godM0D3 – Modular Self-Hosted AI Plugin Platform

THE IDEA

godM0D3 – Modular Self-Hosted AI Plugin Platform

Type: Software Status: Beta Tech Stack: Python 3.11, FastAPI, SQLAlchemy 2 (async), Pydantic v2, PostgreSQL 16 (schema isolation + row-level security), Alembic, Celery 5, Redis 7, React 18, Vite 5, TypeScript, TailwindCSS, WebSockets, JWT, Fernet, Docker Compose, Nginx, n8n, Ollama/OpenAI/Anthropic/vLLM/LM Studio, faster-whisper, pytest, Vitest, Playwright Problem Statement Self-hosted productivity and AI tools are usually separate applications, each with its own login, its own database and its own maintenance; anyone wanting to bring an assistant, invoicing, email and a knowledge base together ends up with many disconnected services or with cloud offerings that pass data on to third parties. Privacy-conscious users such as SMEs and freelancers in particular lack an extensible platform that keeps its modules cleanly isolated from one another and also works with purely local language models. Description godM0D3 is a self-hosted AI work platform built on a FastAPI/React core that discovers, installs and manages full-stack “plugins” and back-end “skills” at runtime. Each plugin gets its own PostgreSQL schema and is additionally secured per user with enforced row-level security — together with the filter in the application, a threefold data isolation. It ships with 12 plugins (including an AI assistant with memory, Telegram and proactive reminders, multi-provider LLM chat across 5 providers, a knowledge base with keyword search, invoicing with PDF/DOCX export, an email client, audio transcription and a training platform with 1,200 bilingual exercises) and 10 skills (Telegram, email, web search, browser reader, RSS, PDF/DOCX, n8n integration and others). LLM tool calling runs through a central registry and also works with purely local models (Ollama, vLLM, LM Studio), so no data has to leave your own system. The stack is built with security in mind (short-lived httpOnly JWT, WebSocket tickets, encrypted credentials, strict checks of secrets and CORS in production, SSRF protection) and starts as a Docker Compose stack with a setup wizard. Technically the platform is close to production readiness (around 569 automated test cases, a 60 % coverage gate, Bandit and CodeQL), but according to its README it is explicitly intended for demonstration purposes only. Use Case Your own fully self-operated AI platform on which you install modules such as an assistant, invoicing, email, a knowledge base or transcription as if from an app store — without data or AI requests ever having to leave your own system. Link: https://github.com/rawk7000/godm0d3 (private repo)