Guía de uso
Local AI
Enable Ollama, run the system check, download models, and use vision tagging + embeddings.
Stratum's intelligence is optional and fully local. Install Ollama, pull two models, and you get image tagging and search-by-meaning — with nothing leaving your machine.
What the AI does
| Capability | Model | Effect |
|---|---|---|
| Vision tagging | llava:7b | Looks at image assets and proposes descriptive tags. |
| Semantic search | nomic-embed-text | Embeds assets so search can match by meaning. |
Neither is required. Without AI, Stratum versions, browses, tags and searches by keyword exactly as before.
Installing Ollama
Download and install Ollama from ollama.com. Once it's running, Stratum can detect it. You don't need to keep a terminal open — Ollama runs as a background service.
The system check
Open Settings → Local AI and press Re-check. Stratum reports on each dependency:
| Item | Required | Meaning |
|---|---|---|
| Git | Yes | Core versioning — must be present. |
| Ollama | No | Enables local AI when running. |
| llava:7b | No | Vision tagging model. |
| nomic-embed-text | No | Embeddings model for semantic search. |
Downloading models
You can pull the models from inside Stratum, or from a terminal:
ollama pull llava:7b
ollama pull nomic-embed-textPrivacy
Every request goes to your own Ollama instance on localhost. Stratum sends nothing to Arkynox or any third party — not images, not queries, not usage data.
Turning it off
Disable local AI any time in Settings. Existing tags and versions stay; you simply lose vision tagging and semantic search until you re-enable it. Keyword search keeps working throughout.