nova-local

command
v1.2.5 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Oct 10, 2026 License: MIT Imports: 19 Imported by: 0

README

nova-local

What it is

nova-local: run local models: what an engine has, one model served at a chosen context, and a worker description nova-swarm accepts

Why use it

Run a model on your own machine, or on one of your fleet.

Install

go install github.com/mas-bandwidth/nova-tools/cmd/nova-local@latest
nova-local version

First run

The transcript below is executed line for line by the docs tests; the full record is in docs/TESTS.md.

Fixture: cmd/nova-local/testdata/: the worker directory (home/) and the placeholder key file (local.key). cmd/nova-local/firstrun_test.go runs each $ line from a checkout root in one sitting against a fake ollama daemon that has gemma4:12b in the shared store of the AI root /ai, a box at load 14.2 with 61 GiB free of 128, and a clock the warm-up advances by three seconds; no engine, network or real time is used. ./gemma.json is a file in the test's own directory.

$ nova-local status
STATUS OK engines=1 answering=1 loaded=0 models=1 mem_used=71940702208 mem_free=65498251264 mem_total=137438953472 wired_cap=unset load1=14.20
STATUS ENGINE name=ollama state=up base=http://127.0.0.1:11434/v1 loaded=0 advertised=1 store=/ai/shared/models/ollama shared=yes

$ nova-local serve --engine ollama --model gemma4:12b --num-ctx 32768 --seed 7
SERVE OK engine=ollama model=gemma4:12b serve_as=gemma4-32k digest=sha256:c0e0c3e5b4a1 num_ctx=32768 keep_alive=30m temperature=0 seed=7 created=yes load=3s mem_used=71940702208 mem_free=65498251264 mem_total=137438953472 wired_cap=unset load1=14.20 engines=1 store=/ai/shared/models/ollama shared=yes

$ nova-local worker --engine ollama --model gemma4-32k --out ./gemma.json --name gemma --harness opencode --harness-args run,--model,ollama/{model},--,{prompt} --worker-dir $PWD/cmd/nova-local/testdata/home --key-file $PWD/cmd/nova-local/testdata/local.key --env-var OLLAMA_API_KEY --usage opencode --deadline 20m
WORKER OK engine=ollama model=gemma4-32k out=./gemma.json workers=1 provider=ollama harness=opencode deadline=20m base=http://127.0.0.1:11434/v1
WORKER NOTE run it one worker at a time (--workers 1): an engine is a queue, and two workers interleave it

Verbs

The nova-local section of the command reference documents every verb's flags, effect and exit codes.

  • status
  • serve
  • worker
  • version
  • help

Spec

The contract is docs/SPEC-LOCAL.md.

Documentation

Overview

nova-local makes a local model engine usable (docs/SPEC-LOCAL.md): what is here (status), one model served at a context the caller chose (serve), and a worker description nova-swarm accepts (worker). It runs no inference, fetches no weights and judges no model. The dispatch, the banner, the help, the refusals and the output envelope are pkg/tool's; the engines (engine.go, ollama.go), the box (box*.go), the shared store (store.go), the serving-host rule (host.go) and the worker description (description.go) are this package's own files.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL