Ready now: Ollama on this machine with qwen3:8b (multilingual MT).
Zen default provider is offline · Ollama — no MyMemory quota, no CF required for local pad.
| Path | State | Zen provider |
|---|---|---|
Ollama qwen3:8b | … | offline / ollama |
| Ollama other (mistral, deepseek…) | … | change model in code / CLI |
qbitOS m2m100-1.2b/ | empty (optional HF) | use edge or ollama |
qbitOS llama-3.1-8b…/ | empty → use qwen3 | ollama |
| Cloudflare m2m100 + Llama | online edge | edge |
# pad (local paper or prod — Ollama must be on this machine) open 'http://127.0.0.1:8790/zen.html?provider=offline&test=short' open 'https://lang.ugrad.ai/zen.html?provider=offline&to=fr,es,de,uk,ar,ja,zh' # CLI cd scripts/language/lang-ugrad-local ./cli/lang-pipe ai --provider ollama --action translate --text "all clear" --target uk # keep Ollama up ollama serve # already pulled: qwen3:8b · mistral · deepseek-r1:7b · …
Not required if Ollama is enough. To fill empty slots later:
bash /Volumes/qbitOS/03.models/07-lang-ugrad/scripts/pull-hints.sh # place GGUF/HF under: # /Volumes/qbitOS/03.models/07-lang-ugrad/m2m100-1.2b/ # /Volumes/qbitOS/03.models/07-lang-ugrad/llama-3.1-8b-instruct-fp8/