Trending Repos

TokenSpeed is an inference engine designed to accelerate large language model (LLM) computations.

Why: It enables faster processing of LLMs, which is crucial for real-time applications and improving overall performance.

ChatGPT Team(Business) 促销码自动扫描工具 — 批量发现/验证/价格收集，支持 17 国 34 个码，最高折扣 71% | ChatGPT Business promo code scanner — batch discovery, validation, price collection, 34 codes across 17 countries, up to 71% off

🔍 The hardest search benchmark in the wild — vague, multi-turn, proactive. 200 long-horizon tasks with persona-driven progressive disclosure, scored by verifiable schema-free knowledge-graph evaluation. No vibes, just triplet F1.

Personal-Model First Self Evolving AI Agent 🐘

KVarN is a native vLLM KV-cache quantization backend for your agents: 3-5x more context, throughput above FP16, and FP16-level accuracy. Calibration-free, one flag.

Fast, lossless LLM inference via dual-view diffusion decoding.

Surreal dream sequence engine — procedural narrative generation from emotional state, memory anchors, and symbolic archetypes

Security-native LLM system for AI-generated application security.

Surreal dream sequence engine — procedural narrative generation from emotional state, memory anchors, and symbolic archetypes

End-to-end AI short-video production pipeline. FastAPI orchestration + Spring Boot gateway with multi-model failover, circuit breaker, metering, and full-stack observability. AI quality gating: prompt anchoring, CLIP consistency, AV sync auto-rescue.

An end-to-end framework for multi-speaker transcription that jointly models who spoke, when, and what.

A recursive evidence-gated cognitive runtime for memory-native AI agents, combining hybrid retrieval, temporal reasoning, async learning, and plug-and-play tools.

Estimate whether a Hugging Face model fits and fine-tunes on your local GPU.

Python port of Claude Code's agent-runtime architecture, on LangChain

LLMs have ended reverse engineering as a high-barrier skill. Case study: reconstructing 6/7 custom HTTP signing parameters of Douyin (TikTok CN) v38.1.0 native library in 30 days using Claude Opus. 大型语言模型（LLMs）已经终结了“逆向工程是一项高门槛技能”的时代。案例研究：使用 Claude Opus，在 30 天内成功还原抖音（中国版 TikTok）v38.1.0 原生库中 7 个自定义 HTTP 签名参数中的 6 个。

Turn Claude Code into its own Meta-Harness — a skill that evolves the scaffolding around a fixed model (memory, retrieval, context, prompts) via a native propose→score→Pareto loop. Native reimplementation of Meta-Harness (Lee et al. 2026).

Don't make LLMs honest. Make every factual claim auditable. — An LLM Claim Auditing Layer with T1-T7 truth gradients. 98.1% business effectiveness on LiarBench v0.2.

MemPrivacy is a privacy-preserving personalized memory management framework for edge-cloud agents.

AgentGuard：An Attribute-Based Access Control Framework for Tool-Use LLM-Based Agent

AI 文献检索与综述 Agent：支持多源检索、代码仓库定位、开放 PDF 下载、证据链与中英双语报告。

Hermes Agent plugin for cron observability. Turn hidden automation into visible spend.

Dataflow-Oriented Reinforcement Learning for (Multi-)Agentic LLMs

使用 QQ/微信 聊天记录结合向量数据库让AI更好的扮演对方的角色，在不微调模型的情况下可以达到可观的效果。把曾经的美好，续成往后的陪伴。 --续温

Official Implementation of LongLive-RAG: A general retrieval-augmented framework for long video generation.

Implementation of D4RT, Efficiently Reconstructing Dynamic Scenes, from Deepmind

Open Agent OS for Claude Code, Codex, and Cursor: meta-agent builder, A2A Hub routing, local ontology, memory and security gates.

🩵uncensored Fully offline AI girlfriends harem Openclaw+Local LLM+GPT-SoVITS+ComfyUI image+Live2D+desktop pet+SilllyTavern Character card import| Dual channels for QQ & Telegram | Dynamic 8GB VRAM scheduling, can run offline完全离线 AI 女友后宫 |小龙虾+本地大模型+GPT-SoVITS+ComfyUI+Live2D+桌宠+酒馆角色卡导入|QQ & Telegram 双通道 | 8GB 显存动态调度断网可跑

Implementation of Kalmanformer, modeling the Kalman gain with a transformer

一个面向架空世界故事创作者的 AI 架空世界与长篇故事引擎。它能从一句设定出发，自动生成层 层嵌套的历史、地理、角色关系与事件网络，并持续写出贴合世界观的剧情。支持 CLI 与可视化界面，可按时间和空间路径扩展世界，让故事像真实宇宙一样不断生长。项目用数字路径管理嵌套的时间、空间、 角色和事件节点，按需生成并补全世界资料，支持本地 Demo、OpenAI/DeepSeek JSON Agent、CLI 和黑白 Web 可视化界面，用于创建可持续展开的故事世界与剧情会话。

training your 0.1B llm from zero

Lightweight agent multiplexer, all in one Web dashboard

Self-evolving context for any coding agent.

Fix "Codex cannot access its local database" / "migration 1 was previously applied but has been modified" / "timed out waiting for state db backfill after 30s" on Codex Desktop (Windows/WSL/Mac/Linux) after the 0.130 → 0.131 update. CRLF/LF sqlx checksum drift in state_5.sqlite + logs_2.sqlite. Sessions preserved. See openai/codex#23787.

GitHub Action + CLI that auto-writes PR descriptions, adds labels, and reviews code using OpenAI

coding agent

Retrieval is CheapShow Me the Code: Executable Multi-Hop Reasoning for Retrieval-Augmented Generation

An open-source End-to-End AI model for self driving

A step by step implementation to build an agent that runs for week to complete a task and self improve.

Implementation of WorldParticle, Unified World Simulation of Lagrangian Particle Dynamics via Transformer

A visual, editable context layer for Codex. AI edit AI's context with surgical precision, giving you more control and freedom over what Codex sees.

Open Source Agentic Security Scanner for your codebase.

Implementation of RigidFormer, Learning Rigid Dynamics using Transformers

An open-source End-to-End AI model for self driving

Progressive agentic reading harness: sentence maps, bounded chunks, Chronicle Markdown books, packets, validation. Python CLI + Next.js demo. MIT.

Implementation of Attention Schema-based Attention Control (ASAC), A Cognitive-Inspired Approach for Attention Management in Transformers

Desktop realtime speech translator for routing translated audio between virtual audio devices in calls.

Biologically inspired language model using Jaccard Surprise as its only training signal. No backprop. No GPU. Online Hebbian learning from corrections. Two-layer cortex with apical feedback. Runs on CPU under 200MB.

Move beyond prompts. Build verifiable, intent-driven Semantic Agents.

artificial-intelligence computer-vision deep-learning emotion-detection facial-expression-recognition tensorflow opencv cnn machine-learning python real-time-ai

Open-source AI Operating System (AIOS) blueprint for Claude Code & Codex — turn any agent into your personal second brain + Chief of Staff: skills, self-verifying workflows, a second-brain wiki, and deterministic tools. Fork it for any role or domain.

Free-tier AI load balancer — auto-rotates 28+ keys across 9 providers (Gemini, Groq, Mistral & more). OpenAI-compatible drop-in. Works with any agent framework.

一个生产级的深度研究 Agent 系统，从零构建多智能体编排、Red-Blue 对抗降噪、 语义级上下文压缩、跨 Agent 共享记忆四大核心能力，配套 165 次独立实验 + Bootstrap 统计显著性检验的完整评测体系。

Sub2API gateway subscription → Codex CLI / Codex App injector. macOS-native, zero-deps Python single-file. Supports codex2api.com and any self-hosted Sub2API instance.

Command-line interface for querying large language models

Clinical Rag AI

Lightweight Python SDK for LLM inference logging and observability

Pool the free tiers of 18 LLM providers (Groq, Cerebras, NVIDIA NIM, Gemini, OpenRouter, Cloudflare, Hugging Face…) behind one OpenAI-compatible /v1 endpoint. Free, zero-config, automatic failover. Codex/agent ready.

When Should Models Change Their Minds? Contextual Belief Management in Large Language Models

FRIDAY is an autonomous cognitive AI operating system for memory, reasoning, voice, cybersecurity, and self-improvement. It runs locally and is designed to move toward general intelligence. Built by a self-taught 17-year-old developer.

Make any web app voice-controllable in 37 languages. Drop-in widget + open-source agent server — visitors speak, the agent talks back and drives their browser in real time.

Docker sandbox backend for deepagents

Your AI's memory grows forever. Your token bill doesn't. A cross-provider memory layer for LLM apps.

AGI-powered social network bot for automated interactions

Deterministic, scripted, resumable multi-agent orchestration for LangChain deepagents — a community port of Claude Code's Dynamic Workflows.

Workflow framework for AI agents. MCP-native typed graphs, visual editor, Python export.

Sovereign agentic CLI — Textual TUI, multi-provider LLM mesh, persistent sessions, skills, MCP, Telegram gateway, autonomous yolo mode

local-llm ollama llama-cpp vllm deep-research research-agent rag retrieval-augmented-generation agent llm-agent local-first privacy offline qwen searxng faiss bm25 python cli ai-research llm-tooling

A Retrieval-Augmented Generation (RAG) based AI Teaching Assistant that delivers accurate, context-aware answers from custom educational resources using LLMs, embeddings and semantic search.

A high-throughput semantic routing engine for AI agents and LLM applications. Executes local intent classification using ONNX embeddings, FAISS indexing, and asynchronous micro-batching.

Sub-second RAG regression testing. Define golden questions, detect lost chunks in CI. pytest for your RAG pipeline.

Implementation and explorations into MPO / DMPO

OWASP ASI-aligned red-teaming and evaluation framework for AI agents

SageSched: Intelligent LLM Request Scheduler with Workload Prediction — QoS-aware dual-queue scheduling for black-box LLM APIs (OpenAI/Azure/Doubao/Gemini)

A practical, continuously updated learning resource covering every layer of Retrieval-Augmented Generation — from how to split a document to how to evaluate whether your system actually works. Each technique comes with runnable Python code, a Feynman-style explanation, step-by-step algorithm breakdown, Mermaid diagrams and worked examples.

OpenTelemetry-native, local-first observability for AI agents. DuckDB on disk, Next.js viewer on localhost, no SaaS. Adapters for Anthropic, OpenAI, LangGraph, AutoGen, CrewAI, and Claude Code's native OTel emit.

Source code for our paper ''When Knowledge Is Not Free: Cost-Aware Evidence Selection in Retrieval-Augmented Generation''

Open-source PII firewall for LLM apps. Detect, anonymize and rehydrate sensitive data before it reaches OpenAI, Anthropic or any LLM provider.

本地中文小说创作 Agent (LangChain 1.0 + LangGraph)