring, automatic switching and experimental conversation continuation in Codex Desktop and VS Code. Open-source Windows app.

Eight-market data collection, quantitative analytics, technical structure, and Hermes AI forecasting system

The Electrocheck fact-checker

AI-powered website builder that uses multi-agent orchestration to generate production-ready websites from natural language prompts.

pxchat plus 增强 | nonebot-plugin-pxchat 的增强补丁：修复稳定性问题（调用超时/上下文污染/模型过滤自动切换/长文本处理/并发加速）。关键词：pxchat plus nonebot插件 聊天插件 增强补丁。

Hands-on LangChain learning repository covering Chat Models, Prompt Templates, Chains, RAG, Agents, and Tools with practical Python examples.

LLM-led e-commerce content review with rule-based safeguards, tenant-isolated MySQL, Redis workers, and auditable publication decisions.

RaDUO turns live radio broadcasts into multilingual STT datasets for ASR research. It records stations in parallel, splits streams into clips, transcribes audio, extracts quality features, and stores searchable metadata. Includes 16 kHz mono WAVs, transcripts, station info, languages, and optional word timings.

Implementation of Self-Play Pretraining with Zero Data

HEIMEL Intent. Realized. HEIMEL is open infrastructure for the transition from intent to consequence.

In-flight circuit breaker & loop recovery for LangGraph, LangChain, OpenAI & Anthropic agents. Catches reasoning loops, prevents runaway LLM costs with budget hard caps, and auto-injects Reflect & Pivot prompts.

一个可复现和可控的学术论文写作工具链，用于规划和起草学术论文。它能够对引用、论点强度、行文表达以及 LaTeX 完整性进行系统性检查。A reproducible and controllable toolchain to plan, draft, and gate an academic paper draft. It performs systematic checks on citations, claim strength, prose, and LaTeX integrity.

Adaptive Prototype Memory: a small prototype learner benchmarked against backprop, plus a chat demo where it learns from a local LLM.

Implementation of a transformer whose attention operates in quasimetric space

Turn a statute book into Akoma Ntoso: acquisition, transcription, anchor-driven structuring, retrieval and comparison for legal text. Apache 2.0.