finrod/README.md

finrod

Tiny web translator backed by Ollama on radagast. Named for Finrod Felagund, Noldor loremaster of the tongues of Men.

Run

uv run uvicorn app:app --host 127.0.0.1 --port 8000 --reload

Then open http://127.0.0.1:8000.

Config

Env var	Default	Notes
OLLAMA_URL	http://radagast.jpnadas.xyz:11434	Ollama base URL
MODEL_FILTER	gemma,aya	Comma-separated substrings; model matches if it contains any. Empty = show all

Models

For Pt/Nl/En translation, three worth A/B-ing:

ssh radagast 'ollama pull gemma2:9b'      # already there — big-and-broad baseline
ssh radagast 'ollama pull gemma3:4b'      # newer multilingual training, ~2-3× faster on Ivy Bridge
ssh radagast 'ollama pull aya-expanse:8b' # Cohere multilingual model, strong on Dutch

All three fit in radagast's 10 GB at Q4. Switching models in the dropdown forces Ollama to load the new weights — expect a 5-15 s pause on the first call after a switch while it reads from disk.

Build / push image

Built locally and pushed to Scaleway — no Gitea runners on the homelab yet, so there's no CI workflow.

One-time setup (host needs buildx + QEMU for cross-arch):

docker buildx create --name multi --use
docker run --privileged --rm tonistiigi/binfmt --install all
docker login rg.nl-ams.scw.cloud -u nologin -p <SCW_SECRET_KEY>

Then on every change:

./build.sh

That builds linux/amd64,linux/arm64 and pushes rg.nl-ams.scw.cloud/valinor/finrod:{latest, <git-short-sha>}. Take note of the short SHA it prints — that's what you pin in valinor.

Deploy

Cluster manifests live in valinor/apps/finrod/ (bjw-s app-template, internal-only ingress at finrod.jpnadas.xyz). To roll out a new build:

1. ./build.sh here — get the SHA
2. Bump tag: in valinor/apps/finrod/values.yaml to that SHA on a branch
3. Merge → ArgoCD syncs the new image

Using :latest works for the first deploy (the pull policy is Always) but pinning to a SHA on each rollout gives you trivial rollback via git revert.