Python agents

Multimodal prompt generator nodes for ComfyUI, designed to generate prompts for QwenImageEdit and Wan2.2. Supports local LLM / local GGUF models (Qwen2.5-VL, Qwen3-VL, Qwen3.5 and Qwen3.6) and Qwen API for image and video prompt generation and enhancement.

6 progressive demos to detect, prevent, and self-correct AI agent hallucinations, from Graph-RAG research to production deployment on Amazon Bedrock AgentCore

Agentic OS done right — create agents, chat, delegate work, stay safe. Self-hosted control plane with pluggable memory, guardrails, and execution.

Autonomous multi-agent system built using LangChain and LangGraph that orchestrates collaborative agents—Supervisor, Researcher, Writer, and Critiquer—to automatically gather information, generate detailed research reports, refine output, and visualise the workflow as a graph. Includes a Streamlit web UI and optional Graphviz visualisation.

🚀 Financial Anomaly Detection with DeepSeek and Isolation Forest – A powerful, locally-run tool for detecting financial anomalies using Isolation Forest and DeepSeek LLM. Features AI-powered insights, interactive time-series visualization, and automated PDF audit reports. 🔍📊

Open-source AI platform for personalized conversations — embeddable chat widget, pluggable LLM engines, PII redaction, voice I/O, and admin-driven intent & preset personalization.

🐱 MiNerva·咪涅瓦 — 替你干活的赛博猫｜An autonomous AI agent that actually does the work.

Customer Conversational Intelligence Platform Powered by an LLM Agent

[ICML2026] The official code of "Beyond Gemini-3-Pro: Revisiting LLM Routing and Aggregation at Scale"

一款专注于慢性病管理的AI平台，通过多Agent协作问答、干预效果仿真预测、认知评估与训练、亲情账号关怀等核心技术，为慢病患者提供专业、个性化的健康管理服务

Proactive 5h usage guard for Claude Code (macOS). Pause before extra wallet on long batch/subagent runs. Account-level window visibility. /usage-guard skill + daemon. Not affiliated with Anthropic.

睡前消息知识库 Agentic RAG

The recursive self-improvement loop for your skills (including itself) — runs a retro after every session and turns corrections into permanent skill improvements, so skills compound instead of plateauing.

Full-stack GraphRAG application with Django backend and React frontend. Build knowledge graphs, detect communities, and chat with documents using hybrid retrieval and LLM-powered reasoning.

灵感象限-Ideasphere：自媒体视频一站式剪辑技能包。去静音剪辑→语音转字幕→LLM纠错→字幕烧录→多平台导出，全程分步确认，批量处理。

Tlamatini is a self-hosted AI developer assistant for local or cloud LLMs: hybrid RAG (FAISS+BM25) over your codebase, a 100+ tool Multi-Turn engine, an 86-agent visual workflow designer, 28 SKILL.md skills, a universal MCP client, and ACPX, a runtime for external coding agents like Claude Code, Codex, Cursor and Gemini CLI.

Search your Claude Code conversation history by content (not just titles), then resume any session. grep + fzf, no index.

Open-source AI agent operating system: one file-native brain (230+ skills, 160+ specialist agents) run across many sealed client arms, with per-client token attribution and opt-in budget caps. Organic, portable, MIT.

Claude as subagents in Codex — scoped workers, resumable sessions, git-ignored logs.

Agent tools from many descriptions in a single integration. Authorize operations, not just tool calls, and approve them when a human is required. The agent gets its own identity and never holds your credentials.

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

12 agent skills for AI-driven software development — a gated idea-to-ship pipeline: 7w3 design, build-ready specs, executable architecture, honest review loops, a parallel worker-fleet lane, QA, safe deploy. Works with Claude Code, Codex, Cursor and any skills-capable agent. No vibecode.

A fully locally deployable, NotebookLM-like application designed for long-form unstructured documents

Adversarial multi-model consensus protocol for Hermes Agent. Independent solve. Adversarial critique. Formal scoring. Verified synthesis.

An AI dev team in your editor — ~48 specialist agents and an enforced, proportional workflow with approval gates. Process, not prompts. Open-source, no lock-in; works in Claude Code, Cursor, Kiro & VS Code.

ScholarChat AI is an AI-powered academic assistant built with RAG. Students can upload syllabus PDFs and ask questions using text or voice. It provides accurate answers based on the uploaded document.

20 runnable LLM agent design patterns in Python, benchmarked, traced, offline-compatible. ReAct, multi-agent, constitutional AI, and more.

MCP server giving AI agents (Claude Code, Copilot, Antigravity) bi-directional sync with Overleaf - free tier included. 24 tools: LaTeX check, format, fix, compile, push.

生产级通用 Text-to-SQL 服务：基于 FastAPI + LangGraph + RAG，将自然语言转换为安全可执行的 SQL，内置 SQL Guard、few-shot/schema 检索、限流、缓存、Langfuse 追踪、Prometheus/Grafana 监控，并基于 CSpider 做准确率评测。

🤖 humux — /ˈhjuːmʌks/human multiplexer, your self-hosted personal AI agent

Academic research skills for literature review, evidence checking, writing, slides, figures, and research workflows.

streamlit region:us

gradio region:us

gradio mcp-server region:us

gradio mcp-server-track mcp-server region:us

gradio mcp-server-track ai-agents leaderboards incentivized-contests Agents-MCP-Hackathon mcp-server region:us

gradio agent-demo-track mcp-server region:us

gradio gradio-custom-component SimpleTextbox designer drag and drop custom designs region:us

gradio mcp-server-track mcp-server region:us

gradio region:us

gradio region:us

gradio region:us

docker region:us

gradio region:us

gradio region:us

docker jupyterlab region:us

gradio region:us

gradio region:us

gradio region:us

gradio region:us

gradio region:us

gradio region:us

gguf moe vlm vision agentic text-generation

transformers pytorch llama text-generation code math

transformers safetensors phi3 text-generation nlp code

transformers safetensors llama text-generation conversational en

transformers safetensors glm text-generation chatglm conversational

transformers pytorch llama text-generation ar arxiv:2309.12053

transformers gguf llama text-generation trendyol llama-2

gguf text-generation ru en endpoints_compatible region:us

transformers safetensors deepseek_v2 text-generation fp8 vllm

mlx safetensors qwen3_next quantized text-generation conversational

transformers safetensors gpt_bigcode text-generation code granite

docker leaderboard modality:text submission:automatic test:public language:icelandic eval:language region:us

gguf instruct finetune chatml axolotl roleplay

gguf MoQ mixture-of-quants GGUF QWEN quantization

transformers safetensors gemma4_assistant text-generation any-to-any arxiv:2607.02770

transformers safetensors qwen2 text-generation code codeqwen

transformers gguf qwen3_5_moe qwen3_5 reasoning agentic

transformers safetensors gemma4_unified image-text-to-text abliterated uncensored

Local AI Agent Runtime

transformers safetensors chatglm glm thudm chat

AI coding agent that shows its work — every step visible, every patch approved by you. Build projects brick by brick, stay in control.

AI smart home assistant for Home Assistant: LangGraph agent + MCP tools + RAG knowledge graph + vision perception

Integration of the advanced llama2 AI model with Telegram to provide real-time chatbot responses. Seamless interaction, smart replies, and scalable deployment.

A Selenium-based chatbot for Kick (Kick.com) that automatically logs in using stored cookies, sends messages from a text file to a specified channel, and supports both headed and headless modes. Features include sequential message sending, status indicators, real-time message display, and adjustable message intervals.

Reactive chat bot framework

Claude Code의 영구 기억 시스템 — 4-layer 파이프라인 (세션 검색 + 메모리 회수 + Memory Compiler). Karpathy LLM-as-Compiler 패턴 실증판.

Remembers your project so you don't have to explain it again. Memory across Claude, Codex, Cursor.

Many-agent, quality-calibrated orchestrator for Kimi Code with 115 vendored official skill packages — a first-class agentic backbone (live ContextGraph, explicit state machine, forward-only rollback) over a deterministic 6-lens verification harness with pure pass/fail gates and human-gated output. No LLM ever computes pass/fail.

Agent-native image-editing SDK for Claude Code. 21 MCP tools + /decompose skill — semantic layer splits, L1–L5 cultural scoring, region inpaint. Powered by ComfyUI, Gemini, or mock.

Python SDK for Lighton API, the 🇪🇺 European industrial-grade retrieval infrastructure powered by its frontier retrieval models developed in-house

Connect any AI agent to your personal Telegram via the official Business API — no userbot, no ban risk, no Premium. Agent reads & searches full history, drafts replies; every reply waits for your ✅. MCP server + 24/7 collector on SQLite/FTS5.

Make the AI coding agents on your machine (Claude Code, Codex, Reasonix, ZCode) work as one team. Zero-dependency, local, MCP + CLI.

Compression with a quality contract — cache-aware, causally-pruned LLM context compression for agentic runtimes, certified non-inferior across 7 domains. Works with any SDK.

rust-in-peace — agentic security review for Rust: autonomous find/triage/patch for unsafe/FFI memory bugs, panic-DoS, and deserialization-trust flaws (Miri / sanitizer / panic / hang detectors). A Rust-security fork of anthropics/defending-code-reference-harness.

Simplified Technical English for Code - extract, adapt, and enforce ASD-STE100 rules for software documentation. 9-agent pipeline, 59-test benchmark (96.6%), multi-language scaffolding.

Security-posture-first list of sandboxes for running AI coding agents' code, ranked by isolation, egress control and secrets handling

Interactive patcher for the Claude Code native binary. Live inline thinking, detailed tool calls, subagent model overrides, custom branding — and your ChatGPT/Codex-plan GPT models as native models. Pure Python.

Voice Agent implementations for the VAmoS Benchmark

Everyone deserves a chief of staff - your personal jarvis | not a chatbot. not another ai agent. | beta

Read-only Abaqus, LAMMPS and Slurm run audits with optional OpenAI agent triage

Heddle — The policy-and-trust layer for MCP tool servers. Turn YAML configs into validated, policy-enforced MCP tools.

transformers safetensors qwen3 text-generation unsloth conversational

AI job-search scout: daily board sweeps, LLM triage, human-in-the-loop by design. It finds the jobs, you still have to apply!

AI code review MCP server with Zero-Trust approach. Multi-model support, caching, multilingual. Works with Windsurf & Claude Desktop

An agent that builds agents — closed-loop AI system that interviews, designs, tests, and ships standalone Python agents.

Medox is an AI agent that helps FR pharmacists navigate drug shortages & find secure therapeutic alternatives. Powered by Mistral.

🛡️ Open-source LLM security library for Python. Protect your AI apps from prompt injection, data leakage, and harmful outputs in 3 lines of code.

An LLM microservice for the learning platform Artemis