THE MAH STACK

Guide to building with the MAH Stack

An open walkthrough of the AI tools, workflows and routines our members plug into their day. Five pillars, one brain.

System Architecture

    +---------------------------------------------------+
    |                  MELBOURNE AI HUB                   |
    |                  THE MAH STACK                      |
    +---------------------------------------------------+
    |                                                    |
    |   +----------+   +----------+   +----------+      |
    |   | CLAUDE   |   | KNOWLEDGE |   | PRODUCT  |      |
    |   | CODE     |   | VAULT    |   | (NOTION) |      |
    |   +----------+   +----------+   +----------+      |
    |        |              |              |            |
    |        +------+-------+------+-------+            |
    |               |              |                    |
    |   +----------+ |   +----------+                  |
    |   | SUPABASE |-+   | STRIPE   |                  |
    |   | (DB/Auth)|     | PAYMENTS |                  |
    |   +----------+     +----------+                  |
    |        |                            |            |
    |   +----------+              +----------+         |
    |   | OLLAMA   |              | AWS EC2  |         |
    |   | (LOCAL   |              | (HOSTING)|         |
    |   |  LLM)    |              +----------+         |
    |   +----------+                                  |
    |                                                    |
    +---------------------------------------------------+
    
    FIVE PILLARS:  CLAUDE CODE | KNOWLEDGE | PRODUCTIVITY
                   BACKEND | INFRASTRUCTURE

Five Pillars, One Brain

1. Claude Code (AI Assistant)

The primary AI co-pilot for code generation, research synthesis, and workflow automation. Members use it for everything from building apps to writing research proposals.

2. Knowledge Vault (Obsidian)

The Hub's collective memory. Notes, research, decisions, and learnings persist across sessions and across agents.

3. Productivity (Notion + Google Workspace)

Task tracking, document creation, calendar, email. The operational layer.

4. Backend (Supabase)

Postgres database, authentication, edge functions, file storage. RLS on every table. The data backbone.

5. Infrastructure (AWS EC2 + Ollama)

Nginx on EC2 (ap-southeast-2) for hosting. Ollama for local LLM inference (GLM models). Self-managed, no vendor lock-in on compute.

Telegram Gateway

The Hub runs a Telegram bot (@TeddyscleaningBot, rebranding to official MAH bot) that serves as the primary interface for community interaction, passive recording, and building management.

Melbourne AI Hub — 317/480 Collins Street, Melbourne, Victoria, Australia

Last modified: August 1990 · Best viewed in Netscape Navigator 1.0 at 640x480

You are visitor number 00000042 since 01-Jan-1990

[ Home ] · [ Contents ] · [ Email ] · [ Prev ] [ Random ] [ Next ] — Melbourne AI WebRing