[mbreiser]

Summary

rig/pixi.toml reads the rig-nodes channel at the v0.4.0 release tag (https://raw.githubusercontent.com/reiserlab/rig-nodes/v0.4.0/channel; 0.3.0 was pinned by the payload commit's SHA only because its tag had landed one commit early) and pins rig_nodes / rig_interfaces ==0.4.0. Re-locked with pixi 0.83 (pixi lock in rig/); the dev environment keeps its git sources on rig-nodes main, now pinned to fa83d39. The mutex 0.20 records in the lock are unchanged (6 before, 6 after; they belong to the dev source-build records, not to an installed environment).

rig-nodes 0.4.0 vs 0.3.0 (reiserlab/rig-nodes#7, tag v0.4.0 on payload commit 6aad10a, CI run 40800724499 green on win-64 / osx-arm64 / linux-64 incl. the clean-directory acceptance):

• the ScanImage JSON connector moved to the private 2P rig repo (rig-nodes#4; a contract change, hence the minor bump) — rig/nodes/rig.py already lost --scanimage / --si-mock in Imaging boundary: the 2P imaging work moves to the private rig repo; gateway keeps the 2P topics (LAB-303, LAB-319) #284, and grep scanimage_node rig/ finds only the emulator's comment;
• host monitor: typed HostStats floats quantized to power-of-two steps (compressible), portable process sampling reads the whole table every 5 s and only watched pids in between (rig-nodes#6);
• no .msg change.

Docs: rig/WINDOWS-TEST.md expected harness count → 60 / 60 (63 / 63 typed), rig/README.md adapters list, the okey-cokey dashboard plan's tiff_run_to_mcap.py line (private since 2026-10-28), the rig-nodes dependency comment in rig/pixi.toml.

Tier

next candidate — together with #285 (pixi run qc), Michael decides the set.

Release notes

none

Test plan

Mac, pixi 0.83.0, with ~/Documents/GitHub/rig-nodes renamed away for the whole run:

• cd rig && pixi install --locked — 4 s, rig_nodes 0.4.0 pyh4616a5c_0 from the tag URL, ros2-distro-mutex 0.19.0 jazzy_26, Python 3.12.16
• pixi run test — 60 / 60; host monitor self_cpu_median 0.006 of a core (self_cpu_max 0.066), /rig/host 3.2 % of the compressed payload — the macOS cost gate (≤ 2 %) that the 0.3.0 sampler failed on this Mac now passes
• pixi install --locked -e typed + pixi run -e typed check-typed — rig_interfaces 0.4.0 h60d57d3_0, usable + serializes, msgs: ArenaFrame, ConnectorReply, DiskStats, FicTracFrame, GpuStats, HostStats, NetStats, ProcStats
• pixi run -e typed test — 63 / 63, typed topics rig_interfaces/msg/ArenaFrame + HostStats, self_cpu_median 0.0049
• a plain pixi install afterwards leaves git status rig/ with only this PR's files (lock as pixi lock wrote it)
• gateway: pixi run rig + pixi run studio-emulator -- --scanimage-cmds — hello_ack advertises the scanimage connector PASS; the ten node-dependent ScanImage checks (ping / setFileStem / grab / …) get no reply because no ScanImage node exists in this graph any more (private repo), as on main since Imaging boundary: the 2P imaging work moves to the private rig repo; gateway keeps the 2P topics (LAB-303, LAB-319) #284 — the gateway's /scanimage/* registration itself is intact (pixi run test 60 / 60 includes test_gateway_registration's kept topics)
• ww14 / ww18: pixi install --locked from the tag URL on the candidate

🤖 Generated with Claude Code