<aside>
Purpose: Turn the Meetup #1 local AI discussion into a practical shared reference for Comox Valley AI builders, learners and organizers.
</aside>
Meetup #1 revealed a strong local technical base. Around seven people in the room were already running parts of a local AI stack. Others wanted to start, but needed a clear on ramp.
This guide creates a shared language for both groups.
| Layer | Plain language meaning | Tools mentioned | Questions to ask |
|---|---|---|---|
| Data | The files, notes, documents, media, code, logs and records you want AI to work with | Hard drives, home labs, Obsidian, file systems | What data is sensitive? Who owns it? Should it ever leave the machine? |
| Storage | Where the data lives | RAID, home lab storage, 160 TB local storage | Is it backed up? Is it encrypted? Can it be restored? |
| Indexing | Turning raw data into something searchable | PostgreSQL, Supabase, Weaviate | How will the system find the right material later? |
| Embeddings | Numerical representations of meaning that let the system compare concepts | Vector database plugins, large concept models | Do you need keyword search, semantic search, concept search, or all three? |
| Retrieval | Finding relevant chunks before the model answers | RAG, Supabase, Neo4j, Qdrant | What should the AI see before it answers? |
| Prompting | The instructions, constraints and context passed to the model | Open WebUI, Python, custom prompts | What role is the model playing? What should it avoid? |
| Inference | The model actually generating an answer | Ollama, LM Studio, Open WebUI, local models, cloud APIs | Should this run locally, in the cloud, or both? |
| Orchestration | Multiple agents, workers or tools coordinating work | Claude Code, MCP server, code factory, orchestrators, workers | What can the system do without you? What requires approval? |
| Evaluation | Checking whether the output is useful, accurate and safe | Custom evaluation scripts, human review | How do you know the answer is good enough? |
| Governance | Rules for access, logging, privacy, accountability and use | Sandbox machines, separate accounts, backups, permissions | Who can see outputs? Who is accountable if something goes wrong? |
Use this prompt:
I want to try running local AI on my computer. Ask me for my operating system, RAM, chip, GPU, storage and technical comfort level. Then recommend the safest beginner setup and model size.
Do not start with the largest model. Large parameter models need stronger hardware. Begin with a smaller model, confirm it works, then increase complexity.
Good first tests: