jetson-arena 0.5.0

Agent and CLI for the Jetson Arena: how a given Jetson device handles a whole model stack. Benchmarks pipelines (e.g. realtime VAD + STT + LLM + TTS on an Orin Nano Super 8GB) for latency, memory signature and quality, then publishes results plus signed-in Docker build recipes. Also maintains the arena public site.

jetson-arena

How does a given Jetson device handle a whole model stack?

Not one model in isolation — the whole pipeline. A realtime VAD → STT → LLM → TTS loop puts four models on one 8GB Orin Nano Super, where they contend for VRAM, memory bandwidth, and the GPU. Each fits alone; the stack is what fails. jetson-arena benchmarks the stack for latency, memory signature, and quality, then publishes the results together with the Docker build recipes that reproduce them — and maintains the arena's public site.

Status

Early. The domain surface is not built yet. What exists today is the agent-first CLI scaffold below, plus the mesh identity and the CI baseline — there is no benchmark runner, device probe, recipe store, or site. The intended shape, and the questions still open, are written down in docs/scope.md. Read that before proposing a design.

Quickstart

uv sync
uv run pytest -n auto                 # run the test suite
uv run jetson-arena whoami            # identity from culture.yaml
uv run jetson-arena learn             # self-teaching prompt (add --json)
uv run teken cli doctor . --strict    # the agent-first rubric gate CI runs

CLI

Verb	What it does
whoami	Report this agent's nick, version, backend, and model from culture.yaml.
learn	Print a structured self-teaching prompt.
explain <path>	Markdown docs for any noun/verb path.
overview	Read-only descriptive snapshot of the agent.
doctor	Check the agent-identity invariants (prompt-file-present, backend-consistency).
cli overview	Describe the CLI surface itself.

Every command supports --json. Results go to stdout, errors/diagnostics to stderr (never mixed). Exit codes: 0 success, 1 user error, 2 environment error, 3+ reserved.

How it is put together

• An agent-first CLI cited from teken (afi-cli) — the runtime package has no third-party dependencies, and the contracts above are enforced in CI by a rubric gate that drives the built CLI as a black box.
• A mesh identity — culture.yaml (suffix + backend) and the matching resident prompt file (AGENTS.colleague.md, since this agent runs backend: colleague). CLAUDE.md is the parallel prompt for Claude Code.
• The guildmaster skill kit (14 skills) under .claude/skills/, vendored cite-don't-import. Provenance ledger: docs/skill-sources.md.
• A build + deploy baseline — pytest, lint, the agent-first rubric gate, and PyPI Trusted Publishing wired into GitHub Actions.

Contributing

Every PR bumps the version — docs-only ones included. CI enforces it, because a merge to main publishes to PyPI and a repeated version is a failed publish. See CLAUDE.md for the full conventions: the CLI contract, the vendored-skills rule, and the cicd PR lane.

License

Apache 2.0 — see LICENSE.