Local AI Operations Assistant Setup
Budget: $250 – $750 NZD
My small creative production studio needs a local, fully private AI stack that will sit on a 16 GB Mac Mini and behave like an operations assistant. I have already chosen the ai-stack-homelab project as the base; it ships with Dockerised n8n, Ollama (running Llama 3.2-class models), Open WebUI, PostgreSQL + pgvector, LiteLLM proxy and MCP integration. Your job is to install, configure and fine-tune that stack so it automates day-to-day admin without ever touching the public cloud.
Immediate focus
• Highest-priority workflow: detect inbound enquiries arriving via email, run them through the local LLM, then log each enquiry as a structured page inside my Notion space. No automatic replies for now—just clean, reliable capture so I can review and follow up manually.
• While you are wiring that, prepare the ground for later extensions such as proposal generation, email summarisation and calendar event management.
Deliverables
1. Docker compose or similar script that spins up the whole ai-stack-homelab environment on macOS and persists data volumes correctly.
2. n8n workflow(s) that:
– monitor the chosen email inbox,
– route new enquiries through Ollama,
– create or update a Notion database entry with parsed details and embeddings in PostgreSQL for future search.
3. Clear environment variables / secrets management so the system survives reboots and OS updates.
4. Brief README (Markdown) explaining how to start, stop, back up and extend the stack.
Acceptance criteria
• All services launch with a single command and remain local to the machine.
• A test email generates a proper Notion page within 60 seconds and no data leaves the network.
• CPU/RAM footprint stays within reasonable bounds for a 16 GB Mac Mini.
If you have hands-on experience with n8n, Notion API, Docker on macOS and running LLMs locally, I’d love to get this humming quickly so I can spend less time on admin and more on creative work.
Immediate focus
• Highest-priority workflow: detect inbound enquiries arriving via email, run them through the local LLM, then log each enquiry as a structured page inside my Notion space. No automatic replies for now—just clean, reliable capture so I can review and follow up manually.
• While you are wiring that, prepare the ground for later extensions such as proposal generation, email summarisation and calendar event management.
Deliverables
1. Docker compose or similar script that spins up the whole ai-stack-homelab environment on macOS and persists data volumes correctly.
2. n8n workflow(s) that:
– monitor the chosen email inbox,
– route new enquiries through Ollama,
– create or update a Notion database entry with parsed details and embeddings in PostgreSQL for future search.
3. Clear environment variables / secrets management so the system survives reboots and OS updates.
4. Brief README (Markdown) explaining how to start, stop, back up and extend the stack.
Acceptance criteria
• All services launch with a single command and remain local to the machine.
• A test email generates a proper Notion page within 60 seconds and no data leaves the network.
• CPU/RAM footprint stays within reasonable bounds for a 16 GB Mac Mini.
If you have hands-on experience with n8n, Notion API, Docker on macOS and running LLMs locally, I’d love to get this humming quickly so I can spend less time on admin and more on creative work.