model-gear 0.6.0

model-gear — run, assess, and switch the local vLLM model.

model-gear

model-gear is the tooling that runs, assesses, and switches the local, OpenAI-compatible vLLM model the Culture mesh consumes. The binary is model — model switch, model assess, model serve, and so on.

The served model is what the lepenseur agent connects to over the acp vllm-local provider. model-gear runs the engine; lepenseur is one consumer of it.

Sibling to culture (the agent mesh), daria (awareness), and steward (alignment).

Install

uv tool install model-gear

Usage

model init --apply          # scaffold a deployment dir (default ~/.model-gear)
model serve --apply         # start the vLLM server (alias: start)
model switch nvidia/Qwen3-32B-NVFP4 --apply   # switch the served model
model status                # current model, container state, /health
model assess                # correctness probes (markdown for a per-model doc)
model benchmark             # decode throughput + prefill latency
model stop --apply          # stop the server

model overview              # tool snapshot + served model + candidate list
model whoami                # tool, machine, served model, container health
model explain switch        # markdown docs for a topic
model doctor                # diagnose docker / compose / .env / health

Every command supports --json. Write verbs (switch, serve, stop, init) are dry-run by default and require --apply to commit — agents call CLIs in loops, so safe-by-default is mandatory.

Running the model locally (vLLM)

model init scaffolds a deployment directory (default ~/.model-gear) from the packaged templates: a docker-compose.yml that stands up the vLLM model as an OpenAI-compatible server on :8000, plus a .env. Tuned for DGX Spark (GB10 Grace Blackwell, 128 GB unified memory) per build.nvidia.com/spark/vllm.

Prerequisites: the NVIDIA Container Toolkit, and docker login nvcr.io with an NGC API key to pull the nvcr.io/nvidia/vllm image.

model init --apply          # writes ~/.model-gear/{docker-compose.yml,.env}
# edit ~/.model-gear/.env to set HF_TOKEN if the model repo is gated
model serve --apply         # first run downloads ~18 GB of weights
model status                # waits/reports until /health is up

Verify it is up:

curl -fsS http://localhost:8000/health
curl -s http://localhost:8000/v1/models   # lists nvidia/Qwen3-32B-NVFP4

Tunables live in the deployment .env (VLLM_MODEL, VLLM_GPU_MEM_UTIL, VLLM_MAX_MODEL_LEN, HF_CACHE, …). VLLM_SERVED_NAME must match the part after vllm-local/ in culture.yaml — model doctor checks this. model switch rewrites these keys for you.

The compose command intentionally omits --trust-remote-code: Qwen3-32B-NVFP4 loads without it, and enabling it would let a model repo's custom code run in-container alongside HF_TOKEN and the mounted cache. Add it back only for a model whose repo ships custom modeling code. If vLLM rejects the nvidia/ ModelOpt checkpoint, set VLLM_MODEL to the vLLM-native RedHatAI/Qwen3-32B-NVFP4 and drop --quantization from the compose command.

Per-model notes

Each runtime model has a doc under docs/ recording how to run it, live test results, and caveats:

• docs/qwen3-32b-nvfp4.md — the current runtime model (nvidia/Qwen3-32B-NVFP4), benchmarked on DGX Spark.
• docs/qwen3.6-27b-nvfp4.md — a candidate (mmangkad/Qwen3.6-27B-NVFP4), load-tested on DGX Spark; loads under the current vLLM image but is slower on decode, so the 32B stays.

The numbers in each doc come from model switch <model> --apply then model assess (correctness) and model benchmark (throughput). model overview --list lists these docs and flags which model is currently served.

lepenseur still gets deployed

The mesh agent served by this model is lepenseur ("le penseur" — the thinker), a local thinking agent. Its runtime identity lives in AGENTS.md (the acp system prompt) and culture.yaml (backend: acp, model: vllm-local/nvidia/Qwen3-32B-NVFP4). model-gear is the repo identity and the tool; lepenseur is the agent that consumes the model model-gear serves.