Hey everyone! I'm Slava (@slavasolodkiy) — just got my Olares One from the Kickstarter batch and spent the past two weeks turning it into a personal AI hub. I'm not a developer or data scientist — just someone who wanted local AI without cloud dependencies. Sharing my experience in case it helps other newcomers (and maybe gives the team some useful feedback).
+Obsidian — knowledge management across all devices, synced via LiveSync on Olares
Turned it on and got a black Ubuntu terminal. Briefly panicked, but it's the normal first-boot experience.
Things that tripped me up:
Running daily:
| App | What I Use It For |
|---|---|
| Ollama | LLM inference engine — serves all my models |
| Open WebUI | Main chat interface, model management, RAG |
| LiteLLM | API proxy — the key to remote access (more below) |
| AnythingLLM | Document Q&A with RAG over my Obsidian vault |
| ComfyUI | Image generation (Stable Diffusion workflows) |
| SearXNG | Private metasearch engine |
| Obsidian LiveSync | CouchDB-based vault sync between all my devices |
Also installed but stopped: Dify, LibreChat, n8n, Perplexica, PhotoPrism, qBittorrent, OnlyOffice, Firefox, Chromium, Windows VM.
Currently running 15 models in Ollama:
| Model | Size | Role |
|---|---|---|
| qwen2.5-coder:32b | 19 GB | Code generation (excellent) |
| qwq:latest | 19 GB | Chain-of-thought reasoning |
| gemma3:27b | ~17 GB | General-purpose powerhouse |
| command-r:35b | ~20 GB | RAG-optimized (Cohere) |
| mistral-small:latest | 14 GB | Multilingual workhorse |
| deepseek-r1:14b | ~9 GB | Fast reasoning |
| phi4:14b | ~9 GB | Summarization, structured output |
| qwen3:8b | 5.2 GB | General chat |
| gemma-3-4b-it | 2.5 GB | Quick tasks, translations |
| nomic-embed-text | 274 MB | Embeddings for RAG |
| + 5 uncensored/experimental models | various | Creative, research |
GPU Management tip: I initially set App Exclusive mode and Ollama started crashing with "Insufficient video memory." Switching back to Time Slicing fixed everything.