🧠 AI & ML

growmos: a living, queryable knowledge graph for codebases and AI agents

Zero‑dependency, shared memory for multi‑agent workflows

Local-first desktop client for DeepSeek Harness, enabling AI agent plugins with vision

Runs AI agents locally

Journeyman: CLI benchmark that profiles LLM agents' process quality

Delivers detailed multi‑axis performance scores

A functional open-source implementation of Claude for seamless browser integration.

Offers reliable, customizable AI tools without restrictions.

P2P multi-agent orchestration with crash-safe SQLite persistence and real-time streaming

Crash recovery and smart routing for distributed systems

Enterprise AI workbench with desktop, sign‑in, policy, audit

Enterprise‑grade AI hub with hardened desktop

Jiaojie: a skill that hands off AI context across models and languages

Facilitates seamless AI handoffs

Offline AI assistant that watches, speaks, sees, writes, and browses.

All offline, no cloud, full control.

LIWM: Auditable, provenance‑based memory for AI agents

Provenance‑gated memory for trustworthy agents

AI agent CLI that runs, manages, and schedules LLM-powered tasks

All‑in‑one terminal for LLM agents, memory, cron, gateway

OpenFlow: a visual builder for AI workflows, letting you design and run pipelines.

Visual AI workflow builder with multi‑agent orchestration.

Local AI coding studio that turns a Qwen model into a private, transparent partner.

Private, local LLM-powered coding companion.

OmniScientist: autonomous AI agent that conducts end-to-end scientific research

Turns raw data into research papers automatically

Electron desktop wrapper for DeepSeek Harness, enabling a local AI workspace

Native desktop UI for LLM development

MADE: a Rust engine for structured multi-agent deliberations

Provides observable, contract‑driven AI coordination

Open-source harness for building AI agents with a plugin architecture

Modular plugin system for AI agents

Session‑isolated personalization memory plugin for DeepSeek LLM

Editable, isolated memory for consistent AI chats

FastAPI service chaining three CrewAI agents for compliant research answers

OpenAI‑compatible API with built‑in compliance checks

FastAPI service exposing a LangGraph ReAct agent via OpenAI API

OpenAI‑compatible endpoint for custom AI agents

Open-source Tauri runtime for desktop AI agents

Local, secure, controllable AI agent execution

Stele Bench: reproducible benchmark suite for coding-agent memory evaluation

Standardized, reproducible tests for agent memory

Lanes: coordination layer for AI agents to avoid duplicate work

Adds visibility, prevents redundant AI effort

36 machine-checked agent skills for ML and AI consulting with a rigorous conformance protocol

Rigorous conformance protocol with numbered gates, executable checkers, and structured delivery blocks

CLOVA: Closed‑Loop Visual Assistant for tool usage and updates

Combines LLM reasoning with visual tool execution

Explainable AI football match predictor with chain‑of‑thought reasoning

Provides transparent predictions with data‑driven reasoning

Forgewright: a web workbench for designing, analyzing, and refining LLM prompts

Discovery, scaffold editing, and reverse analysis in one UI

AI prompt engineering tool that turns vague goals into effective, verifiable prompts or stops when AI is not the right tool

Solves common prompting failures with a thoughtful, multi-phase approach

Evidence-first workflow kit for auditable AI deliverables

Enforces source-backed claims

VERA: sandbox that cryptographically verifies AI agent task completion

Stops AI agents from cheating with false success claims

Codi Agent: Korean AI coding companion that evolves via conversation

Provides multi‑platform Korean AI coding assistance

MedScribe: AI‑driven OCR and LLM tool to extract and structure medical records.

Combines OCR and LLM for quick, editable medical data extraction.

Skill Framework for Claude/Codex agents – routing, hooks, and tools

Rust‑based routing & hook system for AI agents

Slack integration library for Waken AI agents

Real‑time Slack source & output for AI workflows

Harbor: a local‑first personal agent framework for chat and task automation

Modular queue adapters and workers for home‑server agents

LangGraph-based AI agent framework with sandbox execution

Manages sandbox lifecycle, integrates web search & file upload

Biomedical Agent Teams Codex Marketplace plugin for local Codex Desktop

Exposes biomedical-agent-teams skill with 35 agent prompts and more

AI Studio app for running and deploying AI models locally

Simplifies local deployment of AI models with Gemini API integration

AI Studio app for running AI models locally

Simplifies local AI app development

AI Studio app for remote healthcare management

Implements AI-powered healthcare management locally

Web UI for RantaiClaw AI agent, enabling chat and ops via browser

Ready-to-run Next.js UI for self‑hosting AI agents

WordPress chatbot plugin with AI integration, admin panel, and usage analytics

All-in-one solution for AI chatbots with multiple provider support and detailed analytics

AI Agent platform with FastAPI & Gradio for multi‑tool interactions

Combines LlamaIndex agents, REST API, and web UI

LangGraph-powered content generator using Google Gemini

Self‑correcting LLM pipeline with evaluation loop

Proxy server that routes Claude Code API traffic to many AI providers

Lets Claude Code use free, local or alternative models

An innovative approach to agentic LLM systems.

Demonstrates practical implementation of AI-driven decision-making.

AI agent for scientific calculations using local LLMs

Functional prototype with actual code, not just educational demo

A self-hosted temporal memory platform for AI agents and apps

Sophisticated multi-tier memory with evidence attribution

On-device semantic search and RAG for any folder

Local, private, low‑power AI search

A web console to explore and manage Supermemory's AI knowledge base

Live UI with mock backend for memory inspection

Local RAG indexer that guarantees every chunk is stored

Verifies coverage to avoid silent indexing failures

LLM-powered web app for natural-language Q&A on YouTube channel data

Secure LLM queries via Cloudflare Worker, static React UI

DeepResearch Agent: LLM‑powered research assistant with citation checking

End‑to‑end research, retrieval, and citation verification

Self‑hosted AI‑powered markdown knowledge base with hybrid search

Combines vector + keyword search with chat UI

Local RAG‑powered clinical assistant with Streamlit UI

Searchable medical evidence with AI analysis

Reusable website AI chatbot platform with Laravel backend and Next.js front‑end

Ready‑to‑run knowledge‑grounded chatbot solution

Hermes skill-graph plugin adds searchable knowledge graph for agent skills

Typed relationships improve skill discovery

TrueFoundry‑focused RAG server for answering docs queries

Provides grounded answers with source URLs

Local GPU-accelerated pipeline that extracts policy text from PDFs

Runs entirely offline, no cloud APIs needed

Financial PDF extractor using LLMs and embeddings

End‑to‑end pipeline for structured data

Local RAG platform for private research docs

Hybrid BM25+semantic search, no cloud

AI chatbot with real‑time RAG for PDFs and docs

FastAPI, WebSocket, FAISS, Groq

AI PDF Q&A app that answers questions from uploaded PDFs

Combines RAG with Gemini for instant answers

PHANTOM: Zero-copy multi-agent LLM KV-cache serving on Apple Silicon Unified Memory

Optimizes LLM inference by sharing KV-cache across agents without copying

Native macOS app bundling Open WebUI for offline AI with Ollama

Ready‑to‑run offline LLM app for macOS

Lightweight CLI distribution of DeepSeek Harness for local LLM execution

Small inspectable runtime for local DeepSeek LLM use

On-device perception‑to‑reasoning pipeline for Apple Silicon

Combines Core ML vision with swappable LLMs for offline AI

Terminal AI assistant for Northeast Indian languages, runs locally

Local LLM with language ID and vector search

Admission control & idle model reaper for shared local LLM servers

Manages memory, priorities, and external tenants

DeepSee adds vision and smart model routing to DeepSeek Harness

Combines Gemini vision with DeepSeek for seamless coding

DSH Vision Toolkit adds native vision capabilities to DeepSeek agents

Integrates image Q&A, OCR, UI restoration

Vision MCP: Windows tool giving AI agents visual analysis & generation

Integrates image/video AI tools into Windows agents

B2-backed pipeline that tiles, segments, and embeds whole‑slide pathology images

Runs CLAM‑style feature extraction on CPU, storing all data in B2

Multi-camera pedestrian detection, tracking and re-identification system

End‑to‑end CV pipeline with API and UI

KayzeeOCR: layout detection and structured OCR using Qwen2.5-VL-2B

End‑to‑end OCR with JSON output for 17 document element types

Swift OCR microservice using Vision for on-device text extraction

Self‑hosted macOS OCR with API, metrics, and graceful shutdown

AI-powered video analysis pipeline for local object detection and scene understanding

Combines OWLv2 and local LLM for adaptable video analysis without external AI services

AI chest disease detector using X-rays with Streamlit web interface

Complete working solution with pre-trained model and web UI

Smart traffic routing system

Solves traffic congestion

Offline speech-to-text and speaker diarization pipeline

Combines Vosk and SpeechBrain for batch audio processing

Deterministic, explainable multilingual translation via numeric ontology

Provides traceable semantic matches across languages

Swahili TTS pipeline for data prep, baseline evaluation, and fine‑tuning

End‑to‑end workflow for building Swahili speech synthesis models

Turkish TTS model

Working TTS model

Evergreen Research: pipeline that fetches, structures, and digests AI papers weekly.

Automates AI paper ingestion and digest.

YOLO label recovery tool that audits and auto-fixes missing annotations

Model‑free, multi‑teacher pipeline for dataset quality

Dependency‑free TypeScript library for calibrated LLM evaluation certificates

Statistical calibration, abstention thresholds, and drift monitoring in one package

Per-step Integrated Gradients analysis for diffusion-policy robotics models

Enables detailed attribution of vision, language, state inputs

Fast, cheap omni‑modality model serving for text, image, audio, video and more

Extends vLLM for multi‑modal inference

Open-source LLM evaluation harness with sovereign backend and plugin architecture

Pluggable plugins and self‑hosted backends

JudgeShift CX: evaluates bilingual LLM judges for customer‑support replies

real pipeline with dataset, tests, and dashboard

LoRA Merger: compose specialist adapters into a single model for business workflows

Enables deterministic merging of LoRAs for real‑world tasks

Open research platform forecasting salmon‑lice risk in Norwegian aquaculture

Data pipelines, Bayesian models, API, and map UI

Desktop GUI for TabICL to train and forecast tabular models

Easy visual interface for TabICL with LLM constraints

Node.js SDK for AI agent governance and audit trails

Enables policy‑enforced, tamper‑evident AI runs

Benchmark for evaluating event routing systems that decide which source events should wake downstream AI agents

First practical benchmark for comparing event-driven agent architectures on the same data

Real estate price predictor with ML API and web UI

End‑to‑end Azure‑deployed ML model with explainability

Framework for evaluating multi‑region LLM agent handoff performance

Runs realistic distributed LLM experiments with Docker

ZINA detects and edits hallucinated text in multimodal LLM outputs

Fine-grained hallucination tagging and correction

End‑to‑end GNN traffic forecasting on PeMS sensor data

Full pipeline, multi‑GPU training, reproducible results

PyTorch LoHA adapter for LLM fine‑tuning with CLI and full pipeline

Works without PEFT, end‑to‑end fine‑tuning

Elliptical Basis Function interpolation library for smooth fits to scattered data

Learns per-node anisotropic distance metrics

From‑scratch reverse‑mode autodiff engine (scalar & NumPy tensor) with optimizers and experiments

Educational yet functional autodiff toolkit

Contrastive encoder trainer for Llemma‑7B on Lean proof pairs

Ready‑to‑run pipeline for GPU/Apple‑Silicon training

JEPA-based world-model agent scaffold for AI research

End‑to‑end training and evaluation of a JEPA world model

Deletion‑aware RNA nanopore basecalling with student model training

Training and evaluation scripts for accurate RNA basecalling

PersonaDrive: controllable trajectory prediction with multi-dimensional driving personas

Adds persona conditioning to trajectory models

AlphaZero‑style micro‑chess engine with beam‑search inference

Runs a compact GPT‑2 model with search for strong play

UAV multimodal trajectory prediction framework for ProjectAirSim

Provides ready‑to‑use 3D trajectory forecasting

Physics-Informed Neural Network training with precision and optimizer switching

Enables efficient, accurate PINN experiments

SymQNet: fast adaptive Hamiltonian learning via amortized acquisition

RL‑driven VAE speeds quantum system identification

iSCAT‑based NMC battery degradation analysis pipeline

Comprehensive scripts for optical‑electrochemical data processing

U-NSGA-III .NET library for multi‑objective evolutionary optimization

Native C# MOEA with validated performance

Dutch electricity load forecasting app with backtested, calibrated predictions

Real data, transparent backtest, confidence intervals

VRAM‑safe timeline director for MiniMax H3 video generation in ComfyUI

Enables long clips on 8 GB GPUs without OOM

WebRTC‑enabled surgical robot simulator powered by Cosmos‑3 AI

Runs action‑conditioned AI model for realistic surgery demos

Bayesian lag selection tool for VAR models

Implements thesis with CmdStan integration

Reproducible benchmark for forecasting U.S. honey‑bee colony loss

Data, code, and leaderboard for honest out‑of‑sample testing

IDP design tool

Physics-guided

FastAPI service that ranks place recommendations using a scikit‑learn model

Ready‑to‑run ranking API with synthetic data and tests

GNN-driven Monte Carlo framework for magnetic alloy simulations

Combines graph neural nets with MC to study ordering

EEGlas: an explainable EEG dashboard for dementia classification

Combines explainable AI with an interactive UI for clinicians

Diagnostic toolkit for GP boundary bias in Bayesian optimisation

Shows how variance inflation skews acquisition

A Nexus agent plugin that offers grid‑based A*, RRT, and multi‑agent path planning with obstacle handling

Rich algorithm suite in a single, easy‑to‑integrate library